The following are some highlights of this patch:
- In the Worker we only extract a *subset* of the potential contents of the `Usage` dictionary, to avoid having to implement/test a bunch of code that'd be completely unused in the viewer.
- In order to still allow the user to *manually* override the default visible layers in the viewer, the viewable/printable state is purposely *not* enforced during initialization in the `OptionalContentConfig` constructor.
- Printing will now always use the *default* visible layers, rather than using the same state as the viewer (as was the case previously).
This ensures that the printing-output will correctly take the `Usage` dictionary into account, and in practice toggling of visible layers rarely seem to be necessary except in the viewer itself (if at all).[1]
---
[1] In the unlikely case that it'd ever be deemed necessary to support fine-grained control of optional content visibility during printing, some new (additional) UI would likely be needed to support that case.
Given that the Fetch API is supported since Node.js 18 we should be able to use it when downloading l10n files, which allows us to simplify the code and to make it fully `async`.
The fact that the highlight-thickness can only be changed in "free" mode isn't really obvious visually in the toolbar, so attempt to provide at least some indication of the `disabled`-state by "dimming" the slider.
In implementing caret browsing mode in pdf.js, I didn't notice that selectstart isn't always triggered.
So this patch removes the use of selectstart and rely only on selectionchange.
In order to simplify the selection management, the selection code is moved in the AnnotationUIManager:
- it simplifies the code;
- it allows to have only one listener for selectionchange instead of having one by visible page
for selectstart.
I had to add a delay in the integration tests for highlighting (there's a comment with an explanation),
it isn't really nice, but it's the only way I found and in real life there always is a delay between
press and release.
The function caretPositionFromPoint return the position within the last visible element
and sometimes there are some elements on top of the ones in the text layer.
So the idea is to hide the visible elements which aren't in the text layer in order
to get the right caret position.
*Please note:* This is a micro optimization, hence I fully understand if the patch is rejected.
Currently we create two temporary Arrays and have to iterate twice in total when building the final `hexNumbers` Array.
With this patch there's only one temporary Array and a single iteration required to build the final `hexNumbers` Array.
When highlighting, the annotation editor layer is disabled to get pointer events
from the text layer, but the annotation layer must be then disabled either in
order to avoid bad interactions.
Previously we'd simply export this directly from `web/app_options.js`, which meant that it'd be technically possible to *accidentally* modify the `compatibilityParams` Object when accessing it.
To avoid this we instead introduce a new `AppOptions`-method that is used to lookup data in `compatibilityParams`, which means that we no longer need to export this Object.
Based on these changes, it's now possible to simplify some existing code in `AppOptions` by taking full advantage of the nullish coalescing (`??`) operator.
Given that the "PREFERENCE" kind is used e.g. to generate the preference-list for the Firefox PDF Viewer, those options need to be carefully validated.
With this patch we'll now check this unconditionally in development mode, during testing, and when creating the preferences in the gulpfile.
As part of the changes in PR 17686 we "accidentally" enabled source-maps for the *minified* builds, which seems unnecessary since those have never been included in the `pdfjs-dist` output.
Locally this patch reduces the run-time of `gulp minified` by ~15 percent.
Rather than first building the library and then use Terser "manually" to minify the files, we can utilize a Webpack plugin to combine these steps which helps to simplify the gulpfile.
The `handler` method contained this code in two inline functions,
triggered via callbacks, which made the `handler` method big and harder
to read. Moreover, this code relied on variables from the outer scope,
which made it harder to reason about because the inputs and outputs
weren't easily visible.
This commit fixes the problems by extracting the request checking code
into a dedicated private method, and modernizing it to use e.g. `const`/
`let` instead of `var` and using template strings. The logic is now
self-contained in a single method that can be read from top to bottom
without callbacks and with comments annotating each check/section.
- Run the minification in "parallel" since that should be a *tiny* bit more efficient.
- Don't rename the minified files since that seems unnecessary, especially considering that they are only used in the `dist-pre` target where we currently change the name back manually.
After the changes in PR 17637 there's no longer any reason to invoke `tweakWebpackOutput` without an argument, since the `__non_webpack_import__` re-writing was moved into the Babel plugin.
This way we can avoid a (little) bit of unnecessary parsing during building.
The specs are unclear about what kind of xref table format must be used.
In checking the validity of some pdfs in the preflight tool from Acrobat
we can guess that having the same format is the correct way to do.
The pdf in the mentioned bug, after having been changed, wasn't correctly
displayed in neither Chrome nor Acrobat: it's now fixed.
Note how we're using custom `__non_webpack_import__`-calls in the code-base, that we replace during the post-processing stage of the build, to be able to write `import`-calls that Webpack will leave alone during parsing.
This work-around is necessary since we let Babel discards all comments, given that we generally don't need/want them in the builds, hence why we cannot utilize `/* webpackIgnore: true */`-comments in the source-code.
After the changes in PR 17563 it thus seems to me that we should be able to just move this re-writing into the Babel plugin instead.
The `handler` method contained this code in an inline function, which
made the `handler` method big and harder to read. Moreover, this code
relied on variables from the outer scope, which made it harder to reason
about because the inputs and outputs weren't easily visible.
This commit fixes the problems by extracting the directory listing code
into a dedicated private method, and modernizing it to use e.g. `const`/
`let` instead of `var` and using template strings.
The `handler` method contained this code in an inline function, which
made the `handler` method big and harder to read. Moreover, this code
relied on variables from the outer scope, which made it harder to reason
about because the inputs and outputs weren't easily visible.
This commit fixes the problems by extracting the range file serving code
into a dedicated private method, and modernizing it to use e.g. `const`/
`let` instead of `var` and using template strings.
The `handler` method contained this code in an inline function, which
made the `handler` method big and harder to read. Moreover, this code
relied on variables from the outer scope, which made it harder to reason
about because the inputs and outputs weren't easily visible.
This commit fixes the problems by extracting the file serving code into
a dedicated private method, and modernizing it to use e.g. `const`/`let`
instead of `var` and using template strings.
Currently the `web/app.js` file pulls in various build-specific dependencies, via the use of import maps, and those files in turn import from `web/app.js` thus creating undesirable import cycles.
To avoid this we instead pass in a `PDFViewerApplication`-reference, immediately after it's been created, to the relevant code.
Note that we use an ESLint plugin rule, see `import/no-cycle`, that is normally able to catch import cycles. However, in this case import maps are involved which is why this wasn't caught.
Looking at the *built* files you'll notice some lines containing nothing more than a semicolon. This is the result of (mostly top-level) `if`-statements, which include `PDFJSDev`-checks, that evalute to `false` during Babel parsing.
This has always annoyed me a bit, and looking at Babel plugin it seems that we can fix this simply by *removing* the relevant nodes.
This part of the (modern) preprocessor is now dead code, since we no longer use `require` statements anywhere in the main code-base.
Note that as part of the changes leading up to PDF.js version `4` we removed all[1] the remaining `require` statements, and we also have an ESLint rule to ensure that no new ones are accidentally added.
---
[1] With two small exceptions, in benchmarking-code and in the Webpack-example.
Given that we need to pass in a `PDFDataRangeTransport`-instance a number of the needed parameters can be obtained from it, rather than having to specify them manually.
This should be a *tiny* bit more efficient, since it avoids parsing substrings that we don't care about.
*Please note:* I cannot find an ESLint rule to enforce this automatically.
- Ensure that localization works in the GENERIC viewer, even if the necessary locale files cannot be loaded.
This was the behaviour prior to the introduction of Fluent, and it seems worthwhile to keep that (especially since we already bundle the en-US strings anyway).
- Let the `GenericL10n`-implementation use the *bundled* en-US strings directly when no language is provided.
- Remove the `NullL10n`-implementation, and simply fallback to `GenericL10n`, to reduce the maintenance burden of viewer-components localization.
- Indirectly, given the previous point, stop exporting `NullL10n` in the viewer-components since it's now removed.
Note that it was never really intended to be used directly and only existed as a fallback.
*Please note:* This doesn't affect the Firefox PDF Viewer, thanks to the use of import maps.
When an highlight is self-intersecting, the outline was drawn inside.
In order to remove it, we use a svg mask to exclude the shape inside
when drawing the outlines.
That leads to change the outline 1px,white-2px,blue-1px,white to a
2px,white-2px,blue: the part of the stroke which is inside the shape
is removed because of the mask.
All of our static evaluation & dead-code elimination transforms need to
happen in post-order, transforming inner nodes first. This is so that
in complex nested cases all transforms see the simplified version of
their inner nodes.
For example:
async getNimbusExperimentData() {
if (!PDFJSDev.test("GECKOVIEW")) { return null; }
// other code
}
-> [evaluation of PDFJSDev.*]
async getNimbusExperimentData() {
if (!false) { return null; }
// other code
}
-> [!false -> true]
async getNimbusExperimentData() {
if (true) { return null; }
// other code
}
-> [if (true) -> replace with the if branch]
async getNimbusExperimentData() {
return null;
// other code
}
-> [early return -> remove dead code]
async getNimbusExperimentData() {
return null;
// other code
}
This was done correctly in all cases except for our `UnaryExpression`
transform, which was happening in pre-order.
Having this parameter among a list of DOM-elements seems slightly strange now, however this is very old code hence the explanation for why this was done is for historical reasons (as is often the case).
Hence we can simply move this into `AppOptions` instead, which seems more appropriate overall.
Given that only the GENERIC viewer supports opening more than one PDF document, we can simplify things a tiny bit by instead generating the necessary DOM-element in JavaScript.
This unit-test is now failing in up to date versions of Node.js respectively Chromium-browsers, since `CompressionStream` no longer produces consistent data across all environments/browsers.
However logging the compressed TypedArray produced by `writeStream`, with Firefox respectively Chrome, and then feeding *both* of those TypedArray as input to `DecompressionStream` produced the same (correct) result in both browsers.
Hence the *exact* output of `CompressionStream` shouldn't matter, as long as we're able to successfully decompress it when the resulting PDF document is opened with the PDF.js library, and the unit-test is thus extended to check this.
Starting with Chrome 120.0.6099.109 (shipped with Puppeteer 21.8.0+) the
unit test fails in Chrome as well. The issue is tracked in #17399, but
for now we'll only run the unit test in Firefox so we can continue to
update Puppeteer while also still having a browser in which it runs,
until we figure out why the behavior of `CompressionStream` changed.
The `DefaultExternalServices` code, which is used to provide build-specific functionality, is very old. This results in a pattern where we first initialize `PDFViewerApplication.externalServices` and then *override* it for the different builds.
By converting `DefaultExternalServices` into a "regular" class, and leveraging import maps, we can directly initialize the correct instance depending on the build.
Given the simplicity of the `createPreferences` method, we can leverage import maps to directly initialize the correct `Preferences`-instance depending on the build.
Given the simplicity of the `createDownloadManager` method, we can leverage import maps to directly initialize the correct `DownloadManager`-instance depending on the build.
The latest mozilla-central update has test failures, because some CSS variables are not "properly" referenced; in particular:
- Give `--hcm-highlight-selected-filter` a default value, of `none`, similar to the previously existing HCM filter.
- Remove the `--mix-blend-mode` variable, since it's unused.
It isn't really a fix for the mentioned bug but it slightly improve things.
In reducing the memory use, the time spent in the GC is reduced either.
The algorithm to compute the bounding box is the same as before but it has just
been rewritten to be more efficient.
This commit converts the pdfjsdev-loader transform into a Babel plugin,
to skip a AST->string->AST round-trip.
Before this commit, the webpack build process was:
1. Babel parses the code
2. Babel transforms the AST
3. Babel generates the code
4. Acorn parses the code
5. pdfjsdev-loader transforms the AST
6. @javascript-obfuscator/escodegen generates the code
7. Webpack parses the file
8. Webpack concatenates the files
After this commit, it is reduced to:
1. Babel parses the code
2. Babel transforms the AST
3. babel-plugin-pdfjs-preprocessor transforms the AST
4. Babel generates the code
5. Webpack parses the file
6. Webpack concatenates the files
This change improves the build time by ~25% (tested on MacBook Air M2):
- `gulp lib`: 3.4s to 2.6s
- `gulp dist`: 36s to 29s
- `gulp generic`: 5.5s to 4.0s
- `gulp mozcentral`: 4.7s to 3.2s
The new Babel plugin doesn't support the `saveComments` option of
pdfjsdev-loader, and it just always discards comments. Even though
pdfjsdev-loader supported multiple values for that option, it was
effectively ignored due to `acorn` dropping comments by default.
This is in preparation for the next commit, which will convert
preprocessor2.mjs to a Babel plugin. The purpose of this commit
is to help git track the rename regardless of the large amount
of changes.
In the Gulpfile only the exit codes of `test.mjs` child processes
erroneously aren't checked. This causes failures in `test.mjs` to be
logged but not propagated to the master process, which in turn causes
test runners such as GitHub Actions to succeed because they only
monitor the master process. This is easy to reproduce by throwing an
error at the top of `test.mjs` and running `gulp makeref` or `gulp
unittest`: the error is logged, but the task that spawned the child
process succeeds and the master process exits with exit code 0. This is
problematic because it can easily cause errors to go by unnoticed.
This commit fixes the issue by making sure that the `test.mjs`
invocations are handled in the same way as the other child processes
in the file, i.e., if the child process exits with a non-zero exit code
then the master process also exits with a non-zero exit code. After this
patch the error is still logged, but the task now also fails and the
master process exits with exit code 1 to properly signal failure.
This manually ignores some cases where the resulting auto-formatting would not, as far as I'm concerned, constitute a readability improvement or where we'd just end up with more overall indentation.
Please see https://eslint.org/docs/latest/rules/arrow-body-style
For arrow functions that are both simple and short, we can avoid using explicit `return` to shorten them even further without hurting readability.
For the `gulp mozcentral` build-target this reduces the overall size of the output by just under 1 kilo-byte (which isn't a lot but still can't hurt).
The `if` statement is no longer necessary because the Node.js versions
that didn't provide `dns.setDefaultResultOrder` are no longer supported,
but looking into this a bit more it turns out that the entire workaround
is no longer necessary because the issue got fixed in Firefox 105 in bug
1769994. Indeed, Firefox now starts nicely with the workaround removed.
Reverts 60ed3cd297c4045b90f4114a74e5baa4ef1c5056.
In order to do that we must change the text layer opacity to 1 but
it has several implications:
- the selection color must have an alpha component,
- the background color of the span used for highlighted words
must have an alpha component either, but now the opacity is 1
we can use some backdrop-filters in HCM making the highlighted
words more visible.
- fix a regression caused by #17196: the css variable --hcm-highlight-filter
has to live under the #viewer element because in HCM it's overwritten
by js at this level, hence links annotations for example didn't
have the right colors when hovered.
The free highlighting is enabled when the mouse pointer isn't on some text.
Then we draw a shape with smoothed borders corresponding to the movement of
the mouse.
Printing/saving and changing the thickness will come later.
and try to load the font family (guessed from the font name) before trying
the local substitution.
The local(...) command expects to have a real font name and not a predefined
substitution it's why we try the font family.
By always removing the "visibilitychange" listener in the `PDFViewer.#onePageRenderedOrForceFetch`-method we can (ever so slightly) reduce duplication in the code.
Ensure that users cannot provide incorrect values when trying to set the global worker-options.
This patch was prompted by occasionally seeing users manually loading the `pdf.worker.mjs`-file and then assigning it to the `workerSrc`-option, something that obviously doesn't make sense and will cause fake-workers to be used (with poor performance as a result).
When the text of an annotation is extracted in using getTextContent, consecutive white spaces
are just replaced by one space and. So this patch add an option to make sure that white
spaces are preserved when appearance is parsed.
For the case where there's no appearance, we can have a fast path to get the correct string
from the Content entry.
When an existing FreeText is edited, space (0x20) are replaced by non-breakable (0xa0) ones
to make to see all of them on screen.
The system locale (used in OffscreenCanvas) can be different from the one guessed by Fluent,
consequently, in order to avoid any mismatch, we just use an attached canvas element.
The original issue can easily be reproduced locally in adding a lang="ja" in viewer.html
(or with an other language for Japanese users).
With modern JavaScript class features we can move the relevant event handling into private methods, and thus invoke it directly when resetting the toolbar UI-state.
*Please note:* This patch slightly reduces the size of the `web/secondary_toolbar.js` file.
With modern JavaScript class features we can move the relevant event handling into private methods, and thus invoke it directly when resetting the toolbar UI-state.
*Please note:* This patch slightly reduces the size of the `web/toolbar.js` file.
In PR 11912 we started caching images that occur on multiple pages globally, which improved performance a lot in many PDF documents.
However, one slightly annoying limitation of the implementation is the need to re-parse the image once the global-caching threshold has been reached. Previously this was difficult to avoid, since large image-resources will cause cleanup to run on the main-thread after rendering has finished. In PR 16108 we started delaying this cleanup a little bit, to improve performance if a user e.g. zooms and/or rotates the document immediately after rendering completes.
Taking those two PRs together, we now have a situation where it's much more likely that the main-thread has "globally used" images cached at the page-level. Hence we can instead attempt to *copy* a locally cached image into the global object-cache on the main-thread and thus reduce unnecessary re-parsing of large/complex global images, which significantly reduces the rendering time in many cases.
For the PDF document in issue 11878, the rendering time of *the second page* changes as follows (on my computer):
- With the `master`-branch it takes >600 ms to render.
- With this patch that goes down to ~50 ms, which is one order of magnitude faster.
(Note that all other pages are, as expected, completely unaffected by these changes.)
This new main-thread copying is limited to "large" global images, since:
- Re-parsing of small images, on the worker-thread, is usually fast enough to not be an issue.
- With the delayed cleanup after rendering, it's still not guaranteed that an image is available in a page-level cache on the main-thread.
- This forces the worker-thread to wait for the main-thread, which is a pattern that you always want to avoid unless absolutely necessary.
This commit changes the code to use a template string and to use `const`
instead of `var`. Combined with the previous commits this allows for
enabling the ESLint `no-var` rule for this file now.
The test helper code largely predates the introduction of modern
JavaScript features and should be refactored to improve readability.
In particular callbacks make the code harder to understand and maintain.
This commit:
- replaces the callback argument with returning a promise;
- replaces the recursive function calls with a simple loop;
- uses `const`/`let` instead of `var`;
- uses arrow functions for shorter code;
- uses template strings for shorter string formatting code.
The test helper code largely predates the introduction of modern
JavaScript features and should be refactored to improve readability.
In particular callbacks make the code harder to understand and maintain.
This commit:
- replaces the callback argument with returning a promise;
- uses `const` instead of `var`;
- uses arrow functions for shorter code;
- uses template strings for shorter string formatting code;
- uses `Array.includes` for shorter response code checking code.
This test intermittently fails, likely because the auto-print is triggered fast enough
that we don't manage to get it.
So this patch aims to try to set a listener very early in order to be sure that
we'll be aware that a print has been triggered.
It seems this unit-test now fails consistently in "all" up-to-date Node.js versions. We should probably try and understand why, but for now just disable it to get passing CI tests.
There's obviously a few things wrong with the Annotations in the referenced PDF document, however parsing of an Annotation shouldn't just break if the /BS-entry isn't a dictionary.
When opening a pdf from the secondary toolbar, a second color picker is added.
So in order to avoid that, we just stop listening for annotationeditoruimanager
in the toolbar.
The doorhanger for highlighting has a basic color picker composed of 5 predefined colors
to set the default color to use.
These colors can be changed thanks to a preference for now but it's something which could
be changed in the Firefox settings in the future.
Each highlight has in its own toolbar a color picker to just change its color.
The different color pickers are so similar (modulo few differences in their styles) that
this patch introduces a new class ColorPicker which provides a color picker component
which could be reused in future editors.
All in all, a large part of this patch is dedicated to color picker itself and its style
and the rest is almost a matter of wiring the component.
When a pdf as a FreeText without appearance, we use a fake font in order to render it
and that leads to create few new refs for the font.
But then when we're saving, we create some new refs which start at the same number
as the previous created ones.
Consequently, when saving we're using some wrong objects (like a font) to check if
we're able to render the newly added FreeText.
In order to fix this bug, we just remove the persistent refs (which are only used
when rendering/printing) during the saving.
Having just tested PR 17337 locally I noticed that especially the `JpxImage`-test causes a "ridiculous" amount of warning messages to be printed, which doesn't seem helpful.
Given that only actual `Error`s should be relevant here, we can easily disable this logging during the tests.
The test helper code largely predates the introduction of modern
JavaScript features and should be refactored to improve readability.
In particular callbacks and recursive function calls make the code
harder to understand and maintain.
This commit:
- replaces the callback argument with returning a promise;
- replaces the recursive function calls with a simple loop;
- uses `const`/`let` instead of `var`;
- uses template strings for shorter string formatting code;
- improves the error messages to have more details.
The test helper code largely predates the introduction of modern
JavaScript features and should be refactored to improve readability.
In particular callbacks and recursive function calls make the code
harder to understand and maintain.
This commit:
- replaces the callback argument with returning a promise;
- uses `const` instead of `var`;
- uses arrow functions for shorter code;
- uses template strings for shorter string formatting code;
- improves the error messages to have more details.
This unfortunately broke in PR 17060, since I had completely forgotten about https://bugzilla.mozilla.org/show_bug.cgi?id=1632644#c5 when writing that patch.
The easiest solution, while slightly unfortunate, seems to be to add a couple of non-standard hash parameters specifically for the PDF attachment use-case in the Firefox PDF Viewer. (Note that we cannot use "nameddest" here, since we also need to support the stringified destination-Array case.)
It seems this unit-test started failing in Node.js version 20.10 as well. We should probably try and understand why, but for now just disable it to get passing CI tests.
Given that this event listener is only used to trigger rendering after the sidebar has been opened/closed, we can utilize the existing one in the `PDFSidebar` class for this purpose instead. That one is registered on the sidebar DOM-element, and is needed to remove a CSS-class indicating that the sidebar is moving.
It fixes few errors in the CSS for HCM.
It now complies to the specs from UI/UX.
Only the foreground must change in HCM and not the background, similarly to what
we had for the alt-text button before moving it.
After the two previous commits, which removed the remaining call-sites, this method is no longer used and can thus be removed.
As mentioned in the JSDocs for the now removed method, synchronous communication between the viewer and the platform code isn't really a good idea.
Once this patch has landed in mozilla-central some additional clean-up of the platform code will also be possible.
The return value is not, nor has it ever been, used for anything and we should thus be able to just send the message.
Note that the responses are already handled by the "message" event listener registered above.
This commit fixes the JSDoc comment for the `annotationEditorMode` setter.
The types tests fail on that now because the input value was changed from
a number to an object with various properties in recent patches, but the
JSDoc comment was not updated accordingly.
Moreover, the types tests also fail because TypeScript 5.3 assumes that
getters and setters have equal return and input value types, which is
arguably also what one would expect, but our `annotationEditorMode`
getter and setter deviate from that because the getter returns a number
while the setter accepts an object. Given that it seems more important
to document the setter entirely, including the meaning and types of its
properties, and the type of the getter can easily be inferred from this
comment and the other JSDoc comments that have `annotationEditorMode` in
it, we remove the getter type to make the types tests pass again.
- Extend the `fetchData` helper function to also support fetching of "blob" data.
- Use the `fetchData` helper function more in the code-base, when fetching non-PDF data. Given that the Fetch API isn't supported for all protocols, this should improve compatibility for the PDF.js library.
Currently the SVG images for the loading-icons exist in two versions, for the light- respectively dark-theme, which nowadays are the only "duplicated" icons left.
The reason for this is that these icons are being used in `input`-elements, where the regular `mask-image` approach used for all buttons don't work.
To address this we add containers for the `input`-elements, such that we have a "regular" DOM-element where we can use `mask-image`.
The goal is to be able to get these outlines to fill the shape corresponding
to a text selection in order to highlight some text contents.
The outlines will be used either to show selected/hovered highlights.
- Re-factor the existing `fetchData` helper function such that it can fetch more types of data, and it now supports "arraybuffer", "json", and "text".
This only needed minor adjustments in the `DOMCMapReaderFactory` and `DOMStandardFontDataFactory` classes.[1]
- Expose the `fetchData` helper function in the API, such that the viewer is able to access it.
- Use the `fetchData` helper function in the `GenericL10n` class, since this should allow fetching of localization-data even if the default viewer is run in an environment without support for the Fetch API.
---
[1] While testing this I also noticed a minor inconsistency when handling standard font-data on the worker-thread.
This is consistent with the implementation used in the (now removed) webL10n-library, and by only using lowercase language-codes internally in the `L10n`-implementations we should avoid future issues e.g. when users manually set the `locale`-option (in the default viewer).
Some fields, somewhere under the Fields entry in Acroform, could have no name (in T)
but with a parent which has a name but which isn't somewhere under Fields.
As a side-effect, this patch prevents infinite loops because of potential cycles
under Fields.
This commit migrates the font tests away from the bots. Not only are the
font tests broken on the Windows bot since some time, they also run on
Python 2 (end of life since January 2020) and `ttx` 3.19.0 (released in
November 2017). The latter is installed via a submodule, which requires
more complicated logic for finding and running `ttx`.
We solve the issues by implementing a modern workflow that installs the
most recent stable Python and `ttx` (`fonttools` package) versions. This
simplifies the `ttx` driver code as well because it can now assume `ttx`
is available on the path (just like we do for e.g. `node` invocations).
GitHub Actions takes care of creating a virtual environment with
`fonttools` in it so that the `ttx` entrypoint is available. Locally
the font tests can be run in a similar way by creating and sourcing a
virtual environment with `fonttools` in it before running the font
tests, and a README file is included with instructions for doing so.
This commit prepares for running the font tests on GitHub Actions where
we can't spin up headful browsers because there are no display
capabilities on the workers. This will also be useful for porting other
test targets to GitHub Actions at a later time, as well as running the
tests locally in headless mode.
This commit prepares for the introduction of extra options in later
commits by changing the function signatures of the `startBrowser(s)`
functions to take parameter objects instead of plain parameters. This
makes the call sites explicitly state which parameters they pass,
improving overall readability as well.
The current logic is more complicated than it needs to be because it's
passing a callback function to `startBrowsers` instead of a string.
This commit simplifies the logic by passing the base URL as a string to
`startBrowsers` and having it do further augmentation internally,
thereby removing all indirection of the function calls to `makeTestUrl`
and the inner function it returned.
After recent PRs the size and scope of the CI workflow is now reduced, and this patch tries to simplify things further. More specifically we can directly specify the gulp-tasks in the workflow, and thus clean-up the `gulpfile` a tiny bit.
Note that this will technically be slower, since the tests are now run in series (rather than in parallel), however `gulp externaltest` runs so quickly that it really won't matter in practice.
Hopefully this is enough to address the problem of initializing the Worker in Chromium-based browsers.
Locally I've tried to *force* use of `createCDNWrapper` in development mode, by commenting out the `isSameOrigin` checks, and worker-loading fails against `master` and works with this patch.
Currently the background-color of the `editorParamsToolbar`s don't match that of the arrow, which is especially noticable in dark mode (see zoomed-in screen-shots below).
The simplest solution seem to be to just style the `editorParamsToolbar`s like the `secondaryToolbar`, to limit the amount of CSS changes required.
This should *hopefully* fix 17228, by tweaking the build scripts to give the GENERIC viewer something to await to avoid breaking third-party users of the standalone viewer components.
This button is *only* used in the GENERIC viewer, and will currently be visible either in the main or secondary toolbars (depending on the viewer width).
To simplify upcoming changes, and to avoid then having to complicate the relevant CSS rules unnecessarily, let's place the "Open file"-button permanently in the secondary toolbar instead.
(Note that the GENERIC viewer also, since five years, supports drag-and-drop in order to open local files.)
The `fieldObjects`-getter is implemented in the `PDFDocument` class, which means that the `this._localIdFactory`-property that we pass to `AnnotationFactory.create` doesn't actually exist.
The reason that this hasn't caused any bugs, that I'm aware of, is that all /Fields-entries need to be References to actually make sense.
The `fieldObjects`-getter itself is called, from `src/core/worker.js`, in a way that'll ensure that any `MissingDataException`s are handled. However the problem is that the actual data-lookups in `fieldObjects` and `#collectFieldObjects` are done inside of a Promise, which means that `MissingDataException`s won't be handled and parsing could thus break.
To address this we change all data-lookups to be asynchronous instead.
The `viewerCssTheme`-implementation has always been somewhat hacky, and now it's also *partially* broken ever since we've started using CSS nesting.
Trying to support nested media queries would thus require a lot more parsing of the CSS rules, which seems inefficient and thus generally undesirable.[1]
As discussed on Matrix, let's try to remove the `viewerCssTheme`-option and see if there's any (significant) fallout from this.
---
[1] If this option is brought back, it seems to me that it (in Firefox) should probably be set through the platform-code that handles theming.
Depending on the structure of the outline we could potentially need to expand a few levels, especially in long PDF documents, hence it cannot hurt to pause translation in that case as well.
With the changes in PR 17208, where browser-preferences are now handled as "regular" viewer-options, we can tweak the definition of `canvasMaxAreaInBytes` to slightly simplify things in the `PDFViewerApplication.open` method.
Hopefully this will allow us to catch bugs in new Node.js versions earlier, rather than having to wait for bug reports.
Given that `CompressionStream` is (currently) only potentially used when saving a *modified* PDF document, which is unlikely to be a common use-case in Node.js environments, let's just disable the affected unit-test for now.
Given that this branch is only necessary in development mode and *during* building, but is never actually used in the final viewer-bundles, we can utilize the pre-processor to ignore this code.
Currently we *synchronously* fetch a number of browser preferences/options, from the platform code, during the viewer respectively PDF document initialization paths.
This seems unnecessary, and we can re-factor the code to instead include the relevant data when fetching the regular viewer preferences.
The active LTS version is now based on Node.js version 20, hence let's update the relevant workflows to use that one instead; see https://en.wikipedia.org/wiki/Node.js#Releases
Given that we still support Node.js version 18, i.e. the maintenance LTS version, in the PDF.js library we'll keep testing both versions in GitHub Actions to prevent regressions.
I noticed the following warning in the GitHub Actions workflow logs:
`Configuration file not found: .github/linter_config.yml`
The configuration file is called `fluent_linter_config.yml` instead, so
this commit fixes the path so it points to the correct file.
Fixes 487816b.
The current stable version of Python is Python 3.12, see
https://www.python.org/downloads, so we should switch to that since
Python 3.10 is older and only receives security updates.
This commit tweaks the Fluent linter workflow to match the other
workflow files we have, so we make sure the steps have a newline between
them for better readability and align names and descriptions of steps
with how they are called in the other workflow files we have.
There are environments that include *incomplete* polyfills for the `navigator`-object, which may thus cause the PDF.js library to break.
Despite that clearly not being our fault, it may still result in bug reports filed against the PDF.js project; see e.g. 15728.
Currently this even seem to affect *the latest* version of Node.js; see e.g. [here].
*Please note:* Thanks to the pre-processor none of these changes affect the Firefox PDF Viewer, however it does add "overhead" when working with and reviewing the affected code (which is why I'm not crazy about this).
*Please note:* While following the steps in the README still works with this patch, in the sense that the example runs and successfully renders a PDF document, I unfortunately cannot tell if it illustrates Webpack best practices.
- Remove the `errorWrapper`-element, since it simplifies the example and is consistent with the default viewer; see PR 15533.
- Simplify the l10n-handling, since the `NullL10n` should be able to translate everything e.g. without fallback values; see PR 17146.
Note that we must append the textLayer to the DOM *before* enabling the `highlighter` and `accessibilityManager`, to avoid breaking e.g. a pending searching operation.
The least invasive solution, that I was able to come up with, is to introduce a new `TextLayerBuilder` callback-function for this purpose.
Currently the `WidgetAnnotationElement._getKeyModifier` method will always be falsy on Linux, which seems like a simple oversight. Looking at all the other `FeatureTest.platform` accesses we only handle the `isMac`-case specially, and it seems reasonable to do the same thing here.
The reason that this hasn't led to any bug reports is most likely that the `modifier`-property seems completely unused in the scripting-implementation.
Finally, with these changes we can (slightly) simplify the `FeatureTest.platform` implementation.
After PR 17177 the interface of `XfaLayerBuilder` is now inconsistent, since whether or not we directly append the xfaLayer to the DOM now depends on the rendering intent.
I forgot to include `web/l10n_utils.js` in PR 17161, which currently breaks `ConstL10n` since there's no longer a method called `setL10n`; sorry about that!
Most of the strings shouldn't contain special chars (<= 0x1F) so we can
have a fast path which just checks if the string contains at least one such
a char.
To prevent the *standalone* viewer-components from breaking, we need to ensure that the `NullL10n`-interface won't accidentally diverge from the actual `L10n`-implementations.
Looking at the `PDFThumbnailView.setPageLabel` method you'll see that we update e.g. the "aria-label" of the thumbnail-image for documents that contain (valid) pageLabels.
This isn't done in `PDFPageView`, which seems inconsistent, hence this patch.
This patch changes almost all viewer-components[1] to use "data-l10n-id"/"data-l10n-args" for localization, which means that in many cases we no longer need to pass around the `L10n`-instance any more.
One part of the code-base where the `L10n`-instance is still being used "directly" is the AnnotationEditors, however while it might be possible to convert (most of) that code as well that's not attempted in this patch.
---
[1] The one exception is the `PDFDocumentProperties` dialog, since the way it's currently implemented makes that less straightforward to fix without a lot of code changes.
*Please note:* In the Firefox PDF Viewer this findbar is only used for PDF documents placed in e.g. `<iframe>` elements.
By registering a `ResizeObserver` when the `PDFFindBar` is open we slightly unify and simplify how the findbar layout (row vs column) is handled.
This will be especially helpful with upcoming changes, where we'll make use of "data-l10n-id"/"data-l10n-args" to trigger translation in the viewer.
- The old translation engine handled language code casing slightly differently, hence we need to tweak the non-metric locale check in `PDFDocumentProperties` to account for that.
- Use only lowercase names for the pre-defined page names, to improve overall consistency.
*Please note:* These changes only affect the GENERIC build, since `NullL10n` is only a stub elsewhere (see PR 17135).
After the changes in PR 17115, which modernized and improved l10n-handling, the `NullL10n`-implementation is no longer a good fallback for the "proper" `L10n`-classes.
To improve this situation, especially for the *standalone* viewer-components, this patch makes the following changes:
- Let the `NullL10n`-implementation extend an actual `L10n`-class, which is constant and lazily initialized, to ensure that it works *exactly* like the "proper" ones.
- Automatically bundle the "en-US" l10n-strings in the build, via the pre-processor, such that we don't need to remember to manually update them.
- Ensure that the *standalone* viewer-components register their DOM-elements for translation, similar to the default viewer, since this will allow future code improvements by using "data-l10n-id"/"data-l10n-args" in most (if not all) parts of the viewer.
- Remove the `NullL10n` from the `AnnotationLayer`, to avoid affecting bundle size too much.
For third-party users that access the `AnnotationLayer`, as exposed in the main PDF.js library, they'll now need to *manually* register it for translation. (However, the *standalone* viewer-components still works given the point above.)
Use existing helper to calculate the Box
Co-authored-by: Jonas Jenwald <jonas.jenwald@gmail.com>
Ensure that there are non-zero
Co-authored-by: Jonas Jenwald <jonas.jenwald@gmail.com>
Add a reference test for #17147
Given that there's now a bit more asynchronicity in the l10n-initialization in the Firefox PDF Viewer, after PR 17115, try to limit the impact of that by moving it to occur a tiny bit earlier in the default viewer initialization.
In Firefox debug builds, there is an assertion to check that we don't connect
a subelement of an already connected root. Thanks to this assertion, we can see
that the root has already been added to Fluent, hence we don't need to do it
a second time.
We don't need to await anymore on the translation in order to update the
toolbar: it'll be done by Fluent, so we can safely remove the "localized"
event and avoid to wait for it.
*Please note:* This patch contains a couple of micro-optimizations, hence I understand if it's deemed unnecessary.
Move the `AppOptions` initialization into the `Preferences` constructor, since that allows us to remove a couple of function calls, a bit of asynchronicity and one loop that's currently happening in the early stages of the default viewer initialization.
Finally, move the `Preferences` initialization to occur a *tiny* bit earlier since that cannot hurt given that the entire viewer initialization depends on it being available.
Note that CSS-features such as e.g. `flex` didn't exist, or had poor cross-browser support, back when the JavaScript-based solution was initially implemented.
- For the generic viewer we use @fluent/dom and @fluent/bundle
- For the builtin pdf viewer in Firefox, we set a localization url
and then we rely on document.l10n which is a DOMLocalization object.
Currently we're unnecessarily converting data between strings and typed-arrays, when dealing with compressible data, in the `writeStream` function.
Note how we're *first* getting a string-representation of the stream, which involves converting the underlying typed-array into a string, only to immediately convert this back into a typed-array. This seems completely unnecessary, and is easy enough to avoid, and we'll now only do a *single* type-conversion in this function.
The current test fails intermittently only on Windows for unknown
reasons: the code is correct and on Linux it always passes. However, we
have already spent quite a lot of time on this test, so rather than
spending even more time on it I figured we should look at what behavior
the test is trying to check and find an alternative way to do it that
can't trigger this intermittent issue anymore.
This commit changes the test to use a term that only exists once in the
entire document so we cannot accidentally highlight another match
anymore. This doesn't change anything about the behavior that this test
aims to check: we still test searching in the XFA layer, we still test
that the original term is matched case-insensitively and we still test
that that match is actually highlighted. Note that the only objective of
the test is confirming that the search functionality covers the XFA
layer, so the exact phrase/match is not the interesting bit.
Given that we only use standard `import`/`export` statements now, after recent PRs, the "exports" global is unused.
Instead we add "__non_webpack_import__" to the `globals` to avoid having to sprinkle disable statements throughout the code.
Finally, the way that `globals` are defined has changed in ESLint and we should thus explicitly specify them as "readonly"; please find additional details at https://eslint.org/docs/latest/use/configure/language-options#specifying-globals
The previous change that set the timeout had effect because we have seen
quite a few protocol timeouts now correctly being raised in the context
of the active test, however we have also still seen a handful of cases
where this wasn't the case and the one second difference turned out to
be too low (likely because the operation was started slightly after one
second into the test run). We therefore tweak the value to be 75% of the
Jasmine timeout. This should be enough to catch operations that happen
later on in the test run, and if a single operation takes that long any
hope for success is already gone anyway.
It's not necessary because we have configured silent printing for
Firefox and Chrome in the browser arguments we pass in `test.mjs`. This
means that the print dialog is not even shown at all or disappears
automatically once printing is done, so the Escape key press serves no
purpose. Since it has been shown to time out, likely because the page
loses focus during printing, and because the page itself doesn't know
when the printing dialog is shown and we therefore can't possibly do the
key press at the right time anyway, this commit gets rid of it to
stabilize the test.
Those files only contain old debugging code that is not used/imported
anywhere anymore, which is generating code scanning alerts. Moreover,
they rely on globals/platform-specific code and don't import/export
logic properly.
Given the amount of work put into removing `require`-calls from the code-base, let's ensure that new ones aren't accidentally added in the future.
Note that we still have a couple of files where `require` is being used, in particular:
- The Node.js examples, however those will be updated to use `import` in PR 17081.
- The Webpack examples, and related support files, however I unfortunately don't know enough about Webpack to be able to update those. (Hopefully users of that code will help out here, once version `4` is released.)
- The `statcmp`-tool, since *some* of those `require`-calls cannot be converted to `import` without other code changes (and that file is only used during benchmarking).
Please find additional details at https://github.com/import-js/eslint-plugin-import/blob/main/docs/rules/no-commonjs.md
This *finally* allows us to mark the entire PDF.js library as a "module", which should thus conclude the (multi-year) effort to re-factor and improve how we import files/resources in the code-base.
This also means that the `gulp ci-test` target, which is what's run in GitHub Actions, now uses JavaScript modules since that's supported in modern Node.js versions.
For large/complex images it's possible that the image-data arrives in the API *after* the page has been scrolled out-of-view and thus been cleaned-up. In this case we obviously shouldn't cache such page-level data, since it'll first of all be unused and secondly can increase memory usage *a lot*.
Also, ensure that we *immediately* release any `ImageBitmap` data in this case to help reclaim memory faster.
The examples themselves were updated to account for JavaScript modules, which didn't require changing the actual URLs.
However, since it seems that JSFiddle doesn't support JavaScript modules in its separate "JavaScript" editing-area we need to change how we embed the examples to avoid showing a blank "JavaScript"-tab.
When pdfBug is true, the substitution font is used in the text layer in order
to be able to know what is the font really used thanks to the devtools.
And to be sure that fonts are loaded, the font cache isn't cleaned up when
the debugger is active.
The default protocol timeout is 180 seconds according to the
documentation at https://pptr.dev/api/puppeteer.browserconnectoptions,
but the Jasmine timeout we configure in the individual boot files is 30
seconds. The consequence of this is that if a protocol (CDP) error
occurs after 30 seconds Jasmine will fail the test, but the actual
protocol error from Puppeteer is raised much later in the context of
another test, which causes unrelated failures or tracebacks.
This commit fixes the problem by configuring Puppeteer to always use a
lower protocol timeout than the Jasmine timeout so that protocol errors
are always raised in the context of the test that actually triggered it.
It's been loaded as a JavaScript module for a long time, and given that the file is bundled as-is (without building) it seems reasonable to just change the file extension now.
The Windows bot is usually slower than the Linux bot, and therefore
text layer rendering is as well. However, the `autoprint` test awaited
text layer rendering to complete before activating the selector check,
which makes it timing-sensitive and causes it to never resolve because
the page is already printed (and the printed page div removed) by then.
This commit should fix the issue by activating the selector check as
soon as possible, namely as soon as the viewer appears, which should
ensure we're always registering the selector check in time because we're
doing it even before rendering is starting.
Comparing the currently supported browsers/environments, see [the FAQ](https://github.com/mozilla/pdf.js/wiki/Frequently-Asked-Questions#faq-support) and the [MDN compatibility data](https://developer.mozilla.org/en-US/docs/Web/API/structuredClone#browser_compatibility), the `structuredClone` polyfill is *only* needed in Google Chrome versions < 98. Because of some limitations in the core-js polyfill we're currently forced to special-case the `transfer` handling to prevent bugs, and it'd be nice to avoid that.
Note that `structuredClone`, with transfers, is only used in two spots:
- The `LoopbackPort` class, which is only used with fake workers. Given that fake workers should *never* be used in browsers, breaking that edge-case in older Google Chrome versions seem fine.
- The `AnnotationStorage` class, when Stamp-annotations have been added to the document. Given that Google Chrome isn't the main focus of development, breaking *part* of the editing-functionality in older Google Chrome versions should hopefully be acceptable.
To avoid problems with `export` statements in the QuickJS Javascript Engine, we can work-around that by *explicitly* exposing `pdfjsScripting` globally instead.
At this point in time all browsers, and also Node.js, support standard `import`/`export` statements and we can now finally consider outputting modern JavaScript modules in the builds.[1]
In order for this to work we can *only* use proper `import`/`export` statements throughout the main code-base, and (as expected) our Node.js support made this much more complicated since both the official builds and the GitHub Actions-based tests must keep working.[2]
One remaining issue is that the `pdf.scripting.js` file cannot be built as a JavaScript module, since doing so breaks PDF scripting.
Note that my initial goal was to try and split these changes into a couple of commits, however that unfortunately didn't really work since it turned out to be difficult for smaller patches to work correctly and pass (all) tests that way.[3]
This is a classic case of every change requiring a couple of other changes, with each of those changes requiring further changes in turn and the size/scope quickly increasing as a result.
One possible "issue" with these changes is that we'll now only output JavaScript modules in the builds, which could perhaps be a problem with older tools. However it unfortunately seems far too complicated/time-consuming for us to attempt to support both the old and modern module formats, hence the alternative would be to do "nothing" here and just keep our "old" builds.[4]
---
[1] The final blocker was module support in workers in Firefox, which was implemented in Firefox 114; please see https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Statements/import#browser_compatibility
[2] It's probably possible to further improve/simplify especially the Node.js-specific code, but it does appear to work as-is.
[3] Having partially "broken" patches, that fail tests, as part of the commit history is *really not* a good idea in general.
[4] Outputting JavaScript modules was first requested almost five years ago, see issue 10317, and nowadays there *should* be much better support for JavaScript modules in various tools.
The user should *always* provide a correct `GlobalWorkerOptions.workerSrc` value when using the PDF.js library in browser environments. Note that the fallback:
- Has been deprecated ever since PR 11418, first released in version `2.4.456` over three years ago.
- Was always a best-effort solution, with no guarantees that it'd actually work correctly.
- With upcoming changes, w.r.t. outputting JavaScript modules, it'd now be more diffiult to determine the correct value.
The minified default viewer has never been distributed in either official releases or through pdfjs-dist, which means that it's most likely unused, and it's has never been tested nor actively maintained.
Setting the alpha-value explicitly to `1` in `rgb` colors is unnecessary, since that's the default value, and this way we ever so slightly reduce the size of our CSS files.
Unfortunately I've not found a Stylelint rule to enforce this automatically, and the patch was generated using search and replace.
When an editing button is disabled, focused and the user press Enter (or space), an
editor is automatically added at the center of the current page.
Next creations can be done in using the same keys within the focused page.
When an element has the hasOwnCanvas flag we must have an HTML container to attach
the canvas where the element will be rendered.
So the noHTML flag must take this information into account:
- in some cases the noHTML flag is resetted depending on the hasOwnCanvas value;
- in some others, the hasOwnCanvas flag is set depending on the value of noHTML.
To reduced the risk of regressing something else, given that the issue only applies to a (for the default viewer) non-default configuration, this patch is purposely limited to only TextWidget-annotations in the display layer.
It happens only on windows with chrome.
For any reason, click event isn't correctly triggered and it seems work correctly
with pointerup.
And it seems that when drawing a svg on an OffscreenCanvas we need to wait
a little in order to be able to transfer it: it's why this patch adds
a check on the canvas content.
This has been deprecated since version `2.15.349`, which is a year ago.
Removing this will also simplify some upcoming changes, specifically outputting of JavaScript modules in the builds.
This removes the only remaining old and non-standard handling of exports in the `web/`-folder, since some initial attempts at outputting JavaScript modules in the builds have identified this file as a potential problem.
While this uses a hard-coded list, for overall simplicity, I don't believe that that's a big problem since:
- Generating this file automatically would require a bunch more parsing *every single time* that the library is built.
- The official API-surface doesn't change often enough for this to really impede development in any significant way.
- The added unit-test helps ensure that this list cannot accidentally become outdated.
When the editor is invisible (because on a non-rendered page) its parent is null.
But when we undo its deletion, we need to have a parent to attach it.
Given that this is accessed multiple times per page in the viewer, that leads to a number of (strictly speaking unneeded) function calls and allocated Objects for each invocation. By converting `layerProperties` to a, lazily initialized, Object we can avoid this.
When PR 17015 removed the `disabled` handling for the "Save"-button it left a bunch of now unused CSS rules behind, which seems like a simply oversight.
Rather than shipping "dead" CSS rules, let's remove those until such a time that they're actually needed.
The old pre-processor used for CSS, and HTML, files leaves comments intact which unnecessarily contributes to the overall size of the *built* CSS files (note that the built JavaScript files don't include comments).
Rather than trying to "hack" comment removal into the pre-processor it seems easier to use a PostCSS plugin instead. The one potential issue is that it also affects *some* whitespaces, and it's not clear to me if this'll work with the various CSS-related tests that run in mozilla-central.
Please refer to https://www.npmjs.com/package/postcss-discard-comments for additional information.
*For many non-English locales the translated strings will be longer, which is easy to forget about during development/review.*
Note how for some locales (e.g. Swedish) the altText-button end up looking horizontally "cramped", hence it seems reasonable to add a bit of inline padding to improve this.
In the scripting integration tests we use a few different typing
delays, mostly 100 or 200 milliseconds. According to for example
https://www.typingpal.com/en/documentation/school-edition/pedagogical-resources/typing-speed,
a fast typing speed is around 300 characters per minute, which is 5
characters per second and therefore a delay of 200 milliseconds between
each keystroke. Note that this is already above average, so in practice
the delay will be even larger. Therefore the 100 milliseconds variant
is unrealistically fast and therefore not suitable for the integration
tests which aim to simulate the average user behavior.
On top of that, the quick typing speeds are problematic for the tests
that involve validation alert dialogs appearing during typing. In those
tests a handler is registered to close the dialog once it pops up, but
it takes time for Puppeteer to notice the dialog, trigger the handler
and close it. If the typing delay, which is the delay between the key
down and key up events according to the Puppeteer source code at
https://github.com/puppeteer/puppeteer/blob/master/packages/puppeteer-core/src/cdp/Input.ts#L209-L215,
is too short, the key up event will be fired before the dialog is
closed. In that time the text box we're typing in is not focused, so
when the dialog is closed the `page.type()` call on the text box will
never resolve because the key up event never reached the text box.
This commit aims to fix the issues by converting all 100 millisecond
delays to 200 milliseconds. For instance the "must check input for US
zip format" failed pretty consistently locally before and hasn't failed
anymore with a 200 millisecond delay.
This integration test fails often because we wait for scripting to be
ready before we check the printed page, but most of the time the PDF
is already done printing before scripting is reported to be ready.
This happens because the print trigger is on the `Open` event, which is
one of the first events to be dispatched and, most notably, before
scripting is marked as ready; please see
https://github.com/mozilla/pdf.js/blob/master/web/pdf_scripting_manager.js#L176-L191.
Given that the PDF document is only one page, printing it is usually
finished between triggering the `Open` event and scripting reported
to be ready. If this happens the printed page is already destroyed
before we get to our actual test, which will then timeout because it
will never find the printed page in the DOM.
This commit fixes the problem by not awaiting scripting to be ready
because the fact that the printed page appears is already enough to know
that autoprint was triggered (after all, there is no other user
interaction involved here). While we're here we also switch to the
shorter `page.waitForSelector` function.
but keep it for the text area.
Disable pointerdown on the alt-text button to disable dragging the editor
when the button is clicked (especially when slightly moving the mouse
between the down and the up).
The dialog element handles closing with <kbd>Esc</kbd> automatically, however we're not reporting telemetry in that case.
In order to fix that the easiest solution, as far as I'm concerned, seem to be moving the telemetry reporting into the dialog-close handler since it's always invoked.
This patch addresses an edge-case that'll probably never happen, but it nonetheless seems like something that we want to fix.
Note how we're using the `#currentEditor`-field to prevent opening the dialog when it's already active, and it being reset once the dialog has been closed.
By also resetting the `#currentEditor`-field during destruction, instead of waiting until the dialog has actually closed (assuming it's currently open), there's a *tiny* window of time[1] during which we could theoretically allow to (incorrectly) re-open the dialog and thus getting out-of-sync state in the viewer-component.
---
[1] Since the "close" event, on a dialog-element, is dispatched asynchronously by the browser.
When the user edit an existing alt-text and remove it, we want to be able
to save this state and consequently remove the done state from the
alt-text button.
Remove the button from its parent when the editor is removed: it should
help to save few Kb of memory.
Radio-buttons can also be toggled by clicking on their associated `label`-elements, and not only the `input`-elements itself, however it seems that "pointerdown" event listeners don't cover that case.
Hence it's possible that telemetry could miss certain cases of a mouse being used, and the easiest solution seem to be to instead use "click" event listeners and just ignore keyboard-based events.
Rather than trying to be "clever" here, and possibly affect code readability negatively, let's just restore the `collectFields` parameter to address the unneeded parsing that now happens when printing new Annotations.
Given the limitations of the old pre-processor that's used for CSS/HTML files, this unfortunately isn't as "easy" to implement as it is for JavaScript code.
Since this is the first case where we've wanted to do conditional CSS imports, rather than trying to completely re-write the pre-processor, this patch settles for handling it explicitly in the `expandCssImports` function.
Looking at the save-telemetry values they're all boolean *except* for "alt_text_edit" in one instance, since `this.#previousAltText` may be an empty string (looking at the `editAltText` method) and this value may thus become an empty string as well.
When closing a document in the viewer, e.g. by running `PDFViewerApplication.close()` in the console, the `AltTextManager.#finish` method currently throws *unless* the `altText` dialog is actually open.
Similar to e.g. the PasswordPrompt, we should thus only attempt to close the `altText` dialog when it's open.
In the rare situation that an optional content dictionary lacks a /Type-entry we currently throw, which may prevent e.g. Form XObjects from rendering completely.
Fixes https://bugs.ghostscript.com/show_bug.cgi?id=707147
It's a part of the UX specifications. There's a drawing issue in Firefox
(see bug https://bugzilla.mozilla.org/1853288) but setting the
background-clip property to content-box seems to be a good workaround.
Especially on slower bots there is some time between clicking the
element and the actual visibility change, but we didn't await this and
checked the visibility state immediately after clicking. This can be
reproduced 100% of the time by introducing a delay in the `display` and
`hidden` handlers of the `_commonActions` shadow call.
This commit fixes the problem by waiting until the first visibility
change actually happened before continuing with the assertions.
This integration test currently fails intermittently on the bots because
of the fixed timeout in the test, which is sometimes too low on slower
systems. The issue can be reproduced 100% of the time by introducing a
delay just before dispatching the `switchannotationeditormode` event.
Puppeteer also discourages this and instead recommends waiting for a
selector instead, which we now do here. This ensures that the test only
continues if the element under test is available and therefore prevents
any timing problems.
The x/y-coordinates are floats instead of integers like one might
expect. The current approach rounds both the old and the new
coordinates in order to do integer comparison. However, rounding each
coordinate individually causes too much loss of precision because,
depending on the decimal value, they are either rounded up or down
which causes intermittent off-by-one errors.
This commit fixes the problem by comparing coordinate differences
instead of the coordinates themselves. The precision loss is avoided
by subtracting the old from the new coordinate as-is and only rounding
the final result.
This integration test currently fails intermittently on the bots because
of the fixed timeout in the test, which is sometimes too low on slower
systems. The issue can be reproduced 100% of the time by introducing a
delay in the `WidgetAnnotationElement.showElementAndHideCanvas` method.
Puppeteer also discourages this and instead recommends waiting for a
selector instead, which we now do here. This ensures that the test only
continues if the element under test is available and therefore prevents
any timing problems.
We already use `page.$eval` in most other integration tests and it's
simpler because it already takes the selector as argument, so we don't
have to do a separate `querySelector` call ourselves.
When there is no tree, the tags for the new annotions are just put under the root element.
When there is a tree, we insert the new tags at the right place in using the value
of structTreeParentId (added in PR #16916).
Now that modern JavaScript is fully supported also in the worker-thread we no longer need to keep old closures, which slightly reduces the size of the code.
Given that this is a shadowed getter, the `opMap` is already lazily initialized and it shouldn't be necessary to *also* use the `getLookupTableFactory` helper function here. Looking at the history of the code, it seems that this is simply a leftover from before JavaScript classes existed.
Now that modern JavaScript is fully supported also in the worker-thread we no longer need to keep old closures, which slightly reduces the size of the code.
Now that modern JavaScript is fully supported also in the worker-thread we no longer need to keep old closures, which slightly reduces the size of the code.
Now that modern JavaScript is fully supported also in the worker-thread we no longer need to keep old closures, which slightly reduces the size of the code.
Now that modern JavaScript is fully supported also in the worker-thread we no longer need to keep old closures, which slightly reduces the size of the code.
While this cache will not contain a huge amount of data in practice, it's nonetheless a *global* cache that currently will never be cleared.
This patch also removes the existing closure, since it shouldn't really be necessary nowadays given that the code is a JavaScript module which means that only explicitly listed properties will be exported.
When I started looking at PR 16938 it occurred to me that some of the new structTree-methods are synchronously accessing certain dictionary-data (not used during "normal" structTree-parsing), which may not be generally safe since everything in a dictionary could be a reference (and the relevant data may not have been loaded yet).
Rather than suggesting that we make all those new methods even more asynchronous, to me the overall simplest and safest solution is to ensure that the *entire* PDF document has been loaded *before* we begin saving it. In practice this shouldn't really affect "performance" of saving noticeably, since it's always depended on the entire PDF document being downloaded.
Finally note that with the exception of the PDF document possibly not having been fully downloaded when saving is triggered, all other "global" document properties are pretty much guaranteed to already be available at this point.
The unit test is re-enabled because it no longer seems to fail after 10
runs on Linux where this used to fail often. Code inspection also shows
that the code is correct and should raise the previous exception
(anymore). Finally, a lot has changed since this test was disabled such
as new Jasmine versions, new Linux bot OS version and new browser
versions.
While it makes sense to check that the `destDict` parameter is indeed a Dictionary, since that data comes from the PDF document itself, the `resultObj` parameter is an internal PDF.js implementation detail that should always be correct (or tests will fail).
Over time the amount of "document level" data potentially needed during parsing of Annotations have increased a fair bit, which means that we currently need to ensure that a bunch of data is available for each individual Annotation.
Given that this data is "constant" for a PDF document we can instead create (and cache) it lazily, only when needed, *before* starting to parse the Annotations on a page. This way the parsing of individual Annotations should become slightly less asynchronous, which really cannot hurt.
An additional benefit of these changes is that we can reduce the number of parameters that need to be explicitly passed around in the annotation-code, which helps overall readability in my opinion.
One potential drawback of these changes is that the `AnnotationFactory.create` method no longer handles "everything" on its own, however given how few call-sites there are I don't think that's too much of a problem.
The classes were stripped out during when creating the field name but
it led to a wrong name.
Since class components in a path are irrelevant, they're just ignored
when searching for a node in the datasets.
Focus callback must be called only when the element has been blurred.
For example, blur callback (which implies some potential validation) is not called
because the newly focused element is an other tab, an alert dialog, ... so consequently
the focus callback mustn't be called when the element gets its focus back.
While reviewing PR 16898 it occurred to me that it's currently impossible to trigger downloading of FileAttachment annotations using the keyboard.
Hence this patch adds `Ctrl + Enter` as the keyboard shortcut to download those, thus supplementing the existing double-clicking when using a mouse.
The goal is to always have something which is focusable to let the user select
it with the keyboard.
It fixes the mentioned bug because, the annotation layer will now have a container
to attach the canvas for annotations having their own canvas.
`.grab-to-pan-grab:active` is `#viewerContainer` when the mouse is
pressed down. It is supposed to have a `cursor: grabbing` appearance
immediately on mousedown,
`.grab-to-pan-grabbing` is the overlay that is supposed to cover
everything, and also has the `cursor: grabbing` appearance. The "cover
everything" result is achieved through `position:fixed`, `inset:0`, etc.
The block with these CSS properties for "cover everything" is currently
shared by `.grab-to-pan-grab:active` and `.grab-to-pan-grabbing`, but
only "cursor" need to be shared. The original JS and CSS code at
https://github.com/Rob--W/grab-to-pan.js shows that these were supposed
to be associated with the overlay only.
The PR that added this to PDF.js also shows that the "cover everything"
CSS properties were supposed to be limited to the overlay only:
https://github.com/mozilla/pdf.js/pull/4209#discussion-diff-9285917
But the final version of the PR mistakenly merged them together.
This patch rectifies that mistake.
The typescript compiler is now configured to know about the import map
to be able to resolve those imports and find the associated types.
As tsc outputs declaration files using the original module identifiers
and not the resolved ones, tsc-alias is used to post-process the
declaration files by resolving those paths.
Some configurations settings like `paths` cannot be provided through CLI
arguments but only in a configuration file. And when using a
configuration file, only a few options (like `--outDir` can still be
provided) through the CLI.
Using `removeNullCharacters` on the URL should be completely redundant, given the kind of data that we're passing to the `addLinkAttributes` helper function. Note that whenever we're handling a URL, originating in the worker-thread, in the viewer that helper function is always being used.
Furthermore, on the worker-thread all URLs are parsed with the `createValidAbsoluteUrl` helper function, which uses `new URL()` to ensure that a valid URL is obtained. Note that the `URL` constructor will either throw, or in some cases just ignore them, when encountering `\u0000`-characters during parsing.
Hence it should be *impossible* for a valid URL to contain `\u0000`-characters and we can thus simplify the viewer-code a tiny bit. The use of `removeNullCharacters` is most likely a left-over from back when `new URL()` wasn't generally available in browsers.
Testing the `tagged_stamp.pdf` document locally in the viewer, I noticed that e.g. the /Alt entry for the StampAnnotation contains "Secondary text for stamp\u0000".
Elsewhere in the viewer we're skipping null-chars and it's easy enough to do that in the `StructTreeLayerBuilder` class as well. (Note that we generally let the API itself return the data as-is.)
This fixes invalid type references (either due to invalid paths for the
import or missing imports) in the JS doc, as well as some missing or
invalid parameter names for @param annotations.
The issue described in the mentioned bug is reall because
Acrobat is rendering the XFA instead of the Acroform.
The original patch just tried to workaround the issue but it
induces some regressions.
This was added in PR 14899, over a year ago, however it's still completely unused in the PDF.js library/viewer. In hindsight I think that it was a mistake to add unused functionality, and the issue should probably have been WONTFIXed instead, however we probably can't just remove it now.
Thanks to the pre-processor, we can at least exclude this code in the *built-in* Firefox PDF Viewer.
After the changes in PR 16828 the `StampEditor` can now be initialized with a File, in addition to a URL, hence it seems that the `isEmpty` method ought to take that property into account as well.
Looking at this I also noticed that the assignment in the constructor may cause the `this.#bitmapUrl`/`this.#bitmapFile` fields be `undefined` which "breaks" the comparisons in the `isEmpty` method.
We could obviously fix those specific cases, but it seemed overall safer (with future changes) to just update the `isEmpty` method to be less sensitive to exactly how these fields are initialized and reset.
Currently we're repeating virtually the same code *four times* when fetching the bitmap-data, which seems unnecessary.
Also, ensure that the `#bitmapPromise` is always `null`ed by moving that into the `StampEditor.#getBitmapDone` method.
Currently this unit-test will pass just fine if compression is disabled, e.g. by commenting out the relevant code in the `src/core/writer.js` file.
While we don't have a simple way of *directly* checking that the Annotation text-content is compressed, we can however use the resulting file-size as a fairly good proxy. (Note that if compression is disabled the file-size is more than doubled.)
Please note that for performance reasons it's not really advised to use the same worker-thread *in parallel* for parsing multiple PDF documents, since they will then unnecessarily compete for resources.
However, given that it's still possible to do that e.g. when using the global `workerPort` it probably won't hurt to add a unit-test for this particular situation.
Given that the `PDFDocumentLoadingTask.destroy()`-method is documented as being asynchronous, you thus need to await its completion before attempting to load a new PDF document when using the global `workerPort`.
If you don't await destruction as intended then a new `getDocument`-call can remain pending indefinitely, without any kind of indication of the problem, as shown in the issue.
In order to improve the current situation, without unnecessarily complicating the API-implementation, we'll now throw during the `getDocument`-call if the global `workerPort` is in the process of being destroyed.
This part of the code-base has apparently never been covered by any tests, hence the patch adds unit-tests for both the *correct* usage (awaiting destruction) as well as the specific case outlined in the issue.
- The `src/core/unicode.js` exclude ought to have become unnecessary already with PR 16200, which significantly shortened and simplified that file.
- The `src/core/glyphlist.js` exclude no longer seems necessary in practice either, possibly because of improvements in Babel.
The main stamp button will be used to just enter in a add/edit image mode:
- the user can add a new image in using the new button.
- the user can edit an image in resizing, moving it.
In image mode, when the user clicks outside on the page but not on an editor,
then all the selected editors will be unselected.
After the `src/core/`-changes in PR 16779 the `PDFDocumentProxy.getJSActions` method should no longer be able to return *empty* entries, which means that we can simplify the "JavaScript support is not enabled"-warning in the viewer.
Furthermore, improve the auto-printing hack used when scripting is disabled.
Given that the FieldObjects are parsed in parallel, in combination with the existing caching in the `getPage`-method and `annotations`-getter, adding additional caches for this fallback code-path doesn't seem entirely necessary.
We're adding the action in the undo/redo stack whatever the status of the
operation was. This patch aims to add the action only when the image has been
successfully added.
When several editors are selected and the window loses and then gets back its focus,
the previously focused editor is triggering its focus callback making it the only
selected one.
This patch aims to avoid triggering the focus callback called when the main window
gets its focus back.
When moving an element in the DOM, the focus is potentially lost, so we need to make sure
that the focused element before the translation will get back its focus after it.
But we must take care to not execute any focus/blur callbacks because the user didn't
do anything which should trigger such events: it's a detail of implementation. For example,
when several editors are selected and moved, then at the end the same must be selected, so
no element receive a focus event which will set it as selected.
There are 2 rotation we've to deal with: the viewer one and the editor one.
The previous implementation was a bit complex and having to deal with these
rotation would have potentially increase it.
So this patch aims to simplify the implementation and deal with all the possible
cases.
The main idea is to transform the mouse deltas according to the rotations and then
apply the resizing in the page coordinates system.
When resizing an editor we're currently using unidirectional cursors, please refer to https://developer.mozilla.org/en-US/docs/Web/CSS/cursor
Given that editors can (generally) be resized to become either smaller or larger, it seems overall more appropriate to use bidirectional cursors to make this clearer to the user.
Note that as mentioned in the MDN article some environments, which seems to apply to e.g. Windows 11, doesn't differentiate between the two cursor formats and simply use bidirectional ones unconditionally.
One additional benefit of these changes is that the relevant CSS rules become slightly more compact.
We obviously don't want to re-introduce any `require` usage in e.g. the viewer, since we should strive to only use native `import` statements wherever possible.[1]
Hopefully exposing e.g. the library globally in more cases won't break anything, however it's somewhat difficult for me to imagine all the ways in which third-party users may be accessing the PDF.js library. (Given the lack of a runnable test-case in the issue, I also cannot guarantee that this is enough to fully address the problem.)
---
[1] Ideally we should probably not rely on e.g. `pdfjsLib` being globally available in the *built* viewer, and rather always `import` the library instead.
Unfortunately this would require larger (possibly breaking) changes in the builds that we provide, however note that Firefox only recently got support for `import` in workers and that Webpack still only have *experimental* support for outputting "proper" modules.
This method is very old, however with the exception of the auto-print hack (when scripting is disabled) in the viewer it's never actually been used.
Most likely the idea with `PDFDocumentProxy.getJavaScript` was that it'd be useful if scripting support was added, however it turned out that it was a bit too simplistic and instead a number of new methods were added for the scripting use-cases.
Without this patch the password dialog is pretty difficult to use in the GeckoView-viewer, because of a number of missing CSS variables.
*Please note:* This patch makes no effort at actually styling the dialog to better suite the overall look of the GeckoView-viewer, but focuses solely on making it actually usable (since password protected PDF documents are somewhat rare).
If the current PDF document is closed while the password dialog is open, e.g. manually by calling `PDFViewerApplication.close()` from the console, the password dialog wouldn't be closed as intended.
*Please note:* This could only affect the GENERIC viewer, although it's very unlikely to ever happen, since that's the only one that supports opening more than one PDF document.
*Please note:* This situation should never happen in practice, but it nonetheless cannot hurt to fix this.
If the `PasswordPrompt.open` method would ever be called synchronously back-to-back *and* if opening of the dialog fails the first time, then the second invocation would remain pending indefinitely since we just clear out the capability.
Given that the `useOnlyCssZoom` option is essentially just a special-case of the `maxCanvasPixels` functionality, we can combine the two options in order to simplify the overall implementation.
Note that the `useOnlyCssZoom` functionality was only ever used, by default, in the PDF Viewer for the B2G/FirefoxOS project (which was abandoned years ago).
When searching for "endobj"-operators, make sure that we don't accidentally match a "trailer"-string in /Content-streams without /Filter-entries (i.e. streams that contain "raw" and thus human-readable data).
Currently we accidentally accept `cMapUrl` and `standardFontDataUrl` parameters that are empty strings or `null`, since e.g. `new URL(null, document.baseURI)` doesn't throw, when validating the `useWorkerFetch` parameter via the `isValidFetchUrl` helper function.
Please note that we are currently failing gracefully in this case, as intended, however the warning-messages printed in the console are perhaps less helpful without this patch.
When an editor is selected in using the keyboard then it has the focus.
But then if the editor is unselected with Escape key then the focus must
be removed otherwise we still have a blue outline around it.
And add few missing timeout in the integration tests.
This is quite old code, however the error-handling no longer seems necessary for a couple of reasons:
- The `PDFViewerApplication.open` method is asynchronous, which means that it cannot throw a "raw" `Error` and the try-catch is not needed in that case.
- None of the other affected methods should throw, and if they do that'd rather indicate an *implementation* error in the code.
- Finally, and most importantly, with the `PDFViewerApplication.run` method now being asynchronous an (unlikely) `Error` thrown within it will lead to a rejected `Promise` and not affect execution of other code.
We can use modern JavaScript features, in this case optional chaining, to (ever so slightly) simplify how `ViewHistory` errors are handled.
Also, use arrow functions when handling a few other (very rare) errors during loading since that's a tiny bit shorter.
The way that the callback-methods are specified feels unnecessarily verbose, however we can introduce a short-hand to improve this.
Also, adds a couple of new-lines to improve overall readability.
Selected editors can be moved in using the arrows:
- up/down/left/right will move the editors of 1 in page unit;
- ctrl (or meta)+up/down/left/right will move them of 10 in page unit.
The keyboard shortcuts (copy, paste, ...) didn't work correctly when the
main container was not focused.
This patch adds few waitForTimeout in the integration test for FreeText
in order to avoid possible intermittent failures.
Given that the `debugger` is loaded as a module we can use "top level await" in development mode to access the necessary API-functionality, which removes the need to manually pass in the required properties.
- it'll improve the way to resize images: diagonally (in keeping ratio between dimensions)
or horizontally/vertically.
- the resizer was almost invisible in HCM.
- make a resize undoable/redoable.
In order to reproduce the original issue:
- switch to freetext mode
- add a text somewhere
- double click outside and add some text
- repeat the previous step several times
no text is selected during the edition.
The existing Node.js-specific polyfills depend on the `node-canvas` package, which has unfortunately (repeatedly) shown to cause trouble for many users. We attempted to improve the situation by listing the relevant packages as `optionalDependencies`, but that didn't seem to really fix the problem.
With this patch the library should be able to load in Node.js-environments even if polyfilling fails, and any errors will instead occur during rendering. Obviously this is *not* a proper solution, since it basically moves the problem to another part of the code-base.
However for certain "simpler" use-cases, such as e.g. text-extraction, these changes should hopefully improve general usability of the PDF.js library in Node.js-environments.
*Please note:* For most PDF documents rendering should still work though, since `DOMMatrix` is *currently* only used with Patterns and `Path2D` only with Type3-fonts and Patterns.
In Gulp 4, which we use for years now, the `gulp.src()` function
supports the `removeBOM` option to disable the default BOM stripping,
so this commit uses that to get rid of our `vinyl-fs` dependency.
Note that this actually makes disabling BOM stripping work again. It's
currently broken because in `vinyl-fs` 3, that we already use since 2018
in commit 95de23e, the `stripBOM` option was renamed to `removeBOM`, so
the current code doesn't actually disable BOM stripping which we now
confirmed and sadly broke for years without anyone noticing. Most likely
this is because the BOM is not required for UTF-8 documents, but while
not necessary it also can't hurt to have it for tools that use it to
determine if a document is UTF-8.
*Please note:* This only removes the preference itself, however both the viewer-option and the actual implementation is still available.
The `useOnlyCssZoom` functionality was only ever used, by default, in the PDF Viewer for the B2G/FirefoxOS project (which was abandoned years ago). Given that CSS-only zooming can easily make the document look blurry even at low zoom levels, this functionality was only intended for low-powered mobile devices.
Hence it seems reasonable to remove the `useOnlyCssZoom` preference now, since neither the default viewer nor the GeckoView-specific viewer uses this functionality.
Trying to update Stylelint to version `15.10.1`, and beyond, broke linting. Looking at the changes the issue appears to be that the `bin/stylelint.js` file was replaced with `bin/stylelint.mjs` instead, which our `gulp lint` runner wasn't able to automatically find; see https://github.com/stylelint/stylelint/compare/15.10.0...15.10.1
When the flag is set, the appearance has to be generated from the value so it's
useless/meaningless to extract the content from the existing appearance.
When a pdf has /NeedAppearances set to true, the annotation appearance must be
generated from its value and we must take into account the hasOwnCanvas property.
*Please note:* I'm not aware of any bugs caused by this, however that might be more luck than anything else.
In PR 16392 the `incrementalUpdate` function, and all of its various helpers, were made asynchronous. However the call-site in `src/core/worker.js` wasn't updated, which means that we currently reset temporary XRef-entries while saving is ongoing.
By leveraging import maps we can get rid of *most* of the remaining `require`-calls in the `src/display/`-folder, since we should strive to use modern `import`-statements wherever possible.
The only remaining cases are Node.js-specific dependencies, since those seem very difficult to convert unless we start producing a bundle *specifically* for Node.js environments.
With the changes in the previous patch the `isNodeJS`-helper no longer needs to live in its own file, which helps get rid of a closure in the *built* files.
In the last couple of years we've been quicker to remove support for older browsers/environments, which means that at this point in time we don't bundle that many polyfills. (The polyfills are also generally simpler nowadays, ever since we removed support for e.g. Internet Explorer.)
Rather than having to *manually* handle the polyfills, we can actually let Babel take care of bundling the necessary polyfills for us; please refer to https://babeljs.io/docs/babel-preset-env
The only exception here is the Node.js-specific compatibility-code, which is moved into the `src/display/node_utils.js` file. This ought to be fine since workers are not available/used in Node.js-environments.
*Please note:* For the `legacy`-builds this will increase the size of the *built* files, however that seems like a very small price to pay in order to simplify maintenance of the general PDF.js library.
Localization of this button broke in PR 16340, which I assume was completely accidental, since the download-button now tries to access a l10n-id that was removed some time ago (see PR 15617).
Note how loading even the development viewer, i.e. http://localhost:8888/web/viewer-geckoview.html#locale=en-US, currently logs l10n-warnings on the `master` branch.
Having a `require` in this file has never made sense in e.g. the Firefox PDF Viewer and shouldn't really be necessary.
Possibly the idea was to facilitate some kind of third-party bundling, however the *built* `pdf.js` file has always exposed the API-contents globally.
Currently this class contains a few "special" code-paths for the COMPONENTS build-target, which normally wouldn't be a problem. However, in this particular case that means accessing code that we don't want to include unconditionally in all builds.
This is currently implemented using build-time `require`-calls which we nowadays want to avoid, and we should strive to remove all such cases from the code-base. (Generally speaking `import` is the future, and build-tools may not always play well with a mix of both formats.)
We can easily improve things here by using sub-classing for the COMPONENTS build-target, and then use the ability to re-name when exporting (to avoid breaking existing code).
Occasionally some test-suites may fail to start on the bots, however that's not correctly reflected in the botio-output posted to GitHub which makes it easy to accidentally overlook this situation.
Looking at the raw logs when that happens they always seem to contain a line such as `Run NaN tests` which means that we should be able to easily make this situation a *failure* as intended.
In order to reproduce the issue:
- scale down the image
- zoom the page and the image is pixellated
So this patch allow to redraw the image when zooming.
- Take into account the page translation,
- Take into account the correct translation for the editor border,
- Take into account the position of the first glyph in the annotation,
- Take into account the rotation of the editor.
Close#16633.
There's no good reason for getting this option multiple times in the same method. Also, we can slightly re-factor how the `editorStampButton` is made visible.
- Do the /Filter and /DecodeParms lookup in parallel, since that ought to be a *tiny* bit more efficient.
- Avoid code-duplication when `CompressionStream` isn't supported, since we already have a fallback code-path at the end of the function.
This regressed in PR 16659, when the signature of the `PDFViewer.annotationEditorMode`-setter was changed, and it currently leads to an Error being thrown when exiting PresentationMode.
Unfortunately I wasn't able to come up with a *simple* way to just replace the synchronous `require`-call, since we need to ensure that the default preferences are available when bundling starts.
Hence this patch adds a new intermediate parsing-step in all the relevant gulp-tasks, but this shouldn't affect build-times noticeable since the amount of extra parsing is very small.
*Please note:* It's very possible that there's a better way to handle this, however I figured that unblocking further ESM-work is more important than a "perfect" solution.
createImageBitmap doesn't work with svg files (see bug 1841972), so we need to workaround
this in using an Image.
When printing/saving we must rasterize the image, hence we get the biggest bitmap as image
reference to avoid duplications or poor quality on rendering.
The existing code is unable to *correctly* extract the color from the appearance-stream when the ColorSpace-data is "complex". To reproduce this:
- Open `freetexts.pdf` in the viewer.
- Note the purple color of the "Hello World from Preview" annotation.
- Enable any of the Editors.
- Note how the relevant annotation is now black.
Note how we're accidentally using the wrong operator when trying to parse CMYK colors. I'm not aware of any bugs caused by this, since it seems uncommon in practice for annotations to specify text-colors in CMYK format.
When there was a rotation, the generated bbox was wrong because of an inversion
between width and height.
This patch aims to fix this issue in re-writing the FreeText code generation
to have something similar to what Acrobat does.
And fix the name of the font which wasn't the correct one when calling the
evaluator.
- Update the "Getting the Code" section to specifically mention Mozilla Firefox, since while the development viewer *works* it may look slightly "broken" in Chromium-based browsers. (This is caused by a lack of support for unprefixed CSS properties, e.g. `mask-image`, however this does *not* affect the built PDF.js viewer.)
- Remove the Twitter-link, since that account has not been updated since 2016 (i.e. over seven years ago).
Semantically, it is more correct to encode the fragment in the URL
instead of the URL-encoded `file` query parameter. This shouldn't matter
in practice, because `rewriteUrlClosure` in `chromecom.js` decodes the
`file` parameter and restores the fragment. However, as #16625 shows,
there was a case where this did not work as expected.
`PDFViewerApplication` reads from `location.hash` to initialize
`initialBookmark`. But when extensions/chromium/pdfHandler.js prepares
the redirect URL, the reference fragment is encoded instead of bare.
`rewriteUrlClosure` in `chromecom.js` is responsible for decoding the
URL, but that currently runs too late.
To fix this, update `initialBookmark` after rewriting the URL.
This was not a problem in the past because `rewriteUrlClosure` in
`chromecom.js` executed before the initialization of `initialBookmark`.
Given that the PDF.js library has never officially supported/documented that binary data can be provided as a `Buffer`, and that it's been explicitly deprecated in *four* releases, it seems reasonable that we outright reject such data instead (to reduce the amount of Node.js specific code-paths).
We've now been throwing an Error in *three* releases if the `canvasFactory` option is provided, hence it ought to be fine to stop doing that and simply ignore the option instead.
Rather than having to *manually* determine the potential `transfers` at various spots in the API, we can let the `AnnotationStorage.serializable` getter include this.
To further simplify things, we can also let the `serializable` getter compute and include the `hash`-string as well.
These options are completely unused in the PDF.js viewer, and given that the last update of the `GrabToPan`-code from upstream was in 2016 it shouldn't hurt to remove them.
Before this commit, lint-chromium complained without an obvious course
of action:
> Warning: Pref objects doesn't have the same length.
> Error: chromium/preferences_schema is not in sync
With this commit, the error message is more actionable:
> Warning: extensions/chromium/preferences_schema.json does not contain an entry for pref: enableFloatingToolbar
> Error: chromium/preferences_schema is not in sync
This is something that I completely overlooked during review of PR 16593, since the idea is (obviously) that the viewer-components should be usable as-is without the user needing to manually pass in any *additional* parameters.
To support this we can very easily expose the current `FilterFactory`-instance on the `PDFPageProxy`-class[1], and if needed initialize the highlight-filters when initializing the page (again limited to the viewer-components).
In order to minimize the size the of a saved pdf, we generate only one
image and use a reference in each annotation using it.
When printing, it's slightly different since we have to render each page
independantly but we use the same image within a page.
- Modify the text and background colors in popup to fit a11y requirements
- Add a backdrop filter on clickable areas in using a svg filter mapping
canvas colors to Highlight and HighlightText ones.
It occurred to me that we can actually run this unit-test in Node.js environments by making use of the preprocessor to stub out the browser globals there.
Given that nullish coalescing is now available in all environments/browser that we support, we can (ever so slightly) simplify handling of the `TESTING` build-target.
Until now we've not actually had *any* tests that ensure that the *official* PDF.js-viewer API exposes the intended functionality, which means that things can easily break accidentally.
*Please note:* This unit-test cannot (easily) be run in Node.js-environments, since the `external/webL10n/l10n.js` file contains various browser-specific functionality.
These constants were added "speculatively" in PR 10820, almost four years ago, but have never actually been used. We already have issue 10982 that tracks *potentially* extending support for the affected annotation-format, however until that happens I really don't think that we should keep shipping completely unused code in the PDF.js library.
For the MOZCENTRAL build-target, i.e. the Firefox PDF Viewer, this reduces the total bundle size by 1.1 kilo-byte.
Until now we've not actually had *any* tests that ensure that the *official* PDF.js API exposes the intended functionality, which means that things can easily break accidentally.
- Change (most) fields/methods into private ones, since that's now supported.
- Tweak the constructor-parameters, and simplify the sandbox initialization w.r.t. the viewer components.
- Remove some unused function/method parameters.
- Slightly simplify the "updatefromsandbox"-handler by using local variables and inverting some conditions.
Rather than sprinkling pre-processor statements throughout the viewer-code, simply "disable" the relevant `PDFViewer` setters instead.
Also, given that the GeckoView-specific viewer doesn't have a sidebar we don't actually need to explicitly ignore a `pageMode` during loading.
This helper function was added almost two years ago, in PR 13696, and it still has only a single call-site. Furthermore, with the changes made in PR 16572 it also cannot hurt to reduce the size of the `web/l10n_utils.js` file slightly.
Note how the [`ChromeActions.getPreferences` method](https://searchfox.org/mozilla-central/rev/4e8f62a231e71dc53eb50b6d74afca21d6b254e9/toolkit/components/pdfjs/content/PdfStreamConverter.sys.mjs#497-530) returns the preferences as a string, which we then have to convert back into an Object in the viewer.
Back when that code was originally written it wasn't possible to send Objects from the platform-code, however that's no longer the case and we should be able to (eventually) remove this unnecessary string-parsing now.
*Please note that in order to prevent breakage we'll need to land these changes in stages:*
- Land this patch in mozilla-central, as part of regular the PDF.js updates.
- Change the return type in the `ChromeActions.getPreferences` method, in a mozilla-central patch.
- Remove the string-handling from the `FirefoxPreferences._readFromStorage` method.
Please note that we've never had any functionality in the viewer itself that *set* preferences, and we've thus only ever read them.
For the GENERIC viewer it obviously makes sense for the user to be able to modify preferences, e.g. via the console, but that doesn't really apply to the *built-in* Firefox PDF Viewer since preferences are already accessible via `about:config` there. Hence it does seems somewhat strange to expose, a limited part of, the Firefox preference system in this way when we're not even using it.
Note that the unused preference setting-code also include a fair amount of *additional* validation on the platform-side, such as limiting any possible preference changes to the `pdfjs.`-branch and also an explicit white-list of preference names[1], to make sure that this is safe; please see:
- https://searchfox.org/mozilla-central/rev/4e8f62a231e71dc53eb50b6d74afca21d6b254e9/toolkit/components/pdfjs/content/PdfStreamConverter.sys.mjs#458-495
- https://searchfox.org/mozilla-central/rev/4e8f62a231e71dc53eb50b6d74afca21d6b254e9/toolkit/modules/AsyncPrefs.sys.mjs#21-48
Assuming that this patch lands, I'll follow-up with a mozilla-central patch to remove the code mentioned above.
---
[1] This hard-coded list contains preferences that no longer exist, and also at least one (fairly obvious) typo.
This method was added only for consistency with the `register`-method, however it's never actually been used. To avoid including dead code in the builds, let's just remove the `unregister`-method for now.
*Please note:* If this method ever becomes useful, it'll be trivial to revert this commit.
With the changes in PR 16552 we can now move general translation into the `AnnotationLayer` itself, which should improve things ever so slightly in third-party implementations where the default viewer isn't used.
*This is something that I completely overlooked during review of PR 16552, despite leaving a l10n-related comment.*
The new l10n-handling of PopupAnnotations assume that the `AnnotationLayer` is always initialized with a l10n-instance, which might not actually be the case in third-party implementations where the default viewer isn't used.
To work-around that we'll now bundle, and fallback on, the existing `NullL10n`-implementation in GENERIC builds of the PDF.js library. This will only result in a slight file-size increase for the *built* `pdf.js` file, again limited to GENERIC builds, since the `web/l10n_utils.js` file has no dependencies.
Also, tweaks a couple of TESTING pre-processor checks to *only* include that code when running the reference tests.
- it'll help to be able to move popups on screen to let the user read the text
- popups won't inherit some properties from their parent:
- the popup can be misrendered if for example the parent has a clip-path property.
- add an outline to the popup when the parent is focused.
- hide a popup when it's clicked.
Fix handling of /Filter-entries, since the current implementation could potentially corrupt the data if there's multiple filters present.
Please note that filters are applied *sequentially* during decoding, starting from the first one in the Array, hence the first Array-entry needs to be /FlateDecode in order for things to actually work correctly.
To prevent a future bug, if we want to save more "complex" data such as images, also ensure that we include any existing /DecodeParms-entries when updating the /Filter-entry.
The existing unit-test doesn't work as intended, since the page never actually renders. Note how `cleanup` is *not* allowed to run when parsing and/or rendering is ongoing, however an (old) incorrect condition could prevent rendering from ever starting.
This is very old code, which has been slightly re-factored a couple of times (many years ago), however this doesn't appear to affect e.g. the default viewer since the incorrect behaviour seem highly dependent on "unlucky" timing.
Note also how at the start of the `PDFPageProxy.prototype.render`-method we purposely cancel any pending `cleanup`-call, to prevent unnecessary re-parsing for multiple sequential `render`-calls.
Finally, avoid running `cleanup` when document/page destruction has already started since it's pointless in that case.
After PR 16226 the deprecated SVG back-end is now unused in development mode, with the exception of unit-tests, hence we can re-factor how it's exposed in the API to avoid including a useless webpack-closure in e.g. the *built-in* Firefox PDF Viewer.
Given that this API method isn't used anywhere within the PDF.js library itself, except for the unit-tests, we can avoid including what's effectively dead code in e.g. the *built-in* Firefox PDF Viewer.
- Don't attempt to lookup an "SM" entry, since we're only using "SMask" in the `PDFImage` code and I also cannot find any mention in the PDF specification about that being a valid abbreviation for a Soft Mask entry. (There's only a `SM = Smoothness Tolerance` Graphics State parameter, which is obviously something completely different.)
- Don't lookup the /SMask and /Mask entries unless it's actually an inline image, since it's pointless otherwise.
- Last, but most importantly, only check for the *existence* of /SMask and /Mask entries but don't actually fetch the data. Note that if either one exists it'll contain a Stream, and those cannot be cached on the `XRef`-instance, which leads to unnecessary parsing/allocations and in this case we're not using the actual data for anything.
The original `trimCache` functionality was intended to be exposed on the
top-level `puppeteer` module, but due to a bug in Puppeteer this didn't
work correctly and we had to call `trimCache` on the default Puppeteer
node instance instead, which was fortunately exposed. However, since
this didn't feel like intended API usage, this bug was reported and is
now fixed in Puppeteer 20.5.0, so this commits updates Puppeteer to that
version so we can use the intended API.
The full history of this issue can be found at
https://github.com/puppeteer/puppeteer/issues/10174.
This patch is the result of me going through some old issues regarding non-embedded Wingdings support.
There's a few different things wrong in the referenced PDF document:
- The /BaseFont and /FontName entries don't agree on the name of the fonts, with one font using `/BaseFont /Wingdings-Regular` and `/FontName /wg09np` which obviously makes no sense.
To address this we'll compare the font-names against our lists of known ones and ignore /FontName entries that don't make sense iff the /BaseFont entry is a known font-name.
- The non-embedded Wingdings font also set an incorrect /Encoding, in this case /MacRomanEncoding, which should have been fixed by PR 16465. However this doesn't work since the font has *bogus* font-flags, that fail to categorize the font as Symbolic.
To address this we'll also compare the font-name against the list of known symbol fonts.
As far as I can tell there's no particular reason for initializing `KeyboardManager`-instances eagerly, since the user may never use editing, and we can easily do this lazily instead by utilizing shadowed getters.
While it's slightly difficult to trigger in practice, unless the `defaultZoomDelay`-value is increased, it's currently possible to generate thumbnails from *partially* rendered pages when doing *temporary* CSS-only zooming.
We shouldn't dispatch a "pagerendered"-event when doing *temporary* CSS-only zooming, but simply wait until the actual rendering is done.
While I don't believe that this regression has caused any actual bugs, dispatching *duplicate* events is nonetheless inconsistent and should be fixed.
This patch updates the minimum supported browsers as follows:
- Google Chrome 92, which was released on 2021-07-20; see https://support.google.com/chrome/a/answer/10314655
Note that nowadays we usually try, where feasible and possible, to support browsers that are about two years old. By limiting support to only "recent" browsers we reduce the risk of holding back improvements of the *built-in* Firefox PDF Viewer, and also (significantly) reduce the maintenance/support burden for the PDF.js contributors.
*Please note:* As always, the minimum supported browser version assumes that a `legacy`-build of the PDF.js library is being used; see https://github.com/mozilla/pdf.js/wiki/Frequently-Asked-Questions#faq-support
Given that this functionality is only relevant in third-party use-cases, for example the viewer-components, we can avoid needlessly including it in e.g. the MOZCENTRAL build.
This commit makes the following required changes:
- Replace custom cache trimming logic in favor of the (per our request)
newly added `trimCache` method in Puppeteer. Not only does this greatly
simplify our code and prevents having to import Puppeteer internals,
it's also necessary because Puppeteer 20 removed the `BrowserFetcher`
API in favor of the new separate `@puppeteer/browsers` package.
- Start browsers in series instead of in parallel. Parallel browser
starts broke since Puppetter 19.1.0 and it turns out that it has never
been supported officially, so it worked more-or-less by accident.
Starting browsers in series is the supported way, is almost equally
fast and ensures that we avoid any race conditions during startup.
Finally, it also allows us to remove the `browserPromise` state on our
session objects.
Fixes#15865.
This patch does two things:
- Moves the updating of thumbnails into `web/app.js`, via a new `PDFSidebar` callback-function, to avoid having to include otherwise unnecessary parameters when initializing a `PDFSidebar`-instance.
- Only attempt to generate thumbnail-images from pages that are *cached* in the viewer. Note that only pages that exist in the `PDFPageViewBuffer`-instance can be rendered, hence it's not actually meaningful to check every single page when updating the thumbnails.
For large documents, with thousands of pages, this should be a tiny bit more efficient when e.g. opening the sidebar since we no longer need to check pages that we know have not been rendered.
The way that the cleanup was implemented in PR 12613 has always bothered me slightly, since the `isPageCached`-method that I introduced there always felt quite out-of-place in the `IPDFLinkService`-implementations.
By introducing a new "thumbnailrendered" event, similar to the existing "pagerendered" one, we're able to move the cleanup handling into the `PDFViewer`-class instead.
The way that this was implemented in PR 10217 has always bothered me slightly, since the `isPageVisible`-method that I introduced there always felt quite out-of-place in the `IPDFLinkService`-implementations.
Hence this is instead replaced by a callback-function in `PDFFindController`, to handle the page-visibility checks. Note that since the `PDFViewer`-constructor always sets this callback-function, e.g. the viewer-component examples still work as-is.
Now that font-substitution has been implemented, we should be able to do much a better job at supporting non-embedded Wingdings fonts.
Given that this is a Windows-specific font, see https://en.wikipedia.org/wiki/Wingdings, this is however not guaranteed to work (well) on other platforms.
The affected font is non-embedded ZapfDingbats, however the PDF document for some inexplicable reason specifies the encoding as "WinAnsiEncoding" (which is obviously wrong).
To work-around this bug in the PDF generator, we'll simply ignore any explicitly specified named encoding for non-embedded symbol fonts.
- Remove the dependency on fit-curve;
- Improve the way to draw the current line in using a Path2D and
in clearing only the last part of the curve instead of clearing
all the canvas;
- Smooth the curve when drawing to avoid to have some changes after
the drawing ends;
- Make the smoothing a bit less agressive.
Given that inline images may contain "EI"-sequences in the image-data itself, actually finding the end-of-image operator isn't always straightforward.
Here we extend the implementation from PR 12028 to potentially check all of the following bytes, rather than stopping immediately. While we have fairly decent test-coverage for this code, whenever you're changing it there's unfortunately a slightly higher than normal risk of regressions. (You'd really wish that PDF generators just stop using inline images.)
- if the contours count is lower than -1, the glyph is really likely wrong
so just remove it from the font;
- if a contour has the repeat flag then repeats count mustn't be 0.
Looking at the behaviour in Adobe Reader it doesn't appear that attachments are sorted alphabetically, hence it doesn't seem necessary for us to do so either in the viewer.
An additional benefit of *not* sorting the attachments is that any "actual" attachments are now always placed at the top of the list in the sidebar, and if any `FileAttachment`-annotations exist in the document they will now be appended at the end.
The pdf linked in bug 1135277 contains a lot of stroke instructions.
In using the Firefox profiler, this patch helps to reduce the overall
spent time in this function by 30%.
According to https://en.wikipedia.org/wiki/Impact_(typeface) this font should be available on all current versions of Windows, and with the recently added font-substitution we should actually be able to render it correctly (at least on Windows).
The `fontID` handling is quite old and predates the use of the `idFactory` to generate a unique id for each font, hence we can simplify this code a little bit.
When fixing bug 1766987, I thought the field formatted value came from
the result of the format callback: I was wrong. The format callback is ran
but the value is unused (maybe it's useful to set some global vars... or
it's just a bug in Acrobat). Anyway the value to display is the one rendered
in the AP stream.
The field value setter has been simplified and that fixes issue #16409.
This essentially extends PR 11218 to also apply when looking up the final font-reference, via the XRef-table, fails because the font isn't available.
This patch also changes `PartialEvaluator.fallbackFontDict` to simply use "Helvetica" as the default font-name, since that seems generally reasonable given the now existing font-substitution code.
After PR 12563 we're now free to use optional chaining in the worker-thread as well. (This patch also fixes one previously "missed" case in the `web/` folder.)
For the MOZCENTRAL build-target this patch reduces the total bundle-size by `1.6` kilobytes.
Given that the `css` property isn't constant, since it contains document/font ids, we cannot just check it directly. However, we can make use of regular expressions to ensure that the format is generally correct.
Despite this being a *major* version increase, it doesn't appear to require any updates in our test-suites.
Note in particular that the minimum supported browsers/environments were updated, however this isn't a problem given our recent support-changes in the PDF.js library.
Please find additional details at https://github.com/jasmine/jasmine/blob/main/release_notes/5.0.0.md
Originally the `PDFSidebarResizer` class was slightly larger, since the code used to contain e.g. feature testing for older (and no longer supported) browsers.
Given that there's some amount of overlap, when it comes to what DOM-elements and state that these classes need, it now seems reasonable to simply move the sidebar-resizing into the `PDFSidebar` class.
For the MOZCENTRAL build-target this patch reduces the size of the *built* `web/viewer.js` file by just over `1.1` kilobytes.
Similar to other toolbar/secondaryToolbar buttons that open toolbars or dialogs, it seems reasonable to use "aria-controls" for the editor-toolbar buttons as well.
On my computer, it takes few tenths of a second to load a local font.
Since a font can be used several times in a document, the cache will
improve performances.
- Replace FoxitSans with LiberationSans: LiberationSans is already there (for XFA) and we can use
it as a good replacement of FoxitSans.
- For now we just try to substitue standard fonts, the strategy is the following:
* we try to find a font locally from a hardcoded list;
* if it fails then we use Liberation as fallback (only for Helvetica for the moment);
* else we just fallback on the system serif/sansserif/monospace font.
This patch updates the minimum supported browsers as follows:
- Safari 15.4, which was released on 2022-03-15; see https://en.wikipedia.org/wiki/Safari_version_history#Safari_15
Nowadays we usually we try, where feasible and possible, to support browsers that are about two years old. The reasons for limiting support to a *somewhat* more recent Safari version include:
- Throughout the history of the PDF.js project, Safari has always been the worst browser to attempt to support. Compared to other browsers there's a disproportionate number of bugs affecting Safari, especially on iOS, and in most cases those are browser-specific issues that we simply cannot address.[1]
- Safari has often been a lot slower, compared to other browsers, at implementing new web-platform features. Historically this has sometimes blocked usage of new features, for the benefit of the Firefox PDF Viewer, and it's very often meant having to include and maintain polyfills *only* for Safari.
- The current (minimum) supported Safari version lack enough functionality that polyfills placed in the `src/shared/compatibility.js` file are unfortunately not sufficient, but it also requires a bunch of special-cases in both the `gulpfile` and in the `web/`-code.
- Given that the *built-in* Firefox PDF Viewer is the primary development target for the PDF.js library, and the general development pace these days, we need to limit the maintenance "overhead" caused by other browsers.
---
[1] In a few cases a work-around might be possible, however it'd negatively affect e.g. performance, readability, and/or maintainability of the code.
Originally we only used the `structuredClone` polyfill in the `LoopbackPort`-implementation, and that obviously isn't used anywhere within the various image decoders.
At this point in time we've started to use `structuredClone` a little bit more, hence it seems overall simpler to just bundle the polyfill even in the `legacy`-version of the IMAGE_DECODERS built-target.
For some time these checks have only targeted Node.js environments, since the features in question exist in all supported browsers (even when a `legacy`-build is used).
Now that we've updated the minimum supported Node.js version to 18, a number of polyfills are thus (finally) no longer necessary in that environment. Hence for certain *basic* functionality, such as e.g. text-extraction, it's now possible to use either a modern- or a `legacy`-build of the PDF.js library in Node.js environments.
*Please note:* For e.g. canvas-rendering in Node.js environments it's still necessary to use a `legacy`-build, since that functionality requires various polyfills.
This patch updates the minimum supported environments as follows:
- Node.js 18, which was released on 2022-04-19; see https://en.wikipedia.org/wiki/Node.js#Releases
Note also that Node.js 16 will soon reach EOL, and thus no longer receive any security updates.
The /Decode-implementation in the our JPEG decoder, i.e. `src/core/jpg.js`, seems to only handle *inverting* of images properly. To support arbitrary /Decode-entries correctly we'll always use the `PDFImage.decodeBuffer` method, even for "simple" JPEG images, which should be fine since non-default /Decode-entries aren't a very common occurrence.
*Please note:* This patch will lead to a little bit of movement in some existing test-cases, however it should be virtually imperceivable to the naked eye.
The fallback code-path has never really been used, since the `PDFSidebar` is only used in the default viewer (and has never been exposed in e.g. the COMPONENTS-build).
This patch tries to simplify, and improve, the thumbnail styling:
- For rendered thumbnails there's one less DOM-element per thumbnail, which can't hurt in longer documents.
- Use CSS-variables to set the dimensions of all relevant DOM-elements at once.
- Simplify the visual styling of the thumbnails, e.g. remove the border since the viewer no longer has visible borders around pages, since the relevant CSS-rules are quite old code.
These changes also, at least in my opinion, makes the relevant CSS-rules much easier to understand and work with.
- Make it easier to work on e.g. [bug 1690428](https://bugzilla.mozilla.org/show_bug.cgi?id=1690428) without affecting the other sidebarViews.
This property was added in PR 12726 specifically for use in the `getFontType` function, indirectly used by the `PDFDocumentProxy.stats` getter in the API.
In PR 15880 that functionality was removed, but I forgot to remove this now unused font-property.
Now that we no longer depend on the old Babel version in SystemJS we can remove the `static get ...` work-arounds used to define constants, which leads to slightly more compact code.
Now that https://bugzilla.mozilla.org/show_bug.cgi?id=1247687 has landed in Firefox, we're able to use worker-modules during development :-)
This removes the final piece of SystemJS usage from the PDF.js library, thus allowing a fair bit of clean-up, and we now use *only* native `import`/`export` statements everywhere in development mode.
When the `GlobalImageCache` implementation originally landed, back in PR 11912, the image handling was slightly more complex (with e.g. browser-decoding of some JPEG images). At this point it no longer seems necessary to manually handle pageIndexes in this way, and we should be able to simply inline that in the `GlobalImageCache.shouldCache` method.
This commit migrates this functionality away from the bots. Nowadays
it's possible to build and deploy the website to GitHub Pages directly
through the GitHub Actions, which provides a nice simplication of the
process. Not only does this remove the requirement to have a `gh-pages`
branch in the repository, it also avoids the complexity of having to
configure the workflow to commit to Git branches and allows us to remove
the Git committing code from the Gulpfile.
Note that deploying directly though GitHub Actions workflows needs to be
enabled in the repository settings, but this is easy and well documented
on the link below.
The following resources are relevant for this patch:
- Enabling deployment to GitHub Pages directly through GitHub Actions:
https://docs.github.com/en/pages/getting-started-with-github-pages/configuring-a-publishing-source-for-your-github-pages-site#publishing-with-a-custom-github-actions-workflow
- Uploading GitHub Pages artifacts example:
https://github.com/actions/upload-pages-artifact#usage
- Deploying GitHub Pages artifacts example:
https://github.com/actions/deploy-pages#usage
This patch tries to mimic the look of the message-element in the Firefox browser-findbar, and thus makes the following changes:
- Remove the red colour, since it didn't take the light/dark themes into account.
- Display the "notFound" message in bold.
The `pdfjs-dist/lib/` directory contains a README file that explicitly advises against using those files, however based on a fairly large number of issues filed over the years users seem to be (mostly) overlooking that warning.
In particular it unfortunately seems to be somewhat common for users to attempt to "combine" proper builds from `pdfjs-dist/build/` together with individual components from the `pdfjs-dist/lib/web/` directory, which more often than not leads to subtle bugs and general problems.
When we receive bug reports about this it's often not immediately obvious what the problem is, given that many issues lack enough details (such as runnable test-cases), but after some back-and-forth it usually turns out that usage of `pdfjs-dist/lib/` is the culprit.
Considering that keeping the general PDF.js library working is challenging and time-consuming enough nowadays, this patch thus proposes that we stop including the "lib"-build in the `pdfjs-dist` repository to both reduce user confusion and the support burden.
Currently we only prevent triggering the actual text-extraction multiple times in "parallel", when using the "copy all text" feature, however the "copy"-event itself is not prevented.
The result is that if the user selects all text in a long PDF document and then uses the copy-shortcut multiple times in quick succession, we'll actually populate the clipboard with "incomplete" contents (via a `TextLayerBuilder` copy-listener) until all text-extraction finishes.
In PR #16295 one occurrence of this was changed, but a few more remained
in the codebase. This commit fixes the other occurrences so that we
don't use the deprecated way of creating custom events anywhere anymore.
According to MDN, see https://developer.mozilla.org/en-US/docs/Web/API/CustomEvent/initCustomEvent,
using the `CustomEvent.initCustomEvent` method is deprecated and the
`CustomEvent` constructor should be used instead.
Extends d9bf571f5c49e1cac9054cf6b7acfc0b5b719876.
In PR #16327 the `eslint-plugin-mozilla` package was updated so we no
longer have to force-install packages, and the force-install flags for
`npm install` were removed. However, the CI job was missing from this
commit, which we fix here. In general force-installing packages
shouldn't be necessary unless there are problems with dependencies,
which we would like to know about, so especially in the CI job it seems
like a good idea to not force-install packages to catch upcoming defects
early on.
Extends 19526d2322fabd4425688bb7c5504fa9ea015c5c.
This method was added in PR 4938, almost nine years ago, however it doesn't appear to ever have been used.
Given the similarities between the `PDF17` and `PDF20` classes, and how they're used, if the `PDF20.hash` method was actually necessary you'd also expect a similiar method in the `PDF17` class.
The "binary" CMap-format is specific to the PDF.js library, and is used to reduce the size of the built-in CMap data-files.
By moving this code to its own file we can remove the nowadays unnecessary closures, which helps to slightly reduce the size of this code.
The latest version of `eslint-plugin-mozilla` removed the Prettier dependency, see https://bugzilla.mozilla.org/show_bug.cgi?id=1677562, which means that we no longer need to use `npm install --force` in the PDF.js library.
When permissions are enabled and the PDF document doesn't have the COPY-flag set, it shouldn't be possible for the user to trigger the "copy all text" feature.
While this slightly reduces duplication in the CSS rules, some of the auto-formatting done by Prettier is perhaps not great. (Given the overall advantage of using Prettier, we'll probably have to simply accept this.)
Hopefully these changes make sense (since this functionality is new to me), however the existing `xfa`-tests should help avoid any outright regressions.
Some arabic chars like \ufe94 could be searched in a pdf, hence it must be normalized
when creating the search query. So to avoid to duplicate the normalization code,
everything is moved in the find controller.
The previous code to normalize text was using NFKC but with a hardcoded map, hence it
has been replaced by the use of normalize("NFKC") (it helps to reduce the bundle size
by 30kb).
In playing with this \ufe94 char, I noticed that the bidi algorithm wasn't taking into
account some RTL unicode ranges, the generated font wasn't embedding the mapping this
char and the unicode ranges in the OS/2 table weren't up-to-date.
When normalized some chars can be replaced by several ones and it induced to have
some extra chars in the text layer. To avoid any regression, when copying some text
from the text layer, a copied string is normalized (NFKC) before being put in the
clipboard (it works like this in either Acrobat or Chrome).
After the previous patch we now have only *a single* `PRODUCTION` occurrence in the entire code-base, more specifically in the `web/viewer.html` file.
This special build-target can be replaced with any condition that always evaluate to `false`, such as e.g. a comment.
*Please note:* This patch might be considered too hacky, hence I completely understand if it's rejected.
This *special* build-target is very old, and was introduced with the first pre-processor that only uses comments to enable/disable code.
When the new pre-processor was added `PRODUCTION` effectively became redundant, at least in JavaScript code, since `typeof PDFJSDev === "undefined"` checks now do the same thing.
This patch proposes that we remove `PRODUCTION` from the JavaScript code, since that simplifies the conditions and thus improves readability in many cases.
*Please note:* There's not, nor has there ever been, any gulp-task that set `PRODUCTION = false` during building.
To make this functionality work out-of-the-box in custom implementations, see e.g. the "viewer components" examples, it'd be slightly easier if we dynamically create/insert the "hiddenCopyElement" in the `PDFViewer` constructor.
Given that the "copy all text" feature still appears to work just as before with this patch, hopefully I'm not overlooking any reason why doing this would be a bad idea.
I was playing with the new "copy all text" feature, and stumbled upon one document where the copied text was truncated; see http://mirrors.ctan.org/info/lshort/english/lshort.pdf
The problem turns out to be that on [page 83](https://ftp.acc.umu.se/mirror/CTAN/info/lshort/english/lshort.pdf#page=83) the textLayer contains `\u0000` and apparently copying just stops when a null char is encountered.
To fix this we can simply use an existing helper function, and with this patch we're able to successfully copy all the text in that document.
*Please note:* This patch only extends the `PDFFindController` implementation itself to support this functionality, however it's *purposely* not exposed in the default viewer.
This replaces the previous `phraseSearch`-parameter, and a `query`-string will now always be interpreted as a phrase-search.
To enable searching for individual words, the `query`-parameter must instead consist of an Array of strings. This way it's now also possible to combine phrase/word searches, with a `query`-parameter looking something like `["Lorem ipsum", "foo", "bar"]` which will search for the phrase "Lorem ipsum" *and* the words "foo" respectively "bar".
This method was originally added in PR 1320, eleven years ago, however it doesn't appear to ever have been used (not even from the start).
Furthermore, this method also tries to access a property that doesn't exist (`this.out`) and then call a method that also doesn't exist (`writeByteArray`).
In looking at https://bugs.ghostscript.com/show_bug.cgi?id=706451 I noticed that bug2.pdf was pretty
slow to load for such a basic file.
In profiling I noticed that a lot of time is spent in Array.concat, hence this patch use Array.push when
it's possible (it's now ~3 times faster).
The changes in PR 16238 were intended specifically for Node.js environments, however they accidentally applied to older browsers as well.
*Please note:* In up-to-date browsers `Path2D` is available in Workers, which should be connected to the introduction of `OffscreenCanvas`.
By getting the width/height of the first page initially, we can slightly reduce the amount of code needed both in the `hasEqualPageSizes`-check and when building the print-styles.
For the moment there is no real consensus on how we should download a pdf on Android.
Hence we keep this solution for the moment but behind a pref (which will be true on
nightly only).
Currently we repeat the same code in lots of places, to update the "toggled" class and "aria-checked" attribute, when various toolbar buttons are clicked.
For the MOZCENTRAL build-target this patch reduces the size of the *built* `web/viewer.js` file by just over `1.2` kilo-bytes.
Apparently the `structuredClone` polyfill doesn't handle transfers correctly, and `DOMException`s may thus be thrown. This is particularly problematical in Node.js environments, where that exception (obviously) isn't available.
To work-around these issues we'll simply ignore any transfers in `legacy`-builds, since those *may* use the `structuredClone` polyfill. This will obviously lead to slightly higher memory usage in those builds, however this really only affects Node.js environments. (Browsers are only affected if workers are disabled, however that's never been an officially recommended/supported configuration.)
Currently `float: inline-start/inline-end` is only supported in Firefox, see https://developer.mozilla.org/en-US/docs/Web/CSS/float#browser_compatibility, and in order to support other browsers we're thus forced to jump through some hoops.
This leads to slightly less nice code in the *built-in* Firefox PDF Viewer, and this patch attempts to improve the current situation:
- Use Stylelint to forbid direct use of `float: inline-start/inline-end` in the CSS files, to prevent future bugs in the general PDF.js viewer.
- Do a build-time replacement, only in MOZCENTRAL builds, to replace the CSS-variables with raw `float: inline-start/inline-end` instances.
Currently we have two separate image-caches on the worker-thread:
- A local one, which is unique to each `PartialEvaluator.getOperatorList` invocation. This one caches both names *and* references, since image-resources may be accessed in either way.
- A global one, which applies to the entire PDF documents and all its pages. This one only caches references, since nothing else would work.
This patch introduces a third image-cache, which essentially sits "between" the two existing ones. The new `RegionalImageCache`[1] will be usable throughout a `PartialEvaluator` instance, and consequently it *only* caches references, which thus allows us to keep track of repeated image-resources found in e.g. different /Form and /SMask objects.
---
[1] For lack of a better word, since naming things is hard...
This effectively implements some of the changes from https://phabricator.services.mozilla.com/D170496, but using our existing "direction aware" CSS-variable to limit the amount of code changes needed.
The patch changes the minimum supported version of Google Chrome as follows:
- Chrome 88, which was released on 2021-01-19; see https://en.wikipedia.org/wiki/Google_Chrome_version_history
This is done to allow use of modern CSS features, such as e.g. `:is()` and `:where()` in the code-base.
When installing the PDF.js project itself it's currently necessary to use `--force` in order for all packages to install correctly, see issue 15429, hence the same is also necessary when using the `gulp dist-install` command for local development/testing.
*Please note:* This parameter has never been used within the PDF.js library/viewer itself, and it was only ever added for backwards compatibility reasons.
This parameter was added in PR 7475, over six years ago, to try and optionally maintain the previous *default* text-extraction behaviour.
However as part of the general text-extraction improvements in PR 13257, almost two years ago, the `disableCombineTextItems` functionality was accidentally "broken" in various ways. Note how the only (very basic) unit-test was updated in a way that doesn't really make sense, since generally speaking you'd expect that using the option should result in *more* (or at least the same number of) text-items. Furthermore there's also the recent issue 16209, where the option causes almost all textContent to be concatenated together.
Hence this patch proposes that we simply remove the `disableCombineTextItems` option since it's essentially unused/untested functionality, as evident from the fact that it took almost two years for someone to notice that it's broken.
With the changes in PR 16153 we're no longer setting a `<base href>` in the Firefox PDF Viewer, hence it shouldn't be necessary to keep setting a `baseUrl` in the `PDFLinkService`-class.
Given that the original document URL is now kept, the browser itself will handle relative URLs and we can thus slightly reduce the amount of string parsing required when handling various links in the viewer.
Currently if you e.g. enable the `useOnlyCssZoom` option rendering may no longer finish as intended. To reproduce:
- Enable the `useOnlyCssZoom` option.
- Load https://github.com/mozilla/pdf.js/files/1522715/wuppertal_2012.pdf (in the development viewer).
- When rendering starts, *immediately* change the zoom-level.
In this case the document will never finish rendering, since the `postponeDrawing`-functionality will (here incorrectly) abort rendering and with CSS-only zooming rendering is only expected to happen once per page.
To fix this we'll simply ignore any `drawingDelay` when CSS-only zooming is used (regardless if it's triggered via the option or the zoom-level being very large).
Currently the `zoomLayer` isn't rotated correctly in all cases. To reproduce:
- Load https://github.com/mozilla/pdf.js/files/1522715/wuppertal_2012.pdf
- Let the document render.
- Rotate the document *four* times, such that the original rotation is restored.
The easiest solution, as far as I can tell, is that we always set the `transform` just as we did (for years) prior to the changes in PR 15812.
Originally we used helper functions for checking if something was a Dictionary or Stream, and then having an initial `typeof` check probably made sense.
However, given that we're using `instanceof` nowadays the additional check longer seems necessary.
Currently we're *virtually* duplicating the same code, for validating quotation marks, twice in this helper function.
The size decrease is quite small (107 bytes) and this makes the code slightly harder to reader, hence I completely understand if this patch is rejected.
Given that this functionality only applies in the viewer, when `PDFBug` is being enabled and used, it can't hurt to slightly reduce the size of this code.
- Reduce a little bit of duplication by enforcing the max/min scale-values once, at the end, in the `increaseScale`/`decreaseScale` methods.
- Convert the "private" `PDFViewer` scale-related methods into actually private ones, now that JavaScript supports that.
Having just reviewed a patch touching this code, I couldn't help noticing that an `Object` isn't really the optimal data-structure for this and nowadays we can do better by using a `Set` instead.
Given that the viewer always set the `dir`-attribute, to either LTR or RTL, we should be able to use this logical CSS property to (very slightly) reduce the size of the CSS; please see https://developer.mozilla.org/en-US/docs/Web/CSS/inset-block
This is something that I completely overlooked in PR 16162, which in some cases cause the default viewer to incorrectly print warnings.
This can be reproduced with the PAGE scrolling-mode, and/or the PresentationMode, and this patch simply work-around it by checking the visibility as well (since the warning is a best-effort solution anyway).
The `pageColors`-option was removed from the `CanvasGraphics`-constructor in PR 16075, hence the code in the API no longer needs to pass in that option; this is something that I missed during review.
The signatures of these methods were changed in PR 15886, which has now been included in a couple of releases, hence it should hopefully be OK to remove the fallback code-paths now.
Also, the methods are updated slightly to be explicit about what options are supported and we'll no longer pass along any arbitrary options to the "private" methods.
Some of these pre-processor statements are *many* years old, and could thus do with some clean-up. Note that the pre-processor originally didn't support else-if statements, and by using those the code becomes a bit less verbose.
The idea is to apply an overall filter on each page: the main advantage
is to have some filtered images which could help to make them visible for
some users.
During review of PR 16151 this method was simplified, however I overlooked the fact that we now can (and really should) improve this by removing duplication.
Unfortunately I don't believe that we can simply add a default `--scale-factor` CSS-variable to the `container`-element, since that might not be entirely appropriate/correct in all cases.[1]
However, we can at least print a console-error to hopefully make this situation more apparent to users. (This is purposely not using the `warn` helper-function, since those messages can be disabled.)
---
[1] One example is in our reference-tests, where we don't need to add it to the `container`-element itself.
With the previous commit this is now completely unused in API, hence it can be removed. This is done in a separate commit to make it easier to re-instate it, would the need ever arise.
This patch extends PR 16115 to work in all browsers, regardless of their `OffscreenCanvas` support, such that transfer functions will be applied to general rendering (and not just image data).
In order to do this we introduce the `BaseFilterFactory` that is then extended in browsers/Node.js environments, similar to all the other factories used in the API, such that we always have the necessary factory available in `src/display/canvas.js`.
These changes help simplify the existing `putBinaryImageData` function, and the new method can easily be stubbed-out in the Firefox PDF Viewer.
*Please note:* This patch removes the old *partial* transfer function support, which only applied to image data, from Node.js environments since the `node-canvas` package currently doesn't support filters. However, this should hopefully be fine given that:
- Transfer functions are not very commonly used in PDF documents.
- Browsers in general, and Firefox in particular, are the *primary* development target for the PDF.js library.
- The FAQ only lists Node.js as *mostly* supported, see https://github.com/mozilla/pdf.js/wiki/Frequently-Asked-Questions#faq-support
The tag <base> is used to resolve relative URIs within the document.
Newly added SVG filters use a relative URI which then use the URI in base
but this one mismatches with the document URI and consequently filters are
not found in the Firefox viewer.
So this patch just removes <base> and replace few relative URLs by absolute
ones.
In the general PDF.js library multiple PDF documents may be opened on the same web-page, which is why we many years ago started using document-specific identifiers to prevent issues with global data such e.g. with fonts.
Hence we need to treat the identifiers generated by the `FilterFactory` in the same way, since the SVG-filters for two separate PDF documents may otherwise get identical ids.
PDF gradients do not have color stops but an arbitrary PDF function of
the type f(t) -> color. CSS gradients are only based on color stops.
Most PDF gradient functions are produced from color stop oriented
gradients.
Take advantage of this by sampling the PDF function at a higher
frequency but not converting any samples which could be interpolated to
color stops. The sampling frequency is chosen to be the least common
multiple of as many values as practical to exactly re-create the common
case of the PDF function implementing equally spaced linearly
interpolated stops in RGB color space. This also allows for better
approximation of other smooth PDF functions (non-linear, or non-equally
spaced, or in different color space).
Fixes: #10572, #14165
The current value originated in PR 2317, and in the decade that have passed the amount of RAM available in (most) devices should have increased a fair bit.
Nowadays we also do a much better job of detecting repeated images at both the page- and document-level, which helps reduce overall memory-usage in many documents.
Finally the constant is also moved into the `src/shared/util.js` file, since it was implicitly used on both the main- and worker-thread previously.
Currently in PDF documents with large images we immediately cleanup once rendering has finished, in order to reduce memory-usage.
Normally that shouldn't be a big problem, however when e.g. repeated zooming happens in the viewer that could easily lead to a lot of wasted resources (and waiting).
Hence this patch, which introduces a new `PDFPageProxy` method that will slightly delay cleanup after rendering.
The dimensions still need to be fixed (from times to times they're in px)
but it doesn't have to be postponed anymore.
To test it: draw something and when resizing look at the dimensions of the div
in devtools, the units must be %.
This simply extends the approach in PR 10727 to also cover Patterns, which shouldn't be a common occurrence in Type3 fonts (since this is the first issue we've seen).
This patch updates the minimum supported environments as follows:
- Node.js 16, which was released on 2021-04-20; see https://en.wikipedia.org/wiki/Node.js#Releases
Note also that Node.js 14 will very soon reach EOL, and thus no longer receive any security updates.
This was deprecated in PR 15758, which has now been included in three official PDF.js releases.
While PR 15880 did limit the bundle-size impact of this functionality on e.g. the Firefox PDF Viewer, it still leads to some unnecessary "bloat" that these changes remove.
Furthermore, with this being deprecated there'd also be no effort put into e.g. extending the `UNSUPPORTED_FEATURES` list when handling future error cases.
The idea is to encode large image in BMP format (which is very simple and doesn't
require to compute any checksums) and then use createImageBitmap with a BMP blob
(which doesn't suffer of the Canvas/ImageData limits).
From a performance point of view, it isn't crazy (generating a large blob + decoding
it on the main thread is really not ideal) but at least we've something to display
which is a way better than a blank page (and one can notice that most of the time is
spent in decoding the image from the pdf stream).
PDF 32000-1:2008 7.10.5.1 "Type 4 (PostScript Calculator) Functions"
defers to the PostScript Language Reference for the description of these
functions. The PostScript Language Reference, third edition chapter 8
"Operators" defines the `angle` type as a "number of degrees". Section
8.1 defines "angle `sin` real", "angle `cos` real", and "num den `atan`
angle". The documentation for `atan` further states that it will return
an angle in degrees between 0 and 360.
Handle these operators correctly in `PostScriptEvaluator.execute`.
Convert the inputs to `sin` and `cos` from degrees to radians for use
with `Math.sin` and `Math.cos`. Correctly pop two values from the stack
for `atan`, use `Math.atan2`, and convert from radians to (positive)
degrees.
This was deprecated in PR 15943, which has now been included in two official PDF.js releases.
Given that `PDFDataRangeTransport` is somewhat unlikely to be used outside of the *built-in* Firefox PDF Viewer, it doesn't seem necessary to wait longer before removing this.
Also, removes the specific error-message for GENERIC builds to not unnecessarily "advertise" using non-objects when calling the `getDocument`-function.
*Please note:* This patch is written using the GitHub UI, since I'm currently without a dev machine, so hopefully it works correctly.
We introduced the use of OffscreenCanvas in #14754 and this patch aims
to use them for all kind of images.
It'll slightly improve performances (and maybe slightly decrease memory use).
Since an image can be rendered in using some transfer maps but because of
OffscreenCanvas we don't have the underlying pixels array the transfer maps
stuff is re-implemented in using the SVG filter feComponentTransfer.
Rather than repeatedly initializing a `canvasFactory`-instance for every page, move it to the document-level instead.
*Please note:* This patch is written using the GitHub UI, since I'm currently without a dev machine, so hopefully it works correctly.
Currently we repeat the `FeatureTest.isOffscreenCanvasSupported` checks all over the worker-thread code, and with upcoming changes this will become even "worse".
Hence this patch, which changes the *worker-thread* default value for the `isOffscreenCanvasSupported`-parameter to `false` and moves the feature-testing into the `BasePdfManager`-constructor.
*Please note:* This patch is written using the GitHub UI, since I'm currently without a dev machine, so hopefully it works correctly.
Currently some `getCtx` calls will have `isOffscreenCanvasSupported === undefined` set, meaning that `OffscreenCanvas` isn't being used as intended, since no `TextLayerRenderTask._isOffscreenCanvasSupported` property exists.
*Please note:* This patch is written using the GitHub UI, since I'm currently without a dev machine, so hopefully it works correctly.
I noticed several 'Path not found' errors because of a field called #subform[2].
From the XFA specs, the hash is used for a class of elements in the template tree.
When we're looking for a node in the datasets tree, it doesn't make sense to search
for a class. Hence the path element starting with a hash are just skipped.
In order to help to identify a link, we add a border around it with the LinkText color.
And backdrop colors are inverted when the mouse pointer hovers them, this way it should
help to identify the link where the pointer is.
With upcoming changes we'll potentially start to cache `ImageBitmap` data at the document-level, in addition to just at the page-level.
Hence we need to ensure that such data is actually released on clean-up, and rather than duplicating the existing *manual* handling this code is instead moved into the `PDFObjects.clear` method. (In my opinion, this is an overall improvement even without globally cached `ImageBitmap` data.)
*Please note:* This patch is written using the GitHub UI, since I'm currently without a dev machine, so hopefully it's correct and makes sense.
The `Buffer`-object is Node.js specific functionality[1], thus (obviously) not found in browsers. Please note that the PDF.js library has never officially supported/documented that binary data can be passed as a `Buffer`, and that *internally* in the `src/core`-code we only work with standard `Uint8Array`s.
This means that if, in Node.js environments, a `Buffer` is passed to the API we need to wrap it into a `Uint8Array`, which essentially means creating a copy of the data and thus increasing memory usage.
---
[1] Refer to https://nodejs.org/api/buffer.html#buffer
- Pass the `URL`-object directly to `getDocument`, since that's been supported since PR 13166.
- Remove support for the `disableRange`-option in the test-manifest, since it's completely unused. Please note that it's originally added in PR 2719, however there's never actually been any reference tests using it (not even from the start).
Given that the option is `false` by default everywhere (e.g. in the Firefox PDF Viewer) and that we have unit-tests for `disableRange = true`, it doesn't seem necessary to add new reference tests for it now.
Currently we duplicate the same code more than once in the `test/driver.js` file, which we can avoid by adding a new `AnnotationStorage` helper method instead.
Given that this helper function is only used on the worker-thread, there's no reason to duplicate it in both of the *built* `pdf.js` and `pdf.worker.js` files.
Currently these classes take a bunch of parameters (somewhat randomly ordered), probably because this is very old code that's been extended over the years.
Hence this patch changes the constructors to use parameter-objects instead, which improves consistency and (slightly) reduces the amount of code as well.
*Please note:* Also removes the `msgHandler`-property on these classes, since I cannot find a single call-site that accesses it.
Given that the debugging hash-parameters will only be used when the `pdfBugEnabled` option is manually set[1], we can skip a *tiny* bit of asynchronicity for "regular" users.
---
[1] Note that it's enabled by default in the development viewer, i.e. in `gulp server` mode.
Currently this helper function only has two call-sites, and both of them only pass in `ArrayBuffer` data. Given how it's implemented there's a couple of code-paths that are completely unused (e.g. the "string" one), and in particular the intended fast-paths don't actually work.
This patch re-factors and simplifies the helper function, and it'll no longer accept anything except `ArrayBuffer` data (hence why it's also re-named).
Note that at the time when `arraysToBytes` was added we still supported browsers without TypedArray functionality, and we'd then simulate them using regular Arrays.
When printing the pdf in #12233 in Acrobat, we can see that the combo for country
is empty: it's because the V entry doesn't have to be one of the options.
We're using this helper function when reading data from the [`PDFWorkerStreamReader.read`](a49d1d1615/src/core/worker_stream.js (L90-L98)) and [`PDFWorkerStreamRangeReader.read`](a49d1d1615/src/core/worker_stream.js (L122-L128)) methods, and as can be seen they always return `ArrayBuffer` data. Hence we can simply get the `byteLength` directly, and don't need to use the helper function.
Note that at the time when `arrayByteLength` was added we still supported browsers without TypedArray functionality, and we'd then simulate them using regular Arrays.
In the mac case we don't want to care about the scaleFactor threshold
because else if too big another move could start and then subsequent
events aren't considered as wheel events.
It isn't really ideal and at some point we'll need to find a way at
least for the Firefox case to get the real events instead of the fake
wheel ones.
In looking at a profile, I noticed in Marker chart that there's an animation
for loading-icon.gif even if this icon isn't visible.
This patch doesn't completely remove it but just slightly postpones it.
This further extends the web-specific import maps introduced in PR 16009, to allow removing *most* of the build-time `require` statements from the viewer. The few remaining ones are fallbacks used for the COMPONENTS respectively the `legacy` GENERIC builds.
After the compatibility updates in PR 15968 it's no longer strictly necessary to build the `viewer.css` file in order for the *development viewer* to work in Chromium-based browsers.
*Please note:* Given that Chromium-based browsers still don't support the *unprefixed* `mask-image` property the icons won't look right, however the development viewer itself works.
Given that Firefox is the *primary* development target, and that running `gulp generic` locally will generate polyfilled CSS, it seems reasonable to make this simplification here.
Currently there's no toolbar in the GV-viewer, hence invoking the pageLabels functionality isn't meaningful and just leads to unnecessary parsing on both the main- and worker-threads. (And if a toolbar is added at some point, it's not clear to me if we'd want to support pageLabels in the GV-viewer anyway.)
Currently we have a couple of pre-processor checks, specifically for the GV-viewer, spread throughout the code. This works fine when *building* the viewer, however they're obviously ignored in development mode (i.e. `gulp server`).
This leads to a situation where the GV development viewer, i.e. http://localhost:8888/web/viewer-geckoview.html, behaves subtly different from its built version. This could easily lead to bugs, hence this patch introduces a development mode constant to hopefully improve things here.
Finally, in a follow-up to PR 15842, also ignores the `pageMode`-state since there's no sidebar available.
Currently there's no UI for this functionality in the GV-viewer, however we still call the API methods. This potentially leads to a bunch of worker-thread parsing, for PDF documents with these features, despite the result being completely unused.
Given that mobile devices are usually more resource constrained than desktop/laptop computers, not to mentioned battery life, we can avoid doing work that'll just be ignored anyway.
The `DownloadManager.openOrDownloadData` method is written for the default-viewer specifically, assuming a viewer able to handle e.g. URL search/hash parameters. In the viewer components there's obviously no such functionality, and we should thus trigger downloading of PDF attachments directly instead.
Given that the GV-viewer isn't using most of the UI-related components of the default-viewer, we can avoid including them in the *built* viewer to save space.[1]
The least "invasive" way of implementing this, at least that I could come up with, is to leverage import maps with suitable stubs for the GV-viewer.
The one slightly annoying thing is that we now have larger import maps across multiple html-files, and you'll need to remember to update all of them when making future changes.
---
[1] With this patch, the built `viewer.js` size is 391 kB and `viewer-geckoview.js` is 285 kB.
By default we're using worker-thread fetching (in browsers) of this data nowadays, however in Node.js environments or if the user provides custom factories we still fallback to main-thread fetching.
Hence it makes sense, as far as I'm concerned, to move this initialization into the `getDocument` function to ensure that the factories can actually be initialized *before* attempting to load the document.
Also, this further reduces the amount of `getDocument` parameters that we need to pass into into the `WorkerTransport` class.
Currently we're passing all available parameters to this function respectively class, despite that not actually being necessary.
By splitting the parameters we not only improve the structure, and basically "document" the code a little bit, but we can also simplify the `_fetchDocument` function considerably.
This is very old code, where we loop through the user-provided options and build an internal parameter object. To prevent errors we also need to ensure that the parameters are correct/valid, which is especially important for the ones that are sent to the worker-thread such that structured cloning won't fail.[1]
Over the years this has led to more and more code being added in `getDocument` to validate the user-provided options, and at this point *most* of them have at least basic validation. However the way that this is implemented feels slightly backwards, since we first build the internal parameter object and only *afterwards* validate those parameters.[2]
Hence this patch changes the `getDocument` function to instead check/validate the supported options upfront, and then *explicitly* build the internal parameter object with only the needed properties.
---
[1] Note the supported types at https://developer.mozilla.org/en-US/docs/Web/API/Web_Workers_API/Structured_clone_algorithm#supported_types
[2] The internal parameter object may also, because of the loop, end up with lots of unnecessary properties since anything that the user provides is being copied.
These functions invoke the `PDFViewer.currentPageNumber` setter, which already checks that a `pdfDocument` is currently active. Also, given that they're event handlers for the First/Last-page buttons (in the SecondaryToolbar) they can't be invoked before the viewer has been fully initalized.
The default value of the `--scale-select-width` CSS variable has been choosen such that it should be large enough for most locales. This means that in many locales we don't even update the CSS variable at all, and for those locales where we do the update happens *one time* early during the viewer initialization (i.e. before the PDF document has loaded).
*Please note:* Compared to other recent PRs, the effect of these changes ought to be really tiny and are mostly done to promote better coding patterns.
We should be able to let Jasmine simply compare directly against an actually empty Object, rather than using a manually implemented helper function for that.
A number of methods have their Promises cached, to avoid repeated worker round-trips, since they're expected to be called more than once from the default viewer. The way that the caching is currently implemented means that we need to remember to manually clear these Promises on document cleanup/destruction, and it'd be nice to avoid that.
With this patch the relevant Promises are now instead placed in just one `Map`, which is easy to clear, and a new helper method is also introduced to reduce duplication for *simple* `WorkerTransport` methods.
The reasons for making this change are:
- There's no UI available to toggle the cursor-tools in the GeckoView-specific viewer.
- The `HandTool`-implementation basically *simulates* touch scrolling, and is thus unlikely to be helpful/useful anyway.
- PR 15831 already changed the relevant call-sites to handle `PDFViewerApplication.pdfCursorTools` being undefined.
Some of the code in this method is *very* old, and we could thus modernize it a little bit by removing a couple of the loops used to build the `getDocument` argument.
These `@media` rules were most likely just copy-pasted from the regular viewer, however none of them are currently necessary since the GeckoView-specific viewer doesn't have any toolbars.
Note that the whole purpose of these CSS rules are to make the toolbar, of the regular viewer, responsive. If we in the future add toolbars for the GeckoView-specific viewer, these rules most likely wouldn't be usable as-is anyway.
This option was added specifically for third-party users, but has never been used in the PDF.js project itself. Furthermore there's no preference that can be used to enable it, and you need to provide the `removePageBorders` option when initializing a `PDFViewer`-instance.
This patch thus get rid of a little bit more unused code in the Firefox PDF Viewer.
*Unfortunately I missed this during testing/reviewing of PR 15992.*
With the changes in PR 15992 we're now only adding the `loadingIcon`-class when rendering is actually `RUNNING`, in order to improve overall performance.
However when resetting the page, i.e. the `INITIAL` state, we also need to remove the `loadingIcon` completely. Without this patch if you scroll through a document where the pages don't load instantaneously, see e.g. issue 2504, we'll leave the `loadingIcon`-class attached to pages that have had their rendering cancelled *and* also been evicted from the `PDFPageViewBuffer`-instance.
This way we don't have a lot of useless divs and we let the css engine handle the
creation/destruction of the :after pseudo-element.
It'll help to slightly improve performance when zooming.
The only parameter that we actually need here is the `PDFDataRangeTransport`-instance, since the others are not necessary.
- The `url` parameter, as passed to the `getDocument` function in the API, is simply being ignored; see 2d87a2eb1c/src/display/api.js (L447-L458)
- The `length` parameter, as passed to the `getDocument` function in the API, is always being overwritten; see 2d87a2eb1c/src/display/api.js (L519-L525)
Until PR 12563 is deemed safe to land, I'd still like to be able to use worker-modules in the viewer during local development.
Hence this patch which *temporarily* adds a new `workerModules` hash-parameter, only available in non-PRODUCTION mode, that allows using worker-modules in the development viewer.
To enable this functionality, simply use http://localhost:8888/web/viewer.html#workerModules=true
The initial CMap support was added in PR 4259 using the "raw" Adobe files, however they were quickly deemed to be unnecessarily large. As a result PR 4470 introduced the more compact "binary" CMap format, with both of those PRs being included in the very same release (version `0.8.1334`) .
Please note that we've thus never shipped anything *except* the "binary" CMap files with the PDF library, and furthermore note that we've not even once updated the CMap files since they were originally added almost nine years ago.
Requiring users to remember that `cMapPacked = true` is necessary, in addition to setting the `cMapUrl` parameter, in order for CMap loading to work feels like a less than ideal API.
Hence this patch, which suggests that we simply let `cMapPacked` default to `true` now.
If this method was added today, I really can't imagine that we'd support anything *except* objects. Unfortunately we cannot just remove this now, since the code has existed since "forever", however we can deprecate this and limit it to only the GENERIC build.
Furthermore, we can avoid a redundant `PDFViewerApplication.setTitleUsingUrl` call in the Firefox PDF Viewer since the title has already been set previously in that case.
Until just recently the only existing `Path2D` polyfill didn't have support for Node.js and/or the `node-canvas` package. Given that this was just fixed, in the latest version, we can now finally remove our inline-checks at the relevant call-sites; please also see https://github.com/nilzona/path2d-polyfill#usage-with-node-canvas
In PR #15757, a value is automatically converted into a number when it's possible
but the case of numbers like "000123" has been overlooked and their format must
be preserved.
When a script is doing something like "foo.value + bar.value" and the values are
numbers then "foo.value" must return a number but the displayed value must be what
the user entered or what a script set, so this patch is just adding a a field
_orginalValue in order to track the value has it has defined.
Some people are used to use a comma as decimal separator, hence it must be considered
when a value is parsed into a number.
This patch is fixing a regression introduced by #15757.
There's really no need for these "complicated" default value assignments, since `GlobalWorkerOptions` is a local variable at this point, and this is rather a case of too much copy-and-paste.
Note that years ago, when all options were set using a global `PDFJS` object, it's possible that options had been set (from the outside) *before* the object had been properly initialized; see e.g. a89071bdef/src/display/global.js
In general it's always recommended to pass a *parameter object* when calling the `getDocument`-function in the API, since that's the only way to provide additional options, and the fact that it also accepts a URL or TypedArray directly is now mostly for backwards compatibility reasons.
Unfortunately we cannot really remove this, since that code has existed since "forever", however we can limit it to only the GENERIC build to avoid completely unnecessary checks in e.g. the Firefox PDF Viewer.
Finally, note that the default-viewer always provides a *parameter object* when calling the `getDocument`-function and it's thus completely unaffected by these changes.
By being less specific about which *exact* JavaScript features are required for the default vs `legacy` build, we don't need to worry about keeping multiple README files up-to-date.
These README files will now refer back to the FAQ for current browser/environment support information.
This function is only used in PresentationMode these days, but we can still improve it a little bit:
- Use the existing web-platform `deltaMode` constants, rather than defining our own constants for those values.
- Access the `deltaMode` first, before the `delta{X, Y}` properties, to avoid being affected by bug 1392460 (similar to the default viewer).
- Use a `URL`-instance directly, since it's by definition an absolute URL.
- Actually limit the "raw" url-string handling to Node.js environments, as intended.
- Skip the warning, since we're already throwing an Error if the `url`-parameter is invalid.
It seems nicer overall, since we're exporting the `ProgressBar` in the viewer-components, to move this functionality into the `ProgressBar`-class itself rather than handling it "manually" in the default-viewer.
*Please note:* I cannot reproduce the problem reported in bug 1811668, regarding the context menu, and in any case it's not clear that that part is even a PDF Viewer bug.
Looking at bug 1811668 I couldn't help but noticing that the textLayer isn't correct, and it's unfortunately once again a problem with the `adjustType1ToUnicode` function. That's intended to help improve text-selection for fonts without a /ToUnicode-entry, and in many cases it does help (the original PR fixed lots of issues) however it's also caused some problems.
In order to improve text-selection in bug 1811668, we'll now properly ignore fonts that have a predefined *named* encoding specified since that's really the intention with PR 14050.
At the beginning of a search we can an update can be triggered with 0 over 0
found matches.
In the GeckoView context, we can't update the finder whenever we want but only
when it has been required.
The JBIG2 images in this PDF document are corrupt enough that even Adobe Reader warns about it when opening the file.
*Please note:* I don't really know the JBIG2 image format at all, however from a very brief look at the specification it seems that integers should be 32-bit.
In general it's recommended to pass a *parameter object* when calling the `getDocument`-function in the API, since that's the only way to provide additional options, and the fact that it also accepts a URL or TypedArray directly is now mostly for backwards compatibility reasons.
However, the `getDocument`-function also accepts a direct `PDFDataRangeTransport`-instance which just seems unnecessary.
*Please note:* The `PDFDataRangeTransport`-implementation was added specifically for the *built-in* Firefox PDF Viewer, however it's most likely not commonly used by any third-party (given that it requires manual PDF-data loading).
Furthermore, the default-viewer always provides a *parameter object* when calling the `getDocument`-function and it's thus completely unaffected by these changes.
Rather than adding `@media (forced-colors: active) { ... }`-blocks throughout the CSS code, we should utilize CSS variables instead as in our other CSS files.
The relevant TrueType font is missing both /ToUnicode *and* /Encoding entires, either of which would have prevented the (current) broken textLayer rendering.
My first idea was that we could use the `post` table in the TrueType font, see https://developer.apple.com/fonts/TrueType-Reference-Manual/RM06/Chap6post.html, to get the actual glyphNames and amend the fallback ToUnicode-map that way. Unfortunately that didn't work, since the `post` table only contained ".notdef" and "" (i.e. empty string) entries.
Instead we try to use the `name` table in the TrueType font, see https://developer.apple.com/fonts/TrueType-Reference-Manual/RM06/Chap6name.html, to determine if the platform is Windows and thus fallback to generate a ToUnicode-map from the `WinAnsiEncoding`.
When a css variable is update in a node then all the children under this
node are updated.
In order to avoid to update all the UI when a page is rescaling, this
patch moves the --scale-factor from the :root to the viewer container.
Note how all over the `src/core/annotation.js`-code we're assuming that if an `appearance`-entry exists it's also a Stream. However, we're not actually checking that thoroughly enough which causes issues in some badly generated PDF documents.
After the changes in PR 15812 we'll now *intermittently* display completely black canvases during zooming. To reproduce this, try switching to wrapped-scrolling and zoom in/out very quickly using either the mouse-wheel or pinching.
This patch removes the recently introduced `transferPdfData` API-option, and simply enables transferring of TypedArray data *by default* instead of copying it. This will help reduce main-thread memory usage, however it will take ownership of the TypedArrays. Currently this only applies to the following cases:
- TypedArrays passed to the `getDocument`-function in the API, in order to open PDF documents from binary data.
- TypedArrays passed to a `PDFDataRangeTransport`-instance, used to support custom PDF document fetching/loading (see e.g. the Firefox PDF Viewer).
*PLEASE NOTE:* To avoid being affected by this, please simply *copy* any TypedArray data before passing it to either of the functions/methods mentioned above.
Now that we transfer TypedArray data that we previously only copied, we need to be more careful with input validation. Given how the `{IPDFStreamReader, IPDFStreamRangeReader}.read` methods will always return ArrayBuffer data, which is then transferred to the worker-thread[1], the actual TypedArray data passed to the API thus need to have the same exact size as its underlying ArrayBuffer to prevent issues.
Hence we'll check for this and only allow transferring of *safe* TypedArray data, and fallback to simply copying the data just as before. This obviously shouldn't be an issue in the Firefox PDF Viewer, but for the general PDF.js library we need to be more careful here.
---
[1] See e09ad99973/src/display/api.js (L2492-L2506) respectively e09ad99973/src/display/api.js (L2578-L2590)
Note how in the API we're transferring the PDF data that's fetched over the network[1]:
- f28bf23a31/src/display/api.js (L2467-L2480)
- f28bf23a31/src/display/api.js (L2553-L2564)
To support that functionality we have the `PDFDataTransportStream`, `PDFFetchStream`, `PDFNetworkStream`, and `PDFNodeStream` implementations. Here these stream-implementations vary slightly in how they handle `ArrayBuffer`s internally, w.r.t. transferring or copying the data:
- In `PDFDataTransportStream` we optionally, after PR 15908, allow transferring of the PDF data as provided externally (used e.g. in the Firefox PDF Viewer).
- In `PDFFetchStream` we're currenly always copying the PDF data returned by the Fetch API, which seems unnecessary. As discussed in PR 15908, it'd seem very weird if this sort of browser API didn't allow transferring of the returned data.
- In `PDFNetworkStream` we're already, since many years, transferring the PDF data returned by the `XMLHttpRequest` functionality. Note how the `getArrayBuffer` helper function simply returns an `ArrayBuffer` response as-is.
- In `PDFNodeStream` we're currently copying the PDF data, however this is unfortunately necessary since Node.js returns data as a `Buffer` object[2].
Given that the `PDFNetworkStream` has been, indirectly, supporting transferring of PDF data for years it would seem really strange if this didn't also apply to the `PDFFetchStream`-implementation.
Hence this patch simply enables transferring of PDF data, when accessed using the Fetch API, unconditionally to help reduced main-thread memory usage since the `PDFFetchStream`-implementation is used *by default* in browsers (for the GENERIC build).
---
[1] As opposed to PDF data being provided as e.g. a TypedArray when calling `getDocument` in the API.
[2] This is a "special" Node.js object, see https://nodejs.org/api/buffer.html#buffer, which doesn't exist in browsers.
- Scale factor is rounded to only scale by integer percent, hence the unused
ticks are accumulated (like we already do for zoom with the mouse wheel).
- Use the same thing for the pinch-to-zoom on a touchscreen: it led to slightly
refactor the code because it happened to ignore a not so small scale which
led to a not so smooth zooming.
Also, removes the `initialData`-parameter JSDocs for the `getDocument`-function given that this parameter has been completely unused since PR 8982 (over five years ago). Note that the `initialData`-parameter is, and always was, intended to be provided when initializing a `PDFDataRangeTransport`-instance.
Version 16 that we used before is now in maintenance mode, so we should
upgrade to the most recent LTS version. For more information on the
Node.js release schedule please refer to
https://github.com/nodejs/release#release-schedule.
After the changes in PR 15850, the `background-color` of the sidebar is now unnecessarily dark in the light-theme. Hence, we can simply remove this CSS rule to improve things overall (and these changes don't affect the dark-theme much at all).
This is even an overall consistency improvement, given the existing `--sidebar-narrow-bg-color` values.
Given that this is internal functionality, not exposed in the official API, it's not entirely clear (at least to me) why we can't just initialize this directly in `src/display/api.js` instead.
When testing both the development viewer and all the ways in which we run tests, everthing still appears to work just fine with this patch.
*Please note:* The reduced test-case is *not* a perfect reproduction of the original PDF document, since this one fails to open in e.g. Adobe Reader, but I do believe that it captures the most important points here.
For corrupt *and* encrypted PDF documents, it's possible that only some trailer dictionaries actually contain an /Encrypt-entry. Previously we'd could easily miss that, since we generally pick the first not obviously corrupt trailer dictionary, and the solution implemented here is to simply pre-parse all trailer dictionaries to see if there's any /Encrypt-entries.
In most of the cases, showing the loading icon is useless because
it's for a very short time, consequently it doesn't bring any useful
information for the user.
After a delay (400ms), the icon is shown in order to inform the user
that the viewer isn't stuck but it's doing something.
In GeckoView, on an event, a callback must be executed with the result of an action,
but the callback can be used only one time.
So for each FindInPage event, we must trigger only one matches count update.
This fixes a warning reported by CodeQL, and should also make general sense given that we parse the font-data to determine the *actual* `type`/`subtype` rather than trusting the PDF document.
This option/preference was disabled in GENERIC builds, see PR 15812, to avoid landing it *just before* a new release. Hence it should be fine to enable this now.
This was deprecated in PR 15758 but it's unfortunately quite difficult to tell if third-party users are depending on this, e.g. to implement custom error reporting, and if so to what extent.
However, thanks to the pre-processor we can limit *most* of this code to GENERIC builds which still seem like a worthwhile change.
These changes reduce the bundle size of the Firefox PDF Viewer by 3.8 kB in total.
This was deprecated in PR 15758 and given that it's quite unlikely that any third-party users are relying on this functionality, since it was only ever added to support telemetry reporting in the Firefox PDF Viewer, it should hopefully be fine to remove this fairly quickly.
These changes reduce the bundle size of the Firefox PDF Viewer by 4.5 kB in total.
Given that the Fetch API only supports the http/https protocols, worker-thread fetching of CMaps and Standard-fonts may thus fail in certain cases. To improve the default behaviour we'll now also check that the `cMapUrl` and `standardFontDataUrl` options are appropriate, except in Firefox where this should always work.
This tweaks a few name that originated in PR 15812, to improve overall consistency:
- Use the `drawingDelay` parameter-name in all methods that accept a delay.
- Use the `postponeDrawing` variable-name in all relevant methods.
With upcoming background changes elsewhere in the viewer, this should be helpful in separating the styling of the loadingBar. These changes also means that both the "regular" and the "indeterminate" loadingBar now uses the same `background-color` value.
Also, shortens the related CSS variables a little bit since that can't hurt.
*This makes the same kind of changes as in the previous patch, but for the pageNumber-loadingIcon in the main toolbar.*
To display the pageNumber-loadingIcon when rendering starts, if the page is the most visible one, we'll utilize the existing "pagerender" event.
To toggle the pageNumber-loadingIcon as the user moves through the document we'll now instead utilize the "pagechanging" event, which should actually be slightly more efficient overall[1]. Note how we'd, in the old code, only consider the most visible page anyway when toggling the pageNumber-loadingIcon.
---
[1] Even in a PDF document as relatively short/simple as `tracemonkey.pdf`, scrolling through the entire document can easily trigger the "updateviewarea" event more than a thousand times.
Given that we only render one page at a time, this will lead to only *one* page-loadingIcon being displayed at a time even if multiple pages are visible in the viewer. However, this will make it clearer which page is the currently parsing/rendering one.
To simplify toggling of the page-loadingIcon visibility, the existing `PDFPageView.renderingState` is changed into a getter/setter-pair with the latter also handling the page-loadingIcon state.
An additional benefit of these changes is that the `PDFViewer` no longer needs to handling toggling of page-loadingIcon visibility during rendering, since there can only ever be *one* page rendering.
Finally, this may also simplify future changes w.r.t. page-loadingIcon visibility toggling (using e.g. a show-timeout).
The rotation-caching added in PR 15812 completely breaks initialization of PDF documents with varying page sizes, causing all pages to wrongly get the same size; see e.g. `sizes.pdf` from the test-suite.
To fix that without having to e.g. add a new parameter, which feels error prone, this patch changes the `PDFPageView.#setDimensions` method to completely ignore the rotation-caching until the `setPdfPage`-method has been called.
Note how, in the scripting initialization in the viewer, we only ever invoke `PDFPageProxy.getJSActions` *once* per page in order to improve overall performance; see a575aa13b9/web/pdf_scripting_manager.js (L372-L375)
Hence it really shouldn't be necessary to cache its result in the API, especially when that is done *manually* rather than using something like `shadow`.
When we're destroying a `PDFPageProxy`-instance, during full document destruction, we'll force-abort any worker-thread parsing of operatorLists. Hence we should make sure that any pending cancel timeout is always aborted, since a later `PDFPageProxy._abortOperatorList` call should always "replace" a previous one.
*Please note:* Technically this was always wrong, but with the changes in PR 15825 it became *ever so slightly* easier to trigger this thanks to the potentially longer timeout.
Right now, the visible pages are redrawn for each scale change.
Consequently, zooming with mouse wheel or in pinching can be pretty janky
(even on a desktop machine but with a hdpi screen).
So the main idea in this patch is to draw the visible pages only once zooming
is finished.
After the changes in PR 15829 the `loadingIconDiv` is no longer always visible when it should be, specifically in the case where we cancel and re-render a partially parsed/rendered page.
To reproduce this, try opening https://github.com/mozilla/pdf.js/files/1522715/wuppertal_2012.pdf in the viewer and change the zoom level while rendering is ongoing. In this case the `loadingIconDiv` doesn't actually become visible, despite being present in the DOM, since it's no longer at the end of the page-div.
I don't know to what extent this renders PR 15829 "pointless", however we're not repeatedly re-creating and re-inserting the `loadingIconDiv` but rather just *move* the existing element in the DOM.
When trying to find incomplete objects, i.e. those missing the "endobj"-string at the end, there's unfortunately a number of possible operators that we need to check for. Otherwise we could miss e.g. the "trailer" at the end of a corrupt PDF document, which is why the referenced document didn't work.
Currently we do all searching on the "raw" bytes of the PDF document, for efficiency, however this doesn't really work when we need to check for *multiple* potential command-strings. To keep the complexity manageable we'll instead use regular expressions here, but we can at least avoid creating lots of substrings thanks to the `RegExp.lastIndex` property; which is well supported across browsers according to https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/RegExp/lastIndex#browser_compatibility
Note that this repeated regular expression usage could perhaps be slightly less efficient than the old code, however this method is only invoked for corrupt PDF documents.
Given that the `ProgressBar`-constructor was updated, we need to update the "mobile-viewer" example as well; this is yet another thing I missed during review.
We'll no longer import the `SimpleLinkService` dependency unconditionally in the file, since it's only used in COMPONENTS-builds.
Furthermore, for the COMPONENTS-builds, we'll create a `SimpleLinkService`-instance only for those layers that actually need it.
Please note that this functionality has never really mattered for the Firefox PDF Viewer, the GENERIC viewer, or even the "simpleviewer"/"singlepageviewer" component-examples. Hence, in practice this means that only the "pageviewer" component-example[1] have ever really utilized this.
Using factories to initialize various layers in the viewer, rather than simply invoking the relevant code directly, seems (at least to me) like a somewhat roundabout way of doing things.
Not only does this lead to more code, both to write and maintain, but since many of the layers have common parameters (e.g. an `AnnotationStorage`-instance) there's also some duplication.
Hence this patch, which removes the `xfaLayerFactory` and instead uses a lookup-function in the `PDFPageView`-class to access the external viewer-properties as necessary.
Note that this should even be an improvement for the "pageviewer" component-example, since most layers will now work by default rather than require manual configuration.
---
[1] In practice we generally suggest using the "simpleviewer", or "singlepageviewer", since it does *most* things out-of-the-box and given that a lot of functionality really require *a viewer* and not just a single page in order to work.
Please note that this functionality has never really mattered for the Firefox PDF Viewer, the GENERIC viewer, or even the "simpleviewer"/"singlepageviewer" component-examples. Hence, in practice this means that only the "pageviewer" component-example[1] have ever really utilized this.
Using factories to initialize various layers in the viewer, rather than simply invoking the relevant code directly, seems (at least to me) like a somewhat roundabout way of doing things.
Not only does this lead to more code, both to write and maintain, but since many of the layers have common parameters (e.g. an `AnnotationStorage`-instance) there's also some duplication.
Hence this patch, which removes the `textLayerFactory` and instead uses a lookup-function in the `PDFPageView`-class to access the external viewer-properties as necessary.
Note that this should even be an improvement for the "pageviewer" component-example, since most layers will now work by default rather than require manual configuration.
---
[1] In practice we generally suggest using the "simpleviewer", or "singlepageviewer", since it does *most* things out-of-the-box and given that a lot of functionality really require *a viewer* and not just a single page in order to work.
Please note that this functionality has never really mattered for the Firefox PDF Viewer, the GENERIC viewer, or even the "simpleviewer"/"singlepageviewer" component-examples. Hence, in practice this means that only the "pageviewer" component-example[1] have ever really utilized this.
Using factories to initialize various layers in the viewer, rather than simply invoking the relevant code directly, seems (at least to me) like a somewhat roundabout way of doing things.
Not only does this lead to more code, both to write and maintain, but since many of the layers have common parameters (e.g. an `AnnotationStorage`-instance) there's also some duplication.
Hence this patch, which removes the `textHighlighterFactory` and instead uses a lookup-function in the `PDFPageView`-class to access the external viewer-properties as necessary.
Note that this should even be an improvement for the "pageviewer" component-example, since most layers will now work by default rather than require manual configuration.
---
[1] In practice we generally suggest using the "simpleviewer", or "singlepageviewer", since it does *most* things out-of-the-box and given that a lot of functionality really require *a viewer* and not just a single page in order to work.
Please note that this functionality has never really mattered for the Firefox PDF Viewer, the GENERIC viewer, or even the "simpleviewer"/"singlepageviewer" component-examples. Hence, in practice this means that only the "pageviewer" component-example[1] have ever really utilized this.
Using factories to initialize various layers in the viewer, rather than simply invoking the relevant code directly, seems (at least to me) like a somewhat roundabout way of doing things.
Not only does this lead to more code, both to write and maintain, but since many of the layers have common parameters (e.g. an `AnnotationStorage`-instance) there's also some duplication.
Hence this patch, which removes the `structTreeLayerFactory` and instead uses a lookup-function in the `PDFPageView`-class to access the external viewer-properties as necessary.
Note that this should even be an improvement for the "pageviewer" component-example, since most layers will now work by default rather than require manual configuration.
---
[1] In practice we generally suggest using the "simpleviewer", or "singlepageviewer", since it does *most* things out-of-the-box and given that a lot of functionality really require *a viewer* and not just a single page in order to work.
Please note that this functionality has never really mattered for the Firefox PDF Viewer, the GENERIC viewer, or even the "simpleviewer"/"singlepageviewer" component-examples. Hence, in practice this means that only the "pageviewer" component-example[1] have ever really utilized this.
Using factories to initialize various layers in the viewer, rather than simply invoking the relevant code directly, seems (at least to me) like a somewhat roundabout way of doing things.
Not only does this lead to more code, both to write and maintain, but since many of the layers have common parameters (e.g. an `AnnotationStorage`-instance) there's also some duplication.
Hence this patch, which removes the `annotationLayerFactory` and instead uses a lookup-function in the `PDFPageView`-class to access the external viewer-properties as necessary.
Note that this should even be an improvement for the "pageviewer" component-example, since most layers will now work by default rather than require manual configuration.
---
[1] In practice we generally suggest using the "simpleviewer", or "singlepageviewer", since it does *most* things out-of-the-box and given that a lot of functionality really require *a viewer* and not just a single page in order to work.
Please note that this functionality has never really mattered for the Firefox PDF Viewer, the GENERIC viewer, or even the "simpleviewer"/"singlepageviewer" component-examples. Hence, in practice this means that only the "pageviewer" component-example[1] have ever really utilized this.
Using factories to initialize various layers in the viewer, rather than simply invoking the relevant code directly, seems (at least to me) like a somewhat roundabout way of doing things.
Not only does this lead to more code, both to write and maintain, but since many of the layers have common parameters (e.g. an `AnnotationStorage`-instance) there's also some duplication.
Hence this patch, which removes the `annotationEditorLayerFactory` and instead uses a lookup-function in the `PDFPageView`-class to access the external viewer-properties as necessary.
Note that this should even be an improvement for the "pageviewer" component-example, since most layers will now work by default rather than require manual configuration.
---
[1] In practice we generally suggest using the "simpleviewer", or "singlepageviewer", since it does *most* things out-of-the-box and given that a lot of functionality really require *a viewer* and not just a single page in order to work.
Currently we'll only initialize and render the `annotationEditorLayer` once the regular `annotationLayer` has been rendered.
While it obviously makes sense to render the `annotationEditorLayer` *last*, the way that the code is currently written means that if a third-party user disables the `annotationLayer` then the editing-functionality indirectly becomes disabled as well.
Given that this seems like a somewhat arbitrary limitation, this patch simply decouples these two layers while still keeping the rendering order consistent.
By moving this code the "pageviewer"-component example will become slightly more usable on its own, it may simplify a future addition of XFA Foreground document support, and finally also serves as preparation for the following patches.
The container position and dimensions should be almost constant, hence
it's pretty useless to query them on each rescale.
Finally it avoids to trigger some reflows.
First of all, given the screen-sizes of most mobile phones using Spread modes is unlikely to be useful.
Secondly, and more importantly, since there's (currently) no UI available for the user to override a PDF document-specified Spread mode this would result in a bad UX otherwise.
Also, removes an outdated comment from the `apiPageLayoutToViewerModes` helper function.
Previously we'd abort all parsing if an Error was encountered, despite the fact that multiple `startXRefQueue`-entries may be available and that continued parsing could thus eventually be able to find usable data.
Note that in the referenced PDF document the `startxref`-operator, at the end of the file, points to a position in the middle of an arbitrary `stream` which is why things break.
Depending on e.g. the `textLayerMode` option it might not actually be necessary to always initialize this eagerly.
*Please note:* Unfortunately we cannot `shadow` a private field, hence why this is only made semi-"private".
This is done to support upcoming viewer-changes, and in order to prevent third-party users from outright breaking things we'll simply ignore too large values.
It's a follow-up of #14950: some format actions are ran when the document is open
but we must be sure we've everything ready for that, hence we have to run some
named actions before runnig the global format.
In playing with the form, I discovered that the blur event wasn't triggered when
JS called `setFocus` (because in such a case the mouse was never down). So I removed
the mouseState thing to just use the correct commitKey when blur is triggered by a
TAB key.
In order to move the annotations in the DOM to have something which corresponds
to the visual order, we need to have their dimensions/positions which means that
the parent must have some dimensions.
While reviewing recent patches, I couldn't help but noticing that we now have a lot of call-sites that manually access the `PageViewport.viewBox`-property.
Rather than repeating that verbatim all over the code-base, this patch adds a lazily computed and cached getter for this data instead.
This was essentially done only to compensate for the viewer calling `PDFPageProxy.getAnnotations` unconditionally on every annotationLayer-rendering invocation. With the previous patch that's no longer happening, and this API-caching should thus no longer be necessary.
For pages without any annotations, applies e.g. to the `tracemonkey.pdf` document, we'll repeatedly try to re-create the `annotationLayer` on every zoom and rotation operation.
The reason that this happens is because we don't insert the `annotationLayer`-div into the DOM unless there's annotations present on the page, which thus means that we miss the existing `annotationLayer`-caching present in the `PDFPageView` implementation.
This is a very old issue, and the easiest solution is to simply always insert an *empty* (and hidden) `annotationLayer`-div such that the existing code/caching starts working for the "no annotations" case as well.
Note that this is consistent with other layers, since e.g. the `textLayer` and/or `annotationEditorLayer` may be empty. Given that only a limited, by default ten, number of pages are ever active at once the additional DOM-elements shouldn't effect things negatively here.
This is consistent with the `render` methods of the other layers, and reduces overall indentation in the method.
Furthermore, don't "swallow" errors since the `PDFPageView._renderXfaLayer` method is already able to deal with that.
It doesn't seem necessary to have a *separate* `destroy` method given that the `cancel` method always invokes it unconditionally.
In the `PDFPageView.reset` method we currently attempt to call `destroy` directly, however that'll never actually happen since either:
- We're keeping the annotationEditorLayer, in which case we're just hiding the layer and nothing more (and the relevant branch is never entered).
- We're removing the annotationEditorLayer, in which case the `PDFPageView.cancelRendering` method has already cancelled *and* nulled it (and there's thus nothing left to `destroy` here).
*Please note:* Hopefully I'm not overlooking something obvious here, since both reading through the code *and* also adding `console.log(this.annotationEditorLayer);` [before this line](9d4aadbf7a/web/pdf_page_view.js (L438)) suggests that it's indeed unnecessary.
In PR 14877 I forgot to update the horizontal padding, used when computing the scale of the pages, for the case where SpreadModes and PresentationMode are being used together.
Steps to reproduce:
1. Open the viewer with the default `tracemonkey.pdf` document.
1. Enable any SpreadMode.
2. Rotate the document *once*, either clockwise or counterclockwise.
3. Enter PresentationMode.
4. Try swithching page, e.g. by clicking on the document.
Expected result:
The visible pages change as you click.
Actual result:
The visible pages are "stuck" in the current view.
The `PDFPageProxy._pageIndex` property is a "private" one that shouldn't be accessed, since it could theoretically break tomorrow if we re-factor the relevant API code.
Also, try to clean-up and improve consistency in a couple of JSDoc comments.
This change was made in PR 5552, however I cannot tell why we needed to disable searching in PresentationMode. Furthermore, with the changes in PR 13908 which effectively moved where this code is invoked, searching has now (accidentally) been working in PresentationMode in e.g. the Firefox PDF Viewer for well over a year.
So, let's just enable searching unconditionally in PresentationMode to simplify the code.
An annotation editor layer can be destroyed when it's invisible, hence some
annotations can have a null parent but when printing/saving or when changing
font size, color, ... of all added annotations (when selected with ctrl+a) we
still need to have some parent properties especially the page dimensions, global
scale factor and global rotation angle.
This patch aims to remove all the references to the parent in the editor instances
except in some cases where an editor should obviously have one.
It fixes#15780.
The main issue is due to the fact that an editor's parent can be null when
we want to serialize it and that lead to an exception which break all the
saving/printing process.
So this incomplete patch fixes only the saving/printing issue but not the
underlying problem (i.e. having a null parent) and doesn't bring that much
complexity, so it should help to uplift it the next Firefox release.
Rather than handling these parameters separately, which is a left-over from back when streaming of textContent was originally added, we can simply pass either data directly to the `TextLayer` and let it handle things accordingly.
Also, improves a few JSDoc comments and `typedef`-imports.
Compared to the recent PR 15722 for the `textLayer` this one should be a (comparatively) much a smaller win overall, since most documents don't have any structTree-data and the required parsing should be cheaper. However, it seems to me that it cannot hurt to improve this nonetheless.
Note that by moving the `structTreeLayer` initialization we remove the need for the "textlayerrendered" event listener, which thus simplifies the code a little bit.
Also, removes the API-caching of the structTree-data since this was basically done to offset the lack of caching in the viewer.
*Please note:* I don't really expect that this is will be an observable change, since virtually all PDF documents already order e.g. /MediaBox and /CropBox entries correctly.
By normalizing boundingBoxes already on the worker-thread, we can be sure that even a corrupt document won't cause issues.
Note how we're passing the `view`-getter to the `PartialEvaluator.getTextContent` method, in order to detect textContent which is outside of the page, hence it makes sense to ensure that it's formatted as expected.
Furthermore, by normalizing this once on the worker-tread we should no longer have to worry about a possibly negative width/height in the `PageViewport` constructor.
Finally, the patch also simplifies the `view`-getter a little bit.
The idea is just to resuse what we got on the first draw.
Now, we only update the scaleX of the different spans and the other values
are dependant of --scale-factor.
Move some properties in the CSS in order to avoid any updates in JS.
The deprecation is included in the current release, i.e. version `3.1.81`, and given the edge-case nature of this option I really don't think that we need to keep it deprecated for multiple releases.
This patch has been successfully tested in a local, artifact, Firefox build.
*Please note:* The only thing that'll no longer work for PDF documents opened using "data:"-URLs is middle-clicking on internal/outline links, in order to open the destination in a new tab. This is however an extremely small loss of functionality, and as can be seen in the bug the alternative (i.e. doing nothing) is surely much worse.
Currently both of the `AnnotationElement` and `KeyboardManager` classes contain *identical* `platform` getters, which seems like unnecessary duplication.
With the pre-processor we can also limit the feature-testing to only GENERIC builds, since `navigator` should always be available in browsers.
Add a deprecation notification for PDFDocumentLoadingTask.onUnsupportedFeature and PDFDocumentProxy.stats
which are likely useless.
The unsupported feature stuff have initially been added in (#4048) in order to be able to display a
warning bar and to help to have some numbers to know how a feature was used.
Those data are no more used in Firefox.
This is very old code, which is unused (by default) in browsers nowadays since the Font Loading API will always be preferred.
For Node.js environments we use the same constant as elsewhere throughout the code-base, and we can also simplify the Firefox-specific check given that the lowest supported version is `102` (as of this writing).
Finally the old TODO is removed, since the general availability of the Font Loading API has made it redundant.
The use of `Array.prototype.reduce()` is, in my opinion, hurting overall readability since it's not particularly easy to look at the relevant code and immediately understand what's going on here. Furthermore this code leads to strictly speaking unnecessary allocations and parsing, since we could just track the min/max values directly in the relevant loop instead.
This has never really been used anywhere within the PDF.js library[1], and when streaming of textContent was introduced this parameter was effectively made redundant.
Note that when streaming of textContent is used, all text-layout has already happened by the time that this `timeout`-functionality is actually invoked (thus making it pointless).
While the `timeout`-functionality may still "work" when the textContent is provided upfront, although it's never been used/tested, streaming will generally perform better (in e.g. a viewer setting).
*Please note:* While unrelated here, also removes a now unused property that I forgot in PR 15259.
---
[1] At least not since the code was moved into its current file, which happened in PR 6619 and landed seven years ago.
This can't be a particularly common feature, since we've supported Optional Content for over two years and this is the very first TilingPattern-case we've seen.
The reason for the issue is that we use the generic `getFilenameFromUrl` helper function, which was originally intended for regular URLs.
For the filenames we're dealing with in FileAttachments, we really only want to strip the path when one exists[1].
---
[1] See [bug 1230933](https://bugzilla.mozilla.org/show_bug.cgi?id=1230933) for an example of such a case.
With the changes made in PR 14564 this *should* no longer be necessary now, however we still need to keep the `scrollMatches` parameter to handle textLayers with markedContent correctly when searching.
Currently *some* functions in this file have names while others don't, and in a few cases the names are no longer entirely accurate.
For the relevant functions there should really be no need to name them, and if memory serves this was originally done since browsers (many years ago) didn't always handle anonymous functions correctly in stack traces.
Given that this helper function is only used on the worker-thread, there's no reason to duplicate it in both of the `pdf.js` and `pdf.worker.js` files.
Given that this helper function is only used on the worker-thread, there's no reason to duplicate it in both of the `pdf.js` and `pdf.worker.js` files.
Given that these functions are virtually identical, with the latter only adding a BOM, we can combine the two. Furthermore, since both functions were only used on the worker-thread, there's no reason to duplicate this functionality in both of the `pdf.js` and `pdf.worker.js` files.
Having just played around with adding FreeText-annotations and then trying to print, there were `FreeTextAnnotation: OffscreenCanvas is not supported, annotation may not render correctly.` messages printed in the console.
The reason for this is that `FreeTextAnnotation` inherits from `MarkupAnnotation`, however only `WidgetAnnotation` actually defines the `_isOffscreenCanvasSupported` property.
Adding some logging with `console.{time, timeEnd}` around all the constant definitions at the top of the `web/pdf_find_controller.js` file, I noticed that computing `DIACRITICS_EXCEPTION_STR` took close to half the total time.
My first idea was just to try and make it slightly more efficient, by reducing the amount of iterations and intermediate allocations. However, with this constant only being used during "match diacritics" searches it thus seemed like a good candidate for lazy initialization.
*Please note:* Given that this is a micro optimization, I fully understand if the patch is rejected.
Given that this PDF document is an interesting test-case for performance reasons, w.r.t. inline image caching, it probably can't hurt to add it to the test-suite to make it more readily available.
Considering the contents of that PDF document I'm not sure if we can include it directly in the repository, hence why a *linked* test-case was choosen here.
Given that this `assert` is only intended to catch any implementation bugs in our code, and not actually to validate the PDF data directly[1], we can avoid making this function call unconditionally.
---
[1] In those cases, for example a `FormatError` should have been thrown instead.
With modern EcmaScript features, we can define these fields directly instead. Please note that for backwards compatibility purposes they are still public as before, however note that this functionality is *disabled* by default (see the `pdfBug` API option).
Also, we can (slightly) simplify the two loops used in the `toString` method.
These fields were never intended to be public, since modifying them manually would lead to inconsistent state, and with modern EcmaScript features we can now enforce this.
Also, this patch removes a couple of JSDoc comments that we generally don't use.
- For text fields
* when printing, we generate a fake font which contains some widths computed thanks to
an OffscreenCanvas and its method measureText.
In order to avoid to have to layout the glyphs ourselves, we just render all of them
in one call in the showText method in using the system sans-serif/monospace fonts.
* when saving, we continue to create the appearance streams if the fonts contain the char
but when a char is missing, we just set, in the AcroForm dict, the flag /NeedAppearances
to true and remove the appearance stream. This way, we let the different readers handle
the rendering of the strings.
- For FreeText annotations
* when printing, we use the same trick as for text fields.
* there is no need to save an appearance since Acrobat is able to infer one from the
Content entry.
These variables are left-over from the initial implementation, back when `String.prototype.padStart` didn't exist and we thus had to pad manually (using a loop).
This helps improve performance for some PDF documents with a huge number of inline images, e.g. the PDF document from issue 2618.
Given that we no longer create `Stream`-instances unconditionally, we also don't need `Dict`-instances for cached inline images (since we only access the filter).
*Please note:* This only fixes the "wrong letter" part of bug 1799927.
It appears that the simple `computeAdler32` function, used when caching inline images, generates hash collisions for some (very short) TypedArrays. In this case that leads to some of the "letters", which are actually inline images, being rendered incorrectly.
Rather than switching to another hashing algorithm, e.g. the `MurmurHash3_64` class, we simply cache using a stringified version of the inline image data as the cacheKey to prevent any future collisions. While this will (naturally) lead to slightly higher peak memory usage, it'll however be limited to the current `Parser`-instance which means that it's not persistent.
One small benefit of these changes is that we can avoid creating lots of `Stream`-instances for already cached inline images.
The purpose of this patch is twofold:
- Initialize the unicode-category data *lazily* during text-extraction, since this is completely unused during general parsing/rendering.
- Stop exposing this data in the API, since it's unused on the main-thread and it seems like it was *accidentally* included.
Obviously these changes are API-observable, but hopefully no user is depending on this. Furthermore, it's trivial for a user to re-create this unicode-category data manually with a regular expression (from the exposed `unicode` property).
Currently, during text-extraction, we're repeatedly normalizing and (when necessary) reversing the unicode-values every time. This seems a little unnecessary, since the result won't change, hence this patch moves that into the `Glyph`-instance and makes it *lazily* initialized.
Taking the `tracemonkey.pdf` document as an example: When extracting the text-content there's a total of 69236 characters but only 595 unique `Glyph`-instances, which mean a 99.1 percent cache hit-rate. Generally speaking, the longer a PDF document is the more beneficial this should be.
*Please note:* The old code is fast enough that it unfortunately seems difficult to measure a (clear) performance improvement with this patch, so I completely understand if it's deemed an unnecessary change.
In order to support opening certain corrupt PDF documents, particularly hand-edited ones, this patch adds support for letting the `Catalog.getAllPageDicts` method fallback to returning an *empty* dictionary to replace (only) the first /Page of the document.
Given that the viewer cannot initialize/load without access to the first page, this will thus allow e.g. document-level scripting to run as expected. Note that by effectively replacing a corrupt or missing first /Page in this way[1], we'll now render nothing but a *blank* page for certain cases of broken/corrupt PDF documents which may look weird.
*Please note:* This functionality is controlled via the existing `stopAtErrors` option, that can be passed to `getDocument`, since it's easy to imagine use-cases where this sort of fallback behaviour isn't desirable.
---
[1] Currently we still require that a /Pages-dictionary is found though, however it *may* be possible to relax even that assumption if that becomes absolutely necessary in future corrupt documents.
Note that the "trailer"-case is already a fallback, since normally we're able to use the "xref"-operator even in corrupt documents. However, when a "trailer"-operator is found we still expect "startxref" to exist and be usable in order to advance the stream position. When that's not the case, as happens in the referenced issue, we use a simple fallback to find the first "obj" occurrence instead.
This *partially* fixes issue 15590, since without this patch we fail to find any objects at all during `XRef.indexObjects`. However, note that the PDF document is still corrupt and won't render since there's no actual /Pages-dictionary and the /Root-entry simply points to the /OpenAction-dictionary instead.
After the clean-up in PR 15616, the `PdfManager.onLoadedStream` method now only has a single call-site.
Hence why this patch suggests that we remove this method and replace it with an *optional* parameter in `PdfManager.requestLoadedStream` instead. By making the new behaviour opt-in, we'll thus not change any existing call-site.
- Some events, which require a user interaction, will allow those functions to be called.
But after few seconds, if there are no more user interaction, it won't be possible
anymore.
The idea is to give an opportunity to the user to leave the pdf.
- Disable print function when we're printing, the same with saving and disallow to save
on open events.
When a form isn't changed, we used the appearances we had in the file, but when
/NeedAppearances is true, all the appearances have to be regenerated whatever they're.
- In #15373, we implemented copy/paste actions in using the system
clipboard.
For any reasons, on Windows, the clipboard doesn't contain the expected
data when the tests are ran in parallel, hence the tests which are
using the clipboard need to be ran sequentially.
- Make sure that we can paste after having copied.
*This is very old code, and it could thus do with some simplification.*
Note how in the `src/core/worker.js` file we're combining both the `PdfManager.requestLoadedStream` and `PdfManager.onLoadedStream` methods in order to access the stream-data. This seems unnecessary, and it's simple enough to always let the `PdfManager.requestLoadedStream` method return the stream-data as well.
PR 13725 was only intended as a temporary work-around, and it seems that we can now revert that.
- Firefox 102 is the currently maintained ESR-branch, and the PDF.js project only supports the active one.
- Node.js now works, thanks to the `node-canvas` package, and I've confirmed locally that following the STR in issue 13724 generates a correct image.
In the referenced PDF document there are "numbers" which consist only of `-.`, and while that's obviously not valid Adobe Reader seems to handle it just fine.
Letting this method ignore more invalid "numbers" was suggested during the review of PR 14543, so let's simply relax our the validation here.
It appears that PR 15593 broke `issue12402`, and we thus need to partially restore the /Count check.
I completely missed this when looking at the test-results for PR 15593, both locally and on the bots, since the `Driver._getLastPageNumber` method would "swallow" an unavailable page number.
- When we're editing some annotations, keeping the role="text-box" make them visible
as editable and VoiceOver (Mac) is able to read the contents when they're focused;
- Add an attribute "aria-activedescendant" in order to make the content discoverable
by NVDA on Windows.
After PR 14311, and follow-up patches, we no longer require that the /Count entry (in the /Pages dictionary) is either present or even valid in order to parse/render a PDF document.
Hence it seems strange to keep this requirement for *corrupt* PDF documents, when trying to find a usable `trailer` in the `XRef.indexObjects` method.
With the changes in the previous patch we can move the glyph-cache lookup to the top of the method and thus avoid a bunch of, in *almost* every case, completely unnecessary re-parsing for every `charCode`.
This method, and its class, was originally added in PR 4453 to reduce memory usage when parsing text. Then PR 13494 extended the `Glyph`-representation slightly to also include the `charCode`, which made the `matchesForCache` method *effectively* redundant since most properties on a `Glyph`-instance indirectly depends on that one. The only exception is potentially `isSpace` in multi-byte strings.
Also, something that I noticed when testing this code: The `matchesForCache` method never worked correctly for `Glyph`s containing `accent`-data, since Objects are passed by reference in JavaScript. For affected fonts, of which there's only a handful of examples in our test-suite, we'd fail to find an already existing `Glyph` because of this.
When we fail to find a usable PDF document `trailer` *and* there were errors during parsing, try and fallback to a *previous* generation as a last resort during fetching of uncompressed references.
*Please note:* This will not affect "normal" PDF documents, with valid /XRef data, and even most *corrupt* documents should be completely unaffected by these changes.
Given that the new sidebar icon is slightly shorter than the old one, it cannot hurt to ever so slightly tweak the vertical position of the notification icon.
(While the patch also changes the CSS rule used for the horizontal position, this is a no-op and was done to improve consistency between the two values.)
Part of this is very old code, and back when support for parsing the catalog-version was added things became less clear (in my opinion).
Hence this patch tries to improve things, by e.g. validating the header- and catalog-version separately.
Note how we're currently skipping all main-thread cleanup when document destruction has started, but for some reason we're still dispatching the "Cleanup" message.
This seems like a simple oversight, since destruction will already invoke the `BasePdfManager.cleanup` method (on the worker-thread) to fully clear-out all caches.
Given the sheer number of heuristics added to this method over the years, moving the *valid* unicode found case to the top should improve readability of the code.
- Fix Field::getArray in order to collect only the fields which have a value;
- Fix AFSimple_Calculate:
* allow to have a string with a list of field names as argument;
* since a field can be non-terminal, use Field::getArray to collect
the field under it and then apply the calculation on all the descendants.
This code was added all the way back in PR 6698, almost seven years ago, for backwards compatibility reasons. At this point in time, it seems that we can remove that since:
- We have more fine-grained "UnsupportedFeature" reporting elsewhere in the worker-thread code nowadays.
- The GetOperatorList-handling is now using `ReadableStream`s, which means that errors are being forwarded to the main-thread anyway.
- We're also no longer displaying a notification-bar, in the *built-in* Firefox PDF Viewer, for any of these "UnsupportedFeature" messages.
*Please note:* I don't really know what I'm doing here, however the patch appears to fix the referenced issue when comparing the rendering with Adobe Reader (with the caveat that I don't speak the language in question).
When a new PDF document is opened in the GENERIC viewer we (obviously) create a new `AnnotationEditorUIManager`-instance, since those are document-specific, and thus we need to ensure that we actually register the `editorTypes` for each one.
Note how after having found the "%PDF-" prefix we then read both the prefix and the version in the loop, only to then remove the prefix at the end.
It seems better to instead advance the stream position past the "%PDF-" prefix, and then read only the version data.
Finally the loop-condition can also be simplified slightly, to further clean-up some very old code.
*Fixes a regression from PR 15246, sorry about that!*
The return value of all `Annotation.getOperatorList` methods was changed in PR 15246, however I missed updating the error code-path in `Page.getOperatorList` which thus breaks all operatorList-parsing for pages with corrupt Annotations.
Looking at the code on the worker-thread, there doesn't appear to be any particular reason for placing *some* of the properties in a `source`-object when sending them with "GetDocRequest".
As is often the case the explanation for this structure is rather "for historical reasons", since originally we simply sent the `source`-object as-is. Doing that was obviously a bad idea, for a couple of reasons:
- It makes it less clear what is/isn't actually needed on the worker-thread.
- Sending unused properties will unnecessarily increase memory usage.
- The `source`-object may contain unclonable data, which would break the library.
Rather than sending all of these parameters individually and then grouping them together on the worker-thread, we can simply handle that in the API instead.
All of the these constants have been deprecated for a while, and with the upcoming *major* version this seems like a good time to remove them.
For the string-constants we can simply remove them, but the number-constants are left commented out since we don't want to re-number the list to prevent third-party breakage.
The way that we set the width of the `dropdownToolbarButton`-select is very old, and despite some improvements over the years this is still somewhat hacky.
In particular, note how we're assigning the select-element a larger width than its containing `dropdownToolbarButton`-element. This was done to prevent displaying *two* separate icons, i.e. the native and the PDF.js one, since it's the only way to handle this in older browsers (particularly Internet Explorer).
Given the currently supported browsers, there's however a better solution available: use `appearance: none;` to disable native styling of the select-element. [According to MDN](https://developer.mozilla.org/en-US/docs/Web/CSS/appearance#browser_compatibility), this is supported in all reasonably modern browsers.
This way we're able to simplify both the CSS rules and the JS-code that's used to adjust the `dropdownToolbarButton` width in a localization aware way.
Because of https://bugzilla.mozilla.org/show_bug.cgi?id=1582545, the padding-inline is by default 0.
0 is not really enough because of the outline, so just set it to 2px (it was 4px before the patch)
in order to have something visually correct.
I noticed the 256 % 3 (which is equal to 1) so I slighty simplify the code.
The sum of the 16 Uint8 doesn't exceed 2^12, hence we can just take the
sum modulo 3.
This method was originally added in PR 1157 (back in 2012), however its only call-site was then removed in PR 2423 (also in 2012).
Hence this method has been completely unused for nearly a decade, and it should thus be safe to remove it.
This patch first of all makes `isOffscreenCanvasSupported` configurable, defaulting to `true` in browsers and `false` in Node.js environments, with a new `getDocument` parameter. While you normally want to use this, in order to improve performance, it should still be possible for users to control it (similar to e.g. `isEvalSupported`).
The specific problem, as reported in issue 14952, is that the SVG back-end doesn't support the new ImageMask data-format that's introduced in PR 14754. In particular:
- When the SVG back-end is used in Node.js environments, this patch will "just work" without the user needing to make any code changes.
- If the SVG back-end is used in browsers, this patch will require that `isOffscreenCanvasSupported: false` is added to the `getDocument`-call.
While it can't hurt to localize the main error-messages, also localizing the error *details* has always seemed somewhat unnecessary since those are only intended for debugging/development purposes. However, I can understand why that's done since the GENERIC viewer used to expose this information in the UI; via the `errorWrapper` UI that's removed in PR 15533.
At this point, when any errors are simply logged in the console, it no longer seems necessary to keep localizing the error *details* in the default viewer.
*Please note:* The referenced issue is the only mention that I can find, in either GitHub or Bugzilla, of "GoToE" actions.
Hence why I've purposely settled for a very simple, and partial, "GoToE" implementation to avoid complicating things initially.[1] In particular, this patch only supports "GoToE" actions that references the /EmbeddedFiles-dict in the PDF document.
See https://web.archive.org/web/20220309040754if_/https://www.adobe.com/content/dam/acom/en/devnet/pdf/pdfs/PDF32000_2008.pdf#G11.2048909
---
[1] Usually I always prefer having *real-world* test-cases to work with, whenever I'm implementing new features.
This is yet another small piece of clean-up of the `FontLoader`-code, since we've not used this `id`-property for anything ever since PR 6571 (which landed almost seven years ago). Furthermore, by default we're also not even using that code-path now since the Font Loading API will always be used when available.
*Please note:* This is tagged `[api-minor]` since it's technically observable from the outside, however no user ought to be directly interacting with these CSS font rules.
This commit fixes the "Expected null to equal '401R'" errors that
surfaced after the Puppeteer 18 upgrade. Note that even before that
this would have been an improvement because it takes some time between
scripting being reported ready (i.e., triggering the execution of any
OpenActions) and those OpenActions actually completing execution, so
it's only safe to check which element is focused if we know an element
actually became focused.
In the Firefox PDF Viewer this has never been used, with the error message simply printed in the web-console, and (somewhat) recently we've also updated the viewer code to avoid bundling the relevant code there. Furthermore, in the Firefox PDF Viewer we're not even display the *browser* fallback bar any more; see https://bugzilla.mozilla.org/show_bug.cgi?id=1705327.
Hence it seems slightly strange to keep this UI around in the GENERIC viewer, and this patch proposes that we simply remove it to simplify/unify the relevant code in the viewer. In particular this also allows us to remove a couple of l10n-strings, which have always been unused in the Firefox PDF Viewer.
Currently the compatibility-file is loaded using a standard `import`-statement and while its code is enclosed in a pre-processor block, and thus is excluded in e.g. the MOZCENTRAL build-target, it still results in the *built* `pdf.js`/`pdf.worker.js` files having an effectively empty closure as a result.
By moving the checks from `src/shared/compatibility.js` and into `src/shared/util.js` instead, we can load the file using a build-time `require`-statement and thus avoid that closure.
Note that with these changes the compatibility-file will no longer be loaded in development mode, i.e. when `gulp server` is used. However, this shouldn't be a big issue given that none of its included polyfills could be loaded then anyway (since `require`-statements are being used) and that it's really only intended for the `legacy`-builds of the library.
Rather than "manually" looking up the l10n-string and then updating the button, we can (and probably even should) just update the l10n-id and then trigger proper translation for the button DOM-element.
This extends the approach used in PresentationMode to also cover the AnnotationEditor, and tries to handle the combination of both cases correctly.
In order to simplify the overall implementation we simply track the *first* seen "previous" cursorTool, and don't allow it to be reset as long as either PresentationMode or an AnnotationEditor is being used.
Note that this PR only adds the "underscore"-variant of *actually existing* ligatures, however the referenced PDF document also uses a couple of non-standard ones (e.g. `ft`, `Th`, and `fh`) that we cannot easily support without larger changes (since they don't have official Unicode-entries).
Given that it's clearly the PDF document, and its fonts, that's the culprit here it's not entirely clear to me that we actually want to attempt a larger refactoring/rewriting of the `glyphlist.js` code, assuming it's even generally possible. Especially when this patch alone already improves our copy-paste behaviour when compared to both Adobe Reader and PDFium, and that this is only the *second* time this sort of bug has been reported.
Fewer dependencies shouldn't be a bad idea in general, and given that the `node-canvas` package already include a `DOMMatrix` polyfill we can simply use that one instead.
Given that Firefox supports *synchronous* font loading, when the Font Loading API isn't being used, there's really no point including code which if called would just throw in the MOZCENTRAL build. (This is safe, since the `FontLoader.isSyncFontLoadingSupported`-getter always return `true` there.)
After the changes in PR 10539 (which landed over three years ago) the `FontLoader.bind` method can only be called with *a single* font at a time, hence the `_prepareFontLoadEvent` method obviously don't need to support multiple fonts any more.
By having just *one* class, and using pre-processor blocks directly in the relevant methods, we reduce the size of this code in the *built* `pdf.js` file.
Originally, when the `BaseFontLoader` abstraction was added in PR 9982, the idea was probably that additional build-targets would get their own implementations. Given that this hasn't happened in the four years since that landed, it doesn't seem meaningful to keep it around.
The existing `loadingContext` class-property can be simplified slightly, since we've not been using the `id`-property on the requests ever since PR 3477 (which landed nine years ago).
Furthermore, by default we're also not even using that code-path now since the Font Loading API will always be used when available.
Currently the `viewBookmark`-button, which is actually a `href`-element, gets an inconsistent `outline`.
Similarly, the `dialog`-buttons also have an inconsistent `outline` after the changes in PR 15438.
Finally, simplifies a couple of `border` rules since setting a border-width when "none" is being used doesn't seem meaningful.
This was done all the way back in PR 8361, for a mozilla-central test that's since been removed. As can be seen in the following search results, there's no `LoopbackPort` invocation outside of the PDF.js code itself: https://searchfox.org/mozilla-central/search?q=LoopbackPort&path=
Given that the `LoopbackPort` is only used in connection with "fake workers", which is something that we don't officially recommend/support, this doesn't seem like functionality that we want to keep exposing in the public API.
The changes in PR 15438 added a `border-radius` when input-elements are focused, however there's no radius when the same elements are hovered. Having the radius change, and not just the `border-color`, when input goes from hovered to focused feels a bit inconsistent (at least to me).
OperatorList.addOp can trigger a flush if it's required, hence the values passed to it must
be correctly initialized in order to avoid some wrong values in the renderer.
Because of that a clip path was considered as empty, nothing was clipped, hence the wrong
rendering in bug 1791583.
*This effectively replaces PR 15465.*
As outlined in https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Map/forEach, the argument order when iterating through a `Map` is actually `value, key`.
Ignoring the incorrect Array used in the old code, I cannot imagine that this would've worked anyway since we didn't use the actual `setTimeout`-functionRefs to clear the timeouts; please refer to the `setTimeout`/`setInterval` methods in the `SandboxSupportBase.createSandboxExternals` method.
Since there are no script engine with XFA, the FormCalc parser is not used irl.
The bug @nmtigor noticed was hidden by another one (the wrong check on `match`).
Some z-index have been added in the annotation layer because the elements inside are re-ordered
in order to improve accessibility.
Hence we must add a "high" z-index on the annotation editor layer in order to avoid any bad
interaction between the different layers.
Most of the `String.prototype.search` call-sites found throughout the code-base is actually not necessary, since we usually only want a *boolean*, and those can be replaced with `RegExp.prototype.test` instead.
The default outline for a focused text input is not that bad but for any reason when changing
the background color, all the good default border/outline properties are lost (it's the same
behaviour in Edge).
So in order have something consistent in HCM/non-HCM, a 2px-border+1px-outline (on @MReschenberg
advices) is added when an input is focused with different colors depending on HCM.
While working on the above issue, I noticed few bugs I fixed when in HCM:
- input, button and select have some default properties which have been created at a time where
annotation layer didn't exist, hence this patch remove them and set those properties where
they should live;
- some elements (like the main toolbar) is using a box-shadow which is invisible in HCM, hence
it's replaced by a border-bottom in HCM;
- some separators are invisible in HCM, hence use GrayText color to render them correctly;
- the options for the zoom selection were invisible in HCM with Desert (one of the Windows 11
themes).
By force-quitting the browser while the FullScreen API is active, we don't get a chance to exit PresentationMode *cleanly* and some of its state thus remains (via the `ViewHistory`).
To try and improve things here we can skip updating the Scroll/Spread-mode while PresentationMode is active, since they will be changed when entering PresentationMode, which seems to help and is really the best that we can do here (and what the issue describes is very much an edge-case anyway).
In the `legacy`-builds we (obviously) support the currently maintained Firefox ESR version, and looking at the [release history](https://wiki.mozilla.org/Release_Management/Calendar) those are officially supported (by Mozilla) for about 1-1.5 years.
However, for non-Firefox browsers the `legacy`-builds currently attempt to "support" browsers that are approximately *three* years old.[1] Historically, in the PDF.js project, trying to support old browsers have caused some maintenance problems and even delayed adoption of new web-platform features/functionality.
To lessen the support burden, given that the primary purpose of the PDF.js library is still to develop the *built-in* Firefox PDF Viewer, this patch proposes that the upcoming *major* release changes the minimum supported browsers/environments as follows:
- Chrome 85, which was released on 2020-08-25; see https://en.wikipedia.org/wiki/Google_Chrome_version_history
- Firefox ESR (as before); see https://wiki.mozilla.org/Release_Management/Calendar
- Safari 14, which was released on 2020-09-16; see https://en.wikipedia.org/wiki/Safari_version_history#Safari_14
- Node.js 14 (as before), which is now explicitly listed to prevent it from accidentally breaking; see https://en.wikipedia.org/wiki/Node.js#Releases
---
[1] In older browsers some functionality may not be available and generally we'll ask users to update to a modern browser when bugs, specific to old browsers, are being reported.
The [official Chrome extension](https://chrome.google.com/webstore/detail/pdf-viewer/oemmndcbldboiebfnladdacbdfmadadm) has unfortunately not been updated for *three years*, which means that it's currently missing out on years worth of bug fixes, performance improvements, and new features.
In particular, the Chrome extension suffers from a known bug with non-embedded standard fonts; see issue 13669 for details.
For the time being, this patch proposes that we *temporary* make the following changes:
- Remove the mention of the official Chrome extension from the main README, since it seems unfortunate to somewhat prominently recommend users an old and partially non-working extension.
- Don't run the `gulp lint-chromium` task as part of the CI, since in addition to the official extension not having been updated its code is also not being actively maintained.[1]
Once the official Chrome extension has been updated, and it's being actively maintained again, this patch should be simple enough to revert.
---
[1] The last commits, which aren't e.g. linting or general code-maintenance related, happened a year ago now.
There's no point in having this variable defined (implicitly) as `undefined` in e.g. the Firefox PDF Viewer.
By defining it with `var` and using an ESLint ignore, rather than `let`, we can move it into the relevant pre-processor block instead. Note that since the entire viewer-code is placed, by Webpack, in a top-level closure this variable will thus not become globally accessible.
After the changes in PR 15391 one separator may now become visible too soon when the viewer is narrow, applies e.g. to the MOZCENTRAL viewer, since the wrong CSS class is being used.
The reason that this happens is that only the GENERIC viewer includes the "openFile"-buttons, and we thus need the separator to also be conditionally defined.
This is a slightly speculative change, based on something that I happened to notice while browsing MDN, to hopefully prevent PDF.js from outright breaking in older browsers.
According to the following information on MDN, Safari didn't implement support for the necessary features until version 14:
- https://developer.mozilla.org/en-US/docs/Web/API/MediaQueryList#browser_compatibility
- https://developer.mozilla.org/en-US/docs/Web/API/MediaQueryList/change_event#browser_compatibility
Given the browsers that we currently support only older versions of Safari should be affected, hence it seems reasonable to simply disable the functionality rather than trying to polyfill it.
(It's interesting how it's very often Safari which is *much* slower than the other browsers at implementing new features.)
After the changes in PR 14112 the `PDFViewer`-class is now "identical" to the `BaseViewer`-class and the `PDFSinglePageViewer`-class is just a very thin wrapper around the `BaseViewer`-class.
Hence we can rename these files, and also remove the abstract `BaseViewer`-class, which helps reduce some unnecessary "closures" in the *built* viewer.
*Please note:* These changes are made in two separate commits, to allow GitHub to preserve `blame` for the affected files.
After the changes in PR 14112 the `PDFViewer`-class is now "identical" to the `BaseViewer`-class and the `PDFSinglePageViewer`-class is just a very thin wrapper around the `BaseViewer`-class.
Hence we can rename these files, and also remove the abstract `BaseViewer`-class, which helps reduce some unnecessary "closures" in the *built* viewer.
*Please note:* These changes are made in two separate commits, to allow GitHub to preserve `blame` for the affected files.
This patch updates a bunch of older code, that makes conditional function calls, to use optional chaining rather than `if`-blocks.
These mostly mechanical changes reduce the size of the `gulp mozcentral` build by a little over 1 kB.
*Please note:* This is only a, hopefully generally helpful, work-around rather than a proper solution to issue 15292.
There's something that's "special" about the Type1 fonts in the referenced PDF document, since we don't manage to find any actual font programs and thus cannot render anything.
Given that it shouldn't make sense for a Type1 font program to ever be empty, since that means that there's no glyph-data to render, we simply fallback to a standard font to at least try and render *something* in these rare cases.
Given that the change in PR 13393 was slightly speculative, given the lack of test-cases, let's just revert part of that to fix the referenced issue.
Based on a quick look at old issues and existing test-cases, it seems that most (if not all) PDF documents that benefit from using the font-data in this way lack any /ToUnicode maps which should mean that they're unaffected by these changes.
Given that the official Bower website, since almost five years, has been advising users to utilize other tools it doesn't seem entirely necessary to keep including the `bower.json` file in the `pdfjs-dist` repository; see e.g. https://bower.io/blog/2017/how-to-migrate-away-from-bower/
This patch proposes removing the `browserify` example for the following reasons:
- The last `browserify` release was almost two years ago, according to both https://github.com/browserify/browserify/releases and https://www.npmjs.com/package/browserify?activeTab=versions
- The project no longer seems to be actively maintained, since so far this year there's only been *a single* (seemingly trivial) patch merged; see https://github.com/browserify/browserify/commits/master
- Because of the previous points `browserify` doesn't support modern and up-to-date JavaScript features, as evident from e.g. issue 14731 and multiple issues found in https://github.com/browserify/browserify/issues
- Our `browserify` example is most likely not very commonly used, judging by the very low volume of issues/PRs related to it. Looking at the `git` history of that example the only changes have been lint- or maintenance-related.[1]
- Providing an example for a framework that's no longer actively maintained doesn't seem like a good idea in general, since we probably don't want to steer users towards using (possibly) older frameworks.
- Given that we've never used `browserify` in the PDF.js project, it's also quite difficult to provide support for the example.
---
[1] It's interesting to compare with the `webpack` example, since that's generated both issues *and* also PRs (for missing features) from users.
Note that this patch implements the `SetOCGState`-handling in `PDFLinkService`, rather than as a new method in `OptionalContentConfig`[1], since this action is nothing but a series of `setVisibility`-calls and that it seems quite uncommon in real-world PDF documents.
The new functionality also required some tweaks in the `PDFLayerViewer`, to ensure that the `layersView` in the sidebar is updated correctly when the optional-content visibility changes from "outside" of `PDFLayerViewer`.
---
[1] We can obviously move this code into `OptionalContentConfig` instead, if deemed necessary, but for an initial implementation I figured that doing it this way might be acceptable.
It slightly helps to reduce the code size and its complexity.
But the cool thing is that it allows to copy/paste some anntations from a pdf
to an other.
Apparently this is implemented in e.g. Adobe Reader, and the specification does support it, however it cannot be commonly used in real-world PDF documents since it took over ten years for this feature to be requested.
A number of Annotation-types are currently creating their own PopupAnnotations, since they need to use a custom `trigger`-element. However, because of where that check is currently implemented[1] we end up attaching empty/unused containers for those PopupAnnotations to the DOM[2]; see e.g. the `annotation-line.pdf` file in the test-suite for one example.
By instead moving the types-check into the `PopupAnnotationElement` constructor, we can completely skip those PopupAnnotations that are being explicitly handled elsewhere.
Note that I don't *believe* that this is a new issue, although I've not tried to bisect it, but this likely goes back quite some time (possibly even as far as PR 8228).
---
[1] In the `PopupAnnotationElement.render` method.
[2] Please note that the actual Popup-element *itself* isn't being attached/rendered here, just its container which by itself serves no purpose as far as I can tell.
There's three notable exceptions here:
- The `saveDocument` one is converted into a permanent `warn`, since it still works when the `annotationStorage` is empty although it's (obviously) less efficient than `getData`.
- The `fallbackWorkerSrc` functionality (for browsers), since just removing it would risk too much third-party breakage.
- The SVG back-end, since a final decision is yet to be made. (It might be completely removed, or left as-is in an essentially "frozen" state.)
Note that this patch prepends the document title with "* ", rather than only "*" as suggested in the bug, since there's nothing that says that a PDF document cannot specify a title[1] beginning with an asterisk. To reduce possible confusion, having a space between the "editing marker" and the actual document title thus cannot hurt as far as I'm concerned.
In order to notify the viewer when all `AnnotationEditor`s have been removed, we utilize the existing `onAnnotationEditor`-callback to allow the document title to be updated as necessary.
Finally, this patch makes the following (slightly unrelated) changes:
- Rename the `AnnotationStorage.removeKey` method to just `AnnotationStorage.remove` instead. This is consistent with e.g. the `has`-method and should suffice to explain what it does.
- Remove the `AnnotationStorage.hasAnnotationEditors` getter, since the viewer now tracks the necessary state internally. This avoids unnecessarily having to iterate through the `AnnotationStorage`-instance when saving/printing the document.
---
[1] Using either an /Info dictionary or a /Metadata stream.
This functionality has never been used anywhere in the PDF.js library/viewer itself, since it was added in 2013.
Furthermore this functionality is, and has always been, *completely untested* and also unmaintained.
Finally, there's (at least) one old issue about `appendImage` not returning the correct position; see issue 4182.
All-in-all, it seems that keeping very old, untested, unmaintained, and partially broken code around probably isn't what we want here.
(On the off-chance that any future a11y-work requires getting access to image-positions, it'd likely be much better to re-implement the necessary functionality from scratch and also make sure that it's properly tested from the beginning.)
This old method, which is only used with the `imageLayer` functionality, is essentially just a re-implementation of the existing `Util.applyTransform` method.
The password dialog can be cancelled in three different ways:
- By clicking on its "Cancel"-button.
- By pressing the Escape-key.
- By force-opening another dialog, although this shouldn't happen in practice.
Here the "Cancel"-button case is slightly special since it'll trigger `PasswordPrompt.#cancel` *twice*, first directly via the click and secondly via the "close" event on the `dialog`-element.
While this shouldn't, as far as I know, cause any bugs it's nonetheless inconsistent with the other cases outlined above. To improve this we can simply attempt to *close* the password dialog instead, and then rely on the "close" event to run the `PasswordPrompt.#cancel` method.
Currently we simply use the Babel `preset-env` in the `legacy`-builds of the PDF.js library. This has the side-effect of transpiling the code for *very old* browsers/environments, including ones that we (since many years) no longer support which unnecessarily bloats the size of the `legacy`-builds.
For the CSS files we're only targeting *the supported browsers*, and it's thus possible to extend that to also apply to Babel.
One of the most significant changes, with this patch, is that we'll no longer polyfill `async`/`await` in the `legacy`-builds. However, this shouldn't be an issue given the browsers that we currently support in PDF.js; please refer to:
- https://github.com/mozilla/pdf.js/wiki/Frequently-Asked-Questions#faq-support
- https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Statements/async_function#browser_compatibility
Currently in `disableWorker=true` mode it's possible that opening of password-protected PDF documents outright fails, if an *incorrect* password is entered. Apparently the event ordering is subtly different in the non-Worker case, which causes the password-callback to be updated *before* the dialog has been fully closed.
To avoid that we'll utilize a `PromiseCapability` to keep track of the state of the password dialog, such that we can delay both re-opening and (importantly) updating of the password-callback until doing so is safe.
This patch *may* also fix issue 15330, but it's impossible for me to tell.
*This is a follow-up to PR 14869.*
In the old code we're accidentally "swallowing" part of the event-details, which explains why the annotationLayer didn't render.
One thing that made debugging a lot harder was the lack of error messages, from the viewer, and a few `PDFPageView`-methods were updated to improve this situation.
- Remove the `typeof Worker` check, since all browsers have had `Worker` support for many years now; see https://developer.mozilla.org/en-US/docs/Web/API/Worker#browser_compatibility
Furthermore the `new Worker(...)` call is wrapped in try-catch, which means that we'll still fallback to "fake workers" if necessary.
- Limit the `fallbackWorkerSrc` handling, in the `PDFWorker.workerSrc` getter, to only GENERIC builds since that's the only place where it's defined anyway.
Given that the code is written with JavaScript module-syntax, none of this functionality will "leak" outside of this file with these change.
By removing this closure the file-size is decreased, even for the *built* `pdf.worker.js` file, since there's now less overall indentation in the code.
This was moved into the `src/display/`-folder in PR 15110, for the initial editor-a11y patch. However, with the changes in PR 15237 we're again only using `binarySearchFirstItem` in the `web/`-folder and it thus seem reasonable to move it back there.
The primary reason for moving it back is that `binarySearchFirstItem` is currently exposed in the public API, and we always want to avoid that unless it's either PDF-related functionality or code that simply must be shared between the `src/`- and `web/`-folders. In this case, `binarySearchFirstItem` is a general helper function that doesn't really satisfy either of those alternatives.
Currently when the `TextAccessibilityManager.enabled` method is called, we'll update `aria-owns` for any pre-existing elements. This obviously makes sense when e.g. zooming/rotating in the viewer, since the annotationLayer/annotationEditorLayer is kept in those cases.
However when the page is *fully* reset, e.g. as result of going out-of-view and thus being evicted from the cache, we still keep the `#textNodes`-Map around. This causes us to set the `aria-owns` attribute (in the textLayer) for an element that doesn't actually exist any more, which as far as I'm concerned seems incorrect. In this case the element will simply, as already implemented, be re-inserted when the annotationLayer/annotationEditorLayer renders again.
Given that the code is written with JavaScript module-syntax, none of this functionality will "leak" outside of this file with these changes.
For e.g. the `gulp mozcentral` command the *built* `pdf.worker.js` file-size decreases `~2 kB` with this patch, and most of the improvement comes from having less overall indentation in the code.
Given that the code is written with JavaScript module-syntax, none of this functionality will "leak" outside of this file with these changes.
By removing this closure the file-size is decreased, even for the *built* `pdf.worker.js` file, since there's now less overall indentation in the code.
This patch doesn't structurally change the text layer: it just adds some aria-owns
attributes to some spans.
The aria-owns attribute expect to have an element id, hence it's why it adds back an
id on the element rendering an annotation, but this id is built in using crypto.randomUUID
to avoid any potential issues with the hash in the url.
The elements in the annotation layer are moved into the DOM in order to have them in the
same "order" as they visually are.
The overall goal is to help screen readers to present to the user the annotations as
they visually are and as they come in the text flow.
It is clearly not perfect, but it should improve readability for some people with visual
disabilities.
According to MDN `Path2D` is available in all browsers that we currently support, see https://developer.mozilla.org/en-US/docs/Web/API/Path2D#browser_compatibility
Hence only Node.js is currently lagging behind here, and requires that we keep the old code as a fallback in the `compileType3Glyph` function. However, there's an open PR in the `node-canvas` repository for adding `Path2D` support.
As far as I'm concerned, there's two possible solutions here:
- We land this patch now, since it removes unnecessary code in e.g. the Firefox PDF Viewer, which means that compilation of Type3 glyphs will be disabled in Node.js until that PR is landed.[1]
If users report bugs about Type3 glyphs looking "inconsistent" in Node.js and/or being slow to render, we could perhaps encourage them to upvote and otherwise help out getting that PR landed?
- We wait for the mentioned PR to land *first*, before moving forward with this patch. Given that there's been no updates on that PR for almost two months, this alternative may possibly take a while.
---
[1] Note that Type3 fonts are first of all not very common in PDF documents, and secondly that compilation only applies specifically to Type3 glyphs that contain /ImageMask-data (i.e. not all Type3 fonts are affected).
This exports the same constants as the viewer components, but in the default viewer. To avoid bloating the global-scope the constants are added to a new `PDFViewerApplicationConstants` object[1], which also allows us to skip this in builds where it's not actually needed (e.g. the Firefox *built-in* PDF Viewer).
*Please note:* I'm not completely sold on this idea, and thus wouldn't mind the patch being rejected, since we probably don't want to export every single viewer constant this way. (And it may seem a bit arbitrary, to users, why some constants are exported and others are not.)
---
[1] Somewhat similar to the existing `PDFViewerApplicationOptions` structure.
In addition to the existing `LinkTarget` constant, used by the `PDFLinkService`-constructor, this patch exports the following constants in the viewer components:
- `ScrollMode` and `SpreadMode`, since the `BaseViewer` has getters/setters which work with those constants.
- `RenderingStates`, since that one may be helpful when using `PDFPageView` directly.
While this has always worked, as a consequence of the implementation, it's never been officially supported.
In addition to adding basic unit-tests, this patch also introduces a couple of new JSDoc `@typedef`s in the API to avoid overly long lines.
By invoking the `reset` methods *last* in the `Toolbar`/`SecondaryToolbar`-constructors, we ensure that the "toolbarreset"/"secondarytoolbarreset"-events are actually handle when the viewer loads. Note that previously those events were dispatched *before* the relevant event-listeners had been attached.
With this small change we can avoid inconsistent initial toolbar-state, specifically in the case when the viewer is *reloaded* (since Firefox keeps the HTML-state on "soft" reloads).
By doing this in the worker-thread this code will only need to run *once*, whereas currently re-rendering of a page forces this to be repeated (e.g. after it's been scrolled out-of-view and then back into view again).
When a FreeText editor is pasted then it hasn't an editorDiv yet when added
to the layer, hence it's empty.
So this patch just move the call to addToAnnotationStorage to ensure we've
what we need.
An annotation doesn't have to be in the text flow, hence it's likely a bad idea
to insert its text in the text layer. But the text must be visible from a screen
reader point of view so it must somewhere in the DOM.
So with this patch, the text from a FreeText annotation is extracted and added in
a div in its HTML counterpart, and with the patch #15237 the text should be visible
and positioned relatively to the text flow.
Given that this image is intended specifically for the default viewer, we simply use the CSS preprocessor to remove the image reference in the `gulp components` build.
Considering that the issue only affects a CSS file, I don't believe that replacing the *just released* PDF.js version is actually necessary here.
It doesn't make sense to use a page-canvas that's *smaller* than the resulting thumbnail, since that causes the image to be upscaled which results in a blurry thumbnail. Note that this doesn't normally happen, unless a very small zoom-level is used in the viewer.
To improve performance of the sidebar we use the page-canvases to generate the thumbnails whenever possible, since that avoids unnecessary re-rendering when the sidebar is open. This works generally well, however there's an old problem in PDF documents that contain interactive forms (when those are enabled): Note how the thumbnails become partially (or fully) blank, since those Annotations are not included in the OperatorList.[1]
We obviously want to keep using the `PDFThumbnailView.setImage`-method for most documents, however we need a way to skip it only for those pages that contain interactive forms.
As it turns out it's unfortunately not all that simple to tell, after the fact, from looking only at the OperatorList that some Annotations were skipped. While it might have been possible to try and infer that in the viewer, it'd not have been pretty considering that at the time when rendering finishes the annotationLayer has not yet been built.
The overall simplest solution that I could come up with, was instead to include a *summary* of the interactive form-state when doing the final "flushing" of the OperatorList and expose that information in the API.
---
[1] Some examples from our test-suite: `annotation-tx2.pdf` where the thumbnail is completely blank, and `bug1737260.pdf` where the thumbnail is missing the "buttons" found on the page.
Currently all of these factories take a bunch of (randomly ordered) parameters, which first of all doesn't look that nice in the `PDFPageView`-class when some parameters are optional.
Furthermore, it also makes deprecation/removal of any existing parameter a *potentially* breaking change.
Finally, using an Object will provide a small amount of "documentation" at the call-site which isn't really the case with a bunch of "regular" parameters.
Note that all of the `viewer component` examples still work as-is with this patch, which is why I don't believe that we necessarily have to deprecate in the usual fashion.
Currently some `OPS.beginAnnotation` arguments will contain a `Number` value for the `isUsingOwnCanvas`-parameter, or in some cases an `undefined` value, which is inconsistent from an API perspective.
We can undo/redo a command which will at some point add a command in the queue: typically
it can happening when redoing an addition.
So the idea is to lock the queue when undoing/redoing.
Rather than always disable `PDFThumbnailView.setImage` as soon as user has changed the visibility of the Optional Content, we can utilize the new method added in the previous patch to improve thumbnail performance. Note in particular how, in the old code, even *resetting* of the Optional Content to its default state wouldn't enable `PDFThumbnailView.setImage` again.
While slightly unrelated, this patch also removes the `PDFThumbnailViewer._optionalContentConfigPromise`-property since it's completely unused.
This will allow us to improve the `PDFThumbnailView.setImage` handling in the viewer, and thanks to the added caching this should be reasonbly efficient.
Given that Optional Content visibility is only intended/supported to be updated via the `OptionalContentConfig.setVisibility`-method, this patch actually enforces that now.
Note that this will be used by the next patch in the series, and will help prevent inconsistent state in the `OptionalContentConfig`-class.
*Please note:* This patch also uncovered a pre-existing bug, related to iterating through the visibility groups in the constructor, for the `baseState === "OFF"` case.
- After undoing a deletion of several editors, they appeared to be selected (they had a red border)
when in fact they were not, consequently, this patch aims to remove the selectedEditor class when
an editor is removed;
- Add a test with some ink editors.
The previous version was maybe functional but definitely painful to maintain
(maybe more efficient... I don't know) so this patch aims to simplify it and
it adds some basic unit tests.
Given that the SVG back-end is not defined anywhere except in GENERIC builds, we can remove a bit more unnecessary code in e.g. the Firefox PDF Viewer.
The elements in the annotationEditor layer are rearranged to make them
more accessible, but we must draw them in the order they have been created,
hence this patch adds a z-index to the editors.
- This way, the keyboard callbacks are called even if the page has not
the focus, hence the user doesn't have to guess that they have to click
on the page which is a bit painful especially in Ink mode.
- Add two keyboard shortcuts to commit a Freetext editor (ctrl+enter and
escape).
In the referenced PDF document the fonts have /CIDToGIDMap-entries that cannot be loaded. Hence, only when `ignoreErrors` is set, we'll now ignore these corrupt /CIDToGIDMap-entries and fallback to simply assume that no such data is available.
Given that this is *clearly* a case of a corrupt PDF document, there's no guarantee that this will "fix" things in the general case since a /CIDToGIDMap may be *required* in order for some composite fonts to render correctly. However, attempting to render *something* is surely better than skipping a font altogether.
- In the annotationEditorLayer, reorder the editors in the DOM according
the position of the elements on the screen;
- add an aria-owns attribute on the "nearest" element in the text layer
which points to the added editor.
- in using the global clipboard, it'll be possible to copy from a
pdf and paste in an other one;
- it'll allow to edit a previously created annotation;
- copy the editors in the current page.
Previously, we had to set the #allowClick property by hand which was
a bit painful because it's easy to overlook one case or an other.
So with this patch a new editor (for now FreeText one only because the
Ink one is a bit different) is created on the first click if none is selected
on mousedown, else the first click will just commit the data and then the
second will creater a new editor.
The problem is clearly visible when the thickness is at max.
It's mainly because the thickness was not taken into account when
translating the div but it was when the line is drawn on the canvas.
Note that these cases, which are all in older code, were found using the [`unicorn/no-for-loop`](https://github.com/sindresorhus/eslint-plugin-unicorn/blob/main/docs/rules/no-for-loop.md) ESLint plugin rule.
However, note that I've opted not to enable this rule by default since there's still *some* cases where I do think that it makes sense to allow "regular" for-loops.
Given that the SVG back-end is not defined anywhere except in GENERIC builds, we can remove a little bit more unnecessary code in e.g. the Firefox PDF Viewer.
Note how we currently throw a "raw" Error, which is problematical since all of the `PartialEvaluator.loadFont` call-sites expect a Promise to be returned. Furthermore, this also means that we don't benefit from the fallback code-path that now exists below.
*Please note:* Unfortunately I don't have a test-case that fails without this patch, since it's something I happened to notice when reading the code while working on another patch.
Previously it was created only on mouseover event but on a touch screen
there are no fingerover event...
The idea behind creating the ink editor on mouseover was to avoid to have
a canvas on each visible page.
So now, when the editor is created, the canvas has dimensions 1x1 and
only when the user starts drawing the dimensions are set to the page ones.
The current version 2.5 is from September 2014, which is almost 8 years
old now. The new version 3.19.0 is from November 2017, which is still
almost 5 years old, but is a step forward towards eventually using the
most recent version. Note that we currently can't update any further;
see #11802 for the details.
Fortunately using this newer version only required a few changes:
- The `ttx` output regexes needed updating to ignore comments that `ttx`
now puts after some XML nodes (`<!-- ... -->` and `/* ... */`).
- The `ttx` invocation now explicitly uses `python2` (except on Windows
where this alias doesn't exist) since otherwise the font tests can't
be run on modern systems anymore given that `python` is nowadays an
alias for `python3`, and it now points at the new location of the
`ttx.py` file since the `Tools` folder got removed.
- The note about needing a 32-bit Python 2.6 version is dropped since
it's obsolete: this version (and also the existing one already) work
just fine on a 64-bit Python 2.7 as well.
We want to avoid adding regular `id`s to xfaLayer-elements, since that means that they become "linkable" through the URL hash in a way that's not supported/intended. This could end up clashing with "named destinations", and that could easily lead to bugs; see issue 11499 and PR 11503 for some context.
Rather than using `id`s, we'll instead use a *custom* `data-element-id` attribute such that it's still possible to access the DOM-elements directly if needed. *Please note:* This is basically the xfaLayer-equivalent of PR 15057.
- and because of rounding errors it led to slightly resize again and again
the ink container;
- when zooming the size is changing but not the ratio, so in this case we
don't need to change the dimension of the container.
`HTMLSectionElement` is not part of the DOM, so the generated typescript definitions contain a non-existing type.
HTML Section elements have to be handled as simple `HTMLElements`.
fixing punctuation and lint problems
[jsdoc] failing typescript builds - wrong type
Rather than including all of this external code in the PDF.js repository, we should be using the npm package instead.
Unfortunately this is slightly more complicated than you'd hope, since the `fit-curve` package (which is older) isn't directly compatible with modern JavaScript modules.
In particular, the following cases needed to be considered:
- For the development viewer (i.e. `gulp server`) and the unit-tests, we thus need to build a fitCurve-bundle that can be directly `import`ed.
- For the actual PDF.js build-targets, we can slightly reduce the sizes by depending on the "raw" `fit-curve` source-code.
- For the Node.js unit-tests, the `fit-curve` package can be used as-is.
- this way the context menu in Firefox can take into account what we
have in the clipboard, if an editor is selected, ...
- when the user will click on a context menu item, an action will be
triggered, hence this patch adds what is required to handle it;
- some tests will be added in the Firefox' patch.
This extends PR 13461, by also building a fallback bounding box for Type3 fonts that contain a much too small /FontBBox-entry.
*Please note:* While this patch improves things overall, copy-and-pasting still doesn't work perfectly for this document. In particular the lowercase letter "c" cannot be selected/copied, however this can be reproduced in both Adobe Reader and PDFium (in Google Chrome) too, which is caused by a lack of proper /ToUnicode-data in the PDF document.
Rather than forcing the user to *manually* call `setDimensions`, which is also breaking any existing third-party code, it seems that we can simply let the `AnnotationLayer.{render, update}`-methods handle that internally.
As far as I can tell, based on testing manually in the viewer *and* running the browser-tests, everything still appears to work correctly with this patch.
After the changes in PR 15036, the trigger-element created in `FileAttachmentAnnotationElement.render` is now too small. This can be fixed by using the same approach as in PR 15065, and the patch can be tested using the `annotation-fileattachment.pdf` document in the test-suite.
- Simplify how we look-up the DOM-element, which should also be a tiny bit more efficent.
- Use private class-fields, rather than property-names prefixed with underscores.
- Inline the `#updateBar` helper-method directly in the `percent`-setter, since having a separate method doesn't seem necessary in this case.
- Set the `indeterminate`-class on the ProgressBar DOM-element, to simplify the code.
Finally, also (slightly) re-factors the `PDFViewerApplication.progress`-method to make it a bit smaller.
- for example in Dusk theme (Windows 11), black appears to be white, so
the user will draw something in white. But if they want to print or
save the used color must be black.
- fix a bug with the color input which only accepts hex string colors;
- adjust outline color of the selected/hovered editors in HCM.
Now that it's possible to change the font-size, this placeholder string feels a little bit long (especially for larger font-sizes).
Given that Editing is not enabled/released yet, I hope that it should be fine to update this without changing the l10n-id.
This replaces the boolean `annotationEditorEnabled` option/preference with a "proper" `annotationEditorMode` one. This way it's not only possible for the user to control if Editing is enabled/disabled, but also which *specific* Editing-mode should become enabled upon PDF document load.
Given that Editing is not enabled/released yet, I cannot imagine that changing the name and type of the option/preference should be an issue.
Note that even though Puppeteer got a major version bump the changelog
doesn't include compatibility changes that are relevant to us; please
see https://github.com/puppeteer/puppeteer/releases.
Currently we're loading the `web/annotation_layer_builder.css` and `web/xfa_layer_builder.css` files *directly* during the reference tests.
This becomes a problem is we want to reduce duplication in the CSS-files, e.g. by placing *common* rules in the `web/pdf_viewer.css` file.
Given that `gulp components` is already being utilized when running tests, we can thus use that to instead depend on the *entire* viewer-components CSS-file in the reference tests.
Similar to other recent patches, see e.g. PR 15057, we don't want to add these kind of `id`s to DOM-elements since they shouldn't become "linkable" through the URL hash.
*Please note:* This patch can be tested, in the viewer, with e.g. `bug1737260.pdf` from the test-suite.
- As in the annotation layer, use percent instead of pixels as unit;
- handle the rotation of the editor layer in allowing editing when rotation
angle is not zero;
- the different editors are rotated counterclockwise in order to be usable
when the main page is itself rotated;
- add support for saving/printing rotated editors.
Given that Annotations can also have an `OC`-entry, we need to take that into account when generating their operatorLists.
Note that in order to simplify the patch the `getOperatorList`-methods, for the Annotation-classes, were converted to be `async`.
- the annotations must be rendered in the same order as the chronological one.
- fix a bug in document.js which avoids to read a saved pdf correctly in Acrobat:
there is no need to reset the xref state: it's done in worker.js once everything
has been saved.
Given that printing is triggered *synchronously* in browsers, it's thus possible for scripting (in PDF documents) to modify the Annotation-data while printing is currently ongoing.
To work-around that we add a new printing-specific `AnnotationStorage`, where the serializable data is *frozen* upon initialization, which the viewer can thus create/utilize during printing.
For encrypted PDF documents without the required permissions set, this patch adds support for disabling of Annotation-editing. However, please note that it also requires that the `pdfjs.enablePermissions` preference is set to `true` (since PDF document permissions could be seen as user hostile).[1]
As I started looking at the issue, it soon became clear that *only* trying to fix the issue without slightly re-factor the surrounding code would be somewhat difficult.
The following is an overview of the changes in this patch; sorry about the size/scope of this!
- Use a new `AnnotationEditorUIManager`-instance *for each* PDF document opened in the GENERIC viewer, to prevent user-added Annotations from "leaking" from one document into the next.
- Re-factor the `BaseViewer.#initializePermissions`-method, to simplify handling of temporarily disabled modes (e.g. for both Annotation-rendering and Annotation-editing).
- When editing is enabled, let the Editor-buttons be `disabled` until the document has loaded. This way we avoid the buttons becoming clickable temporarily, for PDF documents that use permissions.
- Slightly re-factor how the Editor-buttons are shown/hidden in the viewer, and reset the toolbar-state when a new PDF document is opened.
- Flip the order of the Editor-buttons and the pre-exising toolbarButtons in the "toolbarViewerRight"-div. (To help reduce the size, a little bit, for the PR that adds new Editor-toolbars.)
- Enable editing by default in the development viewer, i.e. `gulp server`, since having to (repeatedly) do that manually becomes annoying after a while.
- Finally, support disabling of editing when `pdfjs.enablePermissions` is set; fixes issue 15049.
---
[1] Either manually with `about:config`, or using e.g. a [Group Policy](https://github.com/mozilla/policy-templates).
Note how the "page"-div, "canvasWrapper"-div, and `textLayer`-div all have *integer* dimensions (rounded down) rather than using the "raw" viewport-dimensions.
Hence it seems reasonable that the same should apply to the "annotationLayer"-div, now that it's explicit dimensions set.
- Let the `Page.save`-method filter out "empty" entries, similar to the `Page._parsedAnnotations`-getter, since that on its own already simplifies the "SaveDocument"-handler a tiny bit.
- The existing `reduce` and `concat` construction isn't exactly a wonder of readability :-)
Thanks to modern JavaScript features it should be possible to replace all of this with `Array.prototype.flat()` instead, which at least to me feels a lot easier to understand.
There are obviously cases where using `concat` makes perfect sense, since that method doesn't change any of the existing Arrays; see https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Array/concat
However, in a few cases throughout the code-base that's not an issue and using `concat` only leads to unnecessary intermediate allocations. With modern JavaScript we can thus replace those with a combination of `push` and spread-syntax, which wasn't originally possible when the code was written.
- each annotation has its coordinates/dimensions expressed in percentage,
hence it's correctly positioned whatever the scale factor is;
- the font sizes are expressed in percentage too and the main font size
is scaled thanks a css var (--scale-factor);
- the rotation is now applied on the div annotationLayer;
- this patch improve the rendering of some strings where the glyph spacing
was not correct (it's a Firefox bug);
- it helps to simplify the code and it should slightly improve the update of
page (on zoom or rotation).
We want to avoid adding regular `id`s to Annotation-elements, since that means that they become "linkable" through the URL hash in a way that's not supported/intended. This could end up clashing with "named destinations", and that could easily lead to bugs; see issue 11499 and PR 11503 for some context.
Rather than using `id`s, we'll instead use a *custom* `data-element-id` attribute such that it's still possible to access the Annotation-elements directly.
Unfortunately these changes required updating most of the integration-tests, and to reduce the amount of repeated code a couple of helper functions were added.
This option is not used, nor has it ever been used, in the *built-in* Firefox PDF Viewer. Hence we can define it only for the environments where it makes sense instead.
This should really have been done as part of PR 12470, since it's now possible to directly set the `defaultUrl`-option without having to fallback to `var`-usage.
- Since the border belongs to the section containing the HTML
counterpart of an annotation, this section must be hidden when
a JS action requires it;
- it wasn't possible to hide a button in using JS.
- Right now, we must select the tool, then click to select a page and
click to start drawing and it's a bit painful;
- so just create a new ink editor when we're hovering a page without one.
Given that the SVG back-end is not defined anywhere except in GENERIC builds, we can remove a little bit of unnecessary code in e.g. the Firefox PDF Viewer.
This appears to be a Microsoft-specific version of the regular Arial font, hence we simply map this to Helvetica in the same way that we treat many other Arial-named fonts.
Apparently the ESLint rule added in PR 15031 wasn't able to catch all cases that can be converted, which is probably not all that surprising given how some of these call-sites look.
- Use `Element.prepend()` to insert nodes before all other ones in the element, rather than using `firstChild` with `insertBefore`-calls; see https://developer.mozilla.org/en-US/docs/Web/API/Element/prepend
- Fix one *incorrect* `insertBefore` call, in the AnnotationLayer-code.
Initially the patch simply changed that to an `Element.before()`-call, however that broke one of the integration-tests. It turns out that the `index` may try to access a non-existent select-child, which triggers undefined behaviour; note the warning in https://developer.mozilla.org/en-US/docs/Web/API/Node/insertBefore#parameters
This only adds the minimum entries required in order to render the referenced document correctly, rather than trying to support "all" Hebrew glyphs, to ensure that all lines in `getGlyphMapForStandardFonts` are covered by tests.
*Please note:* The dates below are still a little ways off, however that obviously won't affect the existing PDF.js releases. Hence I think that we can make these changes now, since by the time of the *next* official PDF.js release they'll likely match up pretty well.[1]
While we "support" some (by now) fairly old browsers, that essentially means that the library (and viewer) will load and that the basic functionality will work as intended.[2]
However, in older browsers, some functionality may not be available and generally we'll ask users to update to a modern browser when bugs (specific to old browsers) are reported.[3]
Since we've previously settled on only supporting browsers/environments that are approximately *three years old*, this patch updates the minimum supported browsers/environments as follows:
- Chrome 76, which was released on 2019-07-30; see https://en.wikipedia.org/wiki/Google_Chrome_version_history
- Firefox ESR (as before); see https://wiki.mozilla.org/Release_Management/Calendar
- Safari 13, which was released on 2019-09-19; see https://en.wikipedia.org/wiki/Safari_version_history#Safari_13
- Node.js 14, which was release on 2020-04-21 (all older versions have reached EOL); see https://en.wikipedia.org/wiki/Node.js#Releases
---
[1] Given that the releases usually happen every two to three months.
[2] Assuming that a `legacy/`-build is being used, of course.
[3] In general it's never a good idea to use old/outdated browsers, since those may contain *known* security vulnerabilities.
Fixes two recent "Code scanning alerts" on GitHub, which likely happened because these calls originally used `parseInt` instead (during initial development).
After the changes in PR 14998, these operators are now no-ops in the `src/display/canvas.js` code and should no longer be necessary.
Given that `beginAnnotations`/`endAnnotations` are not in the PDF specification, but are rather *custom* PDF.js operators, it seems reasonable to stop using them now that they've become no-ops.
The update in https://bugzilla.mozilla.org/show_bug.cgi?id=1773794 failed, because the `editorNone` icon is identical to a pre-existing one. Given that all of the editor-icons are simply placeholders for now, we can just make a tiny change to the SVG-paths to prevent these kind of problems.
While `TextLayerRenderTask` apparently makes sense in TypeScript environments, given that it's being returned by the `renderTextLayer`-function in the API, we really don't want to extend the *public* API by simply exporting the class directly in `src/pdf.js` since it should never be called/initialized manually.
Hence we follow the same pattern as in PR 14013, and add some very basic unit-tests to ensure that `renderTextLayer` always returns a `TextLayerRenderTask`-instance as expected.
In PR #14717, the type was changed from a HTMLElement to a DocumentFragment.
This broke TypeScript projects that use a HTMLElement container.
To remedy this, we extend the type of container to also include HTMLElement.
- Approximate the drawn curve by a set of Bezier curves in using
js code from https://github.com/soswow/fit-curves.
The code has been slightly modified in order to make the linter
happy.
This only applies to *corrupt* PDF documents, where Annotations are missing the required /Rect-entry. Rendering PopupAnnotations unconditionally shouldn't be a problem, since we're not using a `BaseSVGFactory`-instance in that case.
In the "no fontSize available" code-path, in the `ChoiceWidgetAnnotation._getAppearance` method, we don't provide the necessary second argument when calling the `_getTextWidth`-method which will cause errors to be thrown.
Because of a capitalization error, the MIME type wasn't actually being set correctly. However, please note that downloading of font files still worked correctly which is probably why this has gone unnoticed since 2014.
- each annotation must be rendered independently of the others. So
after having rendered each annotation, the canvas states are reset
in order to have something clean to render the next one.
This method/function was added only for the `gulp image_decoders`-builds, and is completely unused elsewhere (e.g. in the Firefox PDF Viewer).
While this only reduces the size of the *built* `pdf.worker.js` file by a little over 1 kB, it can't hurt to remove completely unused code from the "normal" builds.
While calling `JSON.stringify(...)` on a class-instance obviously "works" (as in it doesn't throw), since it's really just an Object, it doesn't really make much sense in the context of the `AnnotationStorage.hash`-getter.
Also, access the *inverse* Viewport-transform correctly in `FreeTextEditor.serialize` to prevent errors being thrown when that method is invoked.
Finally, slightly updates the `AnnotationStorage.serializable`-getter to improve consistency within the class.
*This fixes a regression from PR 14754.*
We didn't lookup the image-data correctly, with the result that we tried to render some ImageMasks using a string rather than the intended TypedArray. To make matters worse, this code-path was apparently not *properly* covered by existing test-cases.
Given the differences between XFA documents and "normal" PDF documents, we don't support editing of the former ones. Hence, when a XFA-document is opened, we temporarily disable the editor-buttons.
- Ensure that the modified-warning won't be displayed, when navigating away from the viewer, if the user has added custom Annotations and then *removed all* of them.
- Ensure that the *initial* editor-buttons state, i.e. the `toggled`-class, is correctly displayed in the toolbar when then viewer loads.
- Tweak the CSS-classes for the editor-buttons, such that they use the correct focus/hover-rules (similar to the sidebar-buttons).
- Remove a no longer accurate comment from the `BaseViewer.annotationEditorMode`-setter.
- Address a couple of *smaller* outstanding review comments, including some re-formatting changes, from PR 14976.
In PR 14710 we only included the JavaScript-part of the polyfill, however we probably need to include the CSS as well to reduce the risk of problems in older browsers.
With the recent CSS-related improvements in the `preprocess`-function we could probably have included this conditionally in the `viewer.css` file. However, considering that the `<dialog>` polyfill-code is only invoked when actually needed it seemed most appropriate/correct to lazy-load the polyfill-CSS as well.
This patch contains a small optimization specifically for the case when `getDocument` is called with TypedArray-data. In that case we'll still hold onto that data, which could obviously be large, even after the "GetDocRequest"-message has been sent to the worker-thread.
In practice this will most likely not affect memory usage in any noticeable way, since the application calling `getDocument` will probably also be keeping a reference to the TypedArray-data. However, it seems like a good idea to ensure that the PDF.js API *itself* won't unnecessarily keep this data alive.
- it's a regression from PR #14247:
- before the PR, the button was rendered on the canvas whatever its status was;
- after the PR, the button image has been moved in an other canvas so when the button is not renderable
(because it has no actions) then the image is not added the HTML element.
- the buttons in the pdf in bug 1737260 or in the pdf in #14308 were not visible
- make the button always renderable but don't add the link element if it's useless.
- right now we're using the font size from the pdf itself but we use an other font
in the annotation layer. So this size doesn't really make sense and leads to bad
rendering (see pdf in #14928);
- use a sans-serif font for the fields containing text (fix issue #14736);
- remove useless padding in text-based fields (fix issue #14301);
- text fields allow/disallow scrolling bars (see bit 24 in Ff entry), so use this
value to hide/show scrollbars in annotation layer.
- it aims to fix https://bugzilla.mozilla.org/show_bug.cgi?id=1771477;
- hangul contains some syllables which are decomposed when using NFD, hence
the text must be correctly shifted in case it contains some of them.
Currently, when range-requests and/or streaming are not supported or for documents opened from `data`-URLs, we'll manually set the `contentDispositionFilename` (assuming it exists and is valid) from the `onOpenWithData`-callback in `PDFViewerApplication.initPassiveLoading`.
However, because of a small oversight in `PDFViewerApplication._initializeMetadata`, this *cached* `contentDispositionFilename` would be ignored and we'd only attempt to use the one returned by `PDFDocumentProxy.getMetadata` in the API (which in the cases outlined above will always be empty).
Also, to ensure that the document properties dialog always displays the *correct* fileName we'll now lookup it using the same exact method as in the viewer itself (via a new callback-function).
With the exception of `isThumbnailViewVisible`, these getters are completely unused. Generally speaking, using the `visibleView`-getter directly works just as well and seems (at least to me) to be overall preferable considering how our classes are usually implemented.
Currently, when non-standard `pageColors` are specified, the thumbnails will look inconsistent depending on how they're created.
The thumbnails that are created by downsizing the *page* canvases will obviously use the `pageColors` as intended, however the thumbnails which are rendered *directly* will always use the default colors.
Over time, as we've been introducing JavaScript code to modify CSS variables, we've been adding shorthand properties to various classes to reduce unnecessary repetition when accessing the document-styles.
Rather than repeating this in multiple places, it seems overall simpler to just introduce a constant and re-use that throughout the viewer instead.
In the `src/display/canvas.js` code the `d1` operator will be used to set the clipping region, and it obviously cannot be empty since that prevents the Type3-glyph from rendering.
Also, the patch removes an outdated comment; refer to PR 12718.
After the changes in https://bugzilla.mozilla.org/show_bug.cgi?id=1757771, that simplified the MOZCENTRAL downloading code, the `ChromeActions.download`-method will no longer invoke the `sendResponse`-callback.
Hence it should no longer be necessary for the `DownloadManager`, in the MOZCENTRAL viewer, to use `FirefoxCom.requestAsync` since no response is ever provided.[1] For the allocated BlobURLs, they should (hopefully) be released when navigating away from the viewer.
---
[1] Note that that was *already* the case, for one of the previous code-paths in the `ChromeActions.download`-method.
After the changes in https://bugzilla.mozilla.org/show_bug.cgi?id=1757771, that simplified the MOZCENTRAL downloading code, the `sourceEventType` is now completely unused and should thus be removed (in my opinion).
Furthermore, with these changes, we also no longer need a *separate* internal "save"-event and can instead just use the older "download"-event everywhere.
This limits the heuristics for handling of incomplete path operators, see PR 9838, to only apply to *sequences* of such operators. In practice a couple of invalid path operators are (hopefully) unlikely to completely break rendering, whereas a sequence of them will easily lead to fairly chaotic rendering artifacts.
The original `ProgressBar`-functionality is very old, and could thus do with some general clean-up.
In particular, while it currently accepts various options those have never really been used in either the default viewer or in any examples. The sort of "styling" that these options provided are *much better*, not to mention simpler, done directly with CSS rules.
As part of these changes, the "progress" is now updated using CSS variables rather than by directly modifying the `style` of DOM elements. This should hopefully simplify future changes to this code, see e.g. PR 14898.
Finally, this also fixes a couple of other small things in the "mobile viewer" example.
While zeroing the temporary `canvas` makes sense, manually clearing the canvas and its context doesn't really accomplish anything since those are tied to the scope of the method.
We need to wait for UI rendering to start *before* getting the CSS variable values, since otherwise the values will be `NaN`.
This is only an issue if the viewer is completely hidden during loading, e.g. in a `display: none` iframe-element.
When pre-processor blocks are being removed, since they don't apply to the current build target, we may currently end up with consecutive blank lines.
While this is obviously not a big issue, it's nonetheless undesirable and we can adjust the `writeLine` function to prevent that.
*This is a follow-up to PR 14886, which "broke" this.*
In addition to fixing the issue, using an Array and `join`-ing it at the end may also be a tiny bit more efficient than using a growing string.
*This patch can be tested, in the viewer, using the `annotation-fileattachment.pdf` document from the test-suite.*
It seems that the code to delay dispatching of the "attachmentsloaded"-event, when `attachmentsCount === 0`, is now effectively broken.[1]
Rather than only using an arbitrary timeout, the new code will instead wait for an "annotationlayerrendered"-event and only *fallback* to using a timeout.
---
[1] The timing of the annotationLayer-rendering changed slightly in PR 14247, and the old code in `PDFAttachmentViewer` wasn't good enough to handle that.
If the computed background/foreground colors are identical, the `canvas` would be rendered mostly blank with only images visible. Hence it seems reasonable to also ignore the `pageColors`-option in this case.
Also, the patch tries to *briefly* outline the various cases in which we ignore the `pageColors`-option in a comment.
Given that the new API-option is an Object named `pageColors`, with `background`/`foreground` keys, it occurred to me that it'd be slightly more consistent if the options/preferences names fully reflected that.
An old shortcoming of the `preprocessCSS`-function is its complete lack of support for our "normal" defines, which makes it very difficult to have build-specific CSS rules. Recently we've started using specially crafted comments to remove CSS rules from the MOZCENTRAL build, but (ab)using the `preprocessCSS`-function in this way really doesn't feel great.
However, it turns out to be surprisingly simple to instead use the "regular" `preprocess`-function for the CSS files as well. The only special-handling that's still necessary is the helper-function for dealing with CSS-imports, but apart from that everything seems to just work.
One reason, as far as I can tell, for having a separate `preprocessCSS`-function was likely that we originally used *lots* of vendor-prefixed CSS rules in our CSS files. With improvements over the years, especially thanks to Autoprefixer and PostCSS, we've been able to remove *almost* all non-standard CSS rules and the need for special-casing the CSS parsing has mostly vanished.
*Please note:* As part of testing this patch I've diffed the output of `gulp generic`, `gulp mozcentral`, and `gulp chromium` against the `master`-branch to check that there was no obvious breakage.
These methods are completely unused in the Firefox PDF Viewer, and were only added to supplement the `PDFViewerApplication.{bindEvents, bindWindowEvents}`-methods since third-party implementations may apparently need to remove event listeners (see PR 8525).
However, in the MOZCENTRAL build that's just dead code and this patch reduces the size of the *built* `web/viewer.js`-file by `~3.5` kB.
*This patch can be tested, in the viewer, using the `annotation-fileattachment.pdf` document from the test-suite.*
Note how the `FileSpec`-implementation already uses `stringToPDFString` during the filename lookup, see cfac6fa511/src/core/file_spec.js (L70)
Hence there's no reason to repeat that again in the `FileAttachmentAnnotationElement`-constructor, and we can thus simplify the "fileattachmentannotation"-event handling a little bit.
This new CSS variable will allow us to simplify a couple of different viewer components, since we no longer need to use JavaScript-based hacks and can directly set the CSS rules instead. In particular:
- The `BaseViewer`-handling, used as part of the code that will center pages *vertically* in PresentationMode, can be simplified.
By using CSS to control the height of the "dummy"-page we avoid unnecessarily invalidating the scale-value, which can reduce *some* unneeded re-rendering while PresentationMode is active.
- The `SecondaryToolbar.#setMaxHeight`-method, and its associated parameters, are no longer necessary and can be completely removed.
Note that in order for things to work correctly in general, the new `--viewer-container-height` CSS variable must potentially be updated on any window-based "resize"-event (even when there's no zooming). While this is currently only done in the default viewer, that shouldn't be an issue since neither PresentationMode nor Toolbar-functionality is included in the "viewer components".
- Use Canvas & CanvasText color when they don't have their default value
as background and foreground colors.
- The colors used to draw (stroke/fill) in a pdf are replaced by the bg/fg
ones according to their luminance.
After recent changes, adding *basic* Spread mode support in PresentationMode has now become reasonably straightforward.
However, documents with *varying* page sizes are non-trivial to handle and would require re-writing (or at least re-factoring) a bunch of the zooming-code.
Hence this PR *purposely* only allow Spread modes to be used, in PresentationMode, for documents where all pages have exactly the same size. While this obviously isn't a fully complete solution, it will however cover the vast majority of all documents and should hopefully be deemed good enough for now.
In PR 14112 usage of this *internal* method was reduced, and it thus can't hurt to clean-up things a little bit more.
Note in particular that we can simplify the call-sites by directly passing in the already available `PDFPageView`-instance, since the `id`-property those contain can replace the previous `pageNumber`-parameter[1].
Given that the method name has always been prefixed with an underscore it was thus never intended to be "public", hence we can now enforce that with modern ECMAScript features.
---
[1] There's already a bunch of other spots, throughout the viewer-code, where we assume that the `PDFPageView.id`-property contains proper page *numbers* (and not e.g. indices); note how we initialize the `PDFPageView`-instances in the `BaseViewer.setDocument`-method.
The current `lastModified`-getter, which only contains a time-stamp, is a fairly crude way of detecting if the stored data has actually been changed. In particular, when the `getRawValue`-method is used, the `lastModified`-getter doesn't cope with data being modified from the "outside".
To fix these issues[1], and to prevent any future bugs in this code, this patch introduces a new `AnnotationStorage.hash`-getter which computes a hash of the currently stored data. To simplify things this re-uses the existing `MurmurHash3_64`-implementation, which required moving that file into the `src/shared/`-folder, since its performance should be good enough here.
---
[1] Given how the `AnnotationStorage.lastModified`-getter was used, this would have been limited to *printing* of forms.
- since resetForm function reset a field value a calculateNow is consequently triggered.
But the calculate callback can itself call resetForm, hence an infinite recursive loop.
So basically, prevent calculeNow to be triggered by itself.
- in Firefox, the letters entered in some fields were duplicated: "AaBb" instead of "AB".
It was mainly because beforeInput was triggering a Keystroke which was itself triggering
an input value update and then the input event was triggered.
So in order to avoid that, beforeInput calls preventDefault and then it's up to the JS to
handle the event.
- fields have a property valueAsString which returns the value as a string. In the
implementation it was wrongly used to store the formatted value of a field (2€ when the user
entered 2). So this patch implements correctly valueAsString.
- non-rendered fields can be updated in using JS but when they're, they must take some properties
in the annotationStorage. It was implemented for field values, but it wasn't for
display, colors, ...
- it fixes#14862 and #14705.
Given that we're (ab)using spread-modes in order to ensure that pages are centered *vertically* in PresentationMode, this re-factoring simplifies the code slightly.
Furthermore, in the event that we *possibly* want to try and support spread-modes in PresentationMode[1] this re-factoring will also prevent future duplication.
---
[1] Note that I'm not particularly keen on doing that, since documents with varying page sizes will be annoying to handle.
Given that the `isNodeJS`-constant will, after PR 14858, *always* be `false` in non-GENERIC builds we can simplify a couple of `getDocument`-parameter default values slightly.
The old format, with inline `PDFJSDev`-checks, wasn't exactly a wonder of readability; which was my fault.
This first of all simplifies the file, since we no longer need dummy-classes and can instead *directly* define the actual classes.
Furthermore, and more importantly, this means that we no longer need to bundle this code in e.g. MOZCENTRAL-builds which reduces the size of *built* `pdf.js` file slightly.
Given that the `compileType3Glyph` function *returns* a function, see `drawOutline`, we'll thus keep the surrounding scope alive. Hence it shouldn't hurt to *explicitly* mark the temporary `Uint8Array`s, used during parsing, as no longer needed. Given the current `MAX_SIZE_TO_COMPILE`-value these `Uint8Array`s may be approximately two mega-bytes large *for every* Type3-glyph.
Interestingly enough this appears to be the very first case of *encoded* dest-strings, in /GoTo destination dictionaries, that we've actually come across. What's really fascinating is that it's less than a week after issue 14847, given that these issues are *somewhat* similar.
Unfortunately newer versions either caused breakage when running the
unit tests manually in a browser or when serving the development viewer.
Given that we hope to use native import maps soon and this dependency
will then be removed anyway, let's pin it for the time being.
This moves the `COMPILE_TYPE3_GLYPHS`/`MAX_SIZE_TO_COMPILE`-checks into the `compileType3Glyph` function itself, which allows for some simplification at the call-site.
These changes also mean that the `COMPILE_TYPE3_GLYPHS`-check is now done *once* per Type3-glyph, rather than everytime that the glyph is being rendered.
There are two notable changes here:
- `dommatrix` is a major version upgrade, but looking through the commit
history of their `package.json` file at https://github.com/thednp/dommatrix/commits/master/package.json
(due to the lack of a changelog) I couldn't find any breaking changes.
- `es-module-shims` is a regular update, but it was previously pinned
for causing intermittent breakage when running the unit tests in a
browser manually. Fortunately this cannot be reproduced anymore with
the most recent version, so we can also put the caret back now.
There's no point in having these variables defined (implicitly) as `undefined` in e.g. the Firefox PDF Viewer.
By defining them with `var` and using ESList ignores, rather than `let`, we can move them into the relevant pre-processor block instead. Note that since the entire viewer-code is placed, by Webpack, in a top-level closure these variables will thus *not* become globally accessible.
Given that `webViewerOpenFileViaURL` only has a single call-site, and also isn't a particularly large/complex function, it doesn't seem necessary for this to be a separate function and hence it's simply inlined instead.
Also, changes the "no valid build-target was set"-case to throw unconditionally since the only way that it could ever be hit is if there are bugs in the `gulpfile`-code.
- it aims to fix https://bugzilla.mozilla.org/show_bug.cgi?id=1264608;
- it's only a partial fix for #3351;
- some tiled images have some spurious white lines between the tiles.
When the current transform is applyed the corners of an image can have
some non-integer coordinates leading to some extra transparency added
to handle that. So with this patch the current transform is applied on the
point and on the dimensions in order to have at the end only integer values.
As it turns out, most of the code-paths in the `PDFImage`-class won't actually pass the TypedArray (containing the image-data) to the `ColorSpace`-code. Hence we *generally* don't need to force the image-data to be a `Uint8ClampedArray`, and can just as well directly use a `Uint8Array` instead.
In the following cases we're returning the data without any `ColorSpace`-parsing, and the exact TypedArray used shouldn't matter:
- b72a448327/src/core/image.js (L714)
- b72a448327/src/core/image.js (L751)
In the following cases the image-data is only used *internally*, and again the exact TypedArray used shouldn't matter:
- b72a448327/src/core/image.js (L762) with the actual image-data being defined (as `Uint8ClampedArray`) further below
- b72a448327/src/core/image.js (L837)
*Please note:* This is tagged `api-minor` because it's API-observable, given that *some* image/mask-data will now be returned as `Uint8Array` rather than using `Uint8ClampedArray` unconditionally. However, that seems like a small price to pay to (slightly) reduce memory usage during image-conversion.
Currently we're using *both* ids and classes when specifying the button icons, which seems completely unnecessary and only serves to bloat the CSS code.
In general you always need to be careful about CSS specificity, but in these cases that should not actually be a problem.
Also, while not a button, this patch makes a similar simplification for the `pageNumber`-input.
Note how in the `viewer.html` file we're specifying class names for most buttons, despite that not really being necessary. First of all, in many cases those class names are *identical* to the element-ids. Secondly, looking through the CSS rules they are only ever used when specifying button icons.
All-in-all, we should be able to simplify the HTML-code and use the element-ids in the CSS rules instead. Obviously ids have a higher CSS specificity than classes, but given how the old classes were being used that shouldn't be a problem here.
Also, the patch generalizes the styling for buttons (e.g. `viewBookmark`) that are *actually* link-elements.
Finally, while slightly unrelated, this patch also removes a little bit of duplication when specifying the border for `toolbarField`s.
Initially I considered updating the `NameOrNumberTree`-implementation to handle encoded keys, however that quickly became somewhat messy (especially in the `NameOrNumberTree.get`-method) since only NameTrees using string-keys.
Hence the easiest solution, as far as I'm concerned, was thus to just update the `Catalog.destinations`-getter instead. Please note that in the referenced PDF document the `Catalog.destination`-method will thus fallback to fetch all destinations, which should be fine since this is the very first case of encoded keys that we've seen.
Also changes the `NameOrNumberTree.getAll`-method to prevent a possible run-time error, although we've so far not seen such a case, for any non-Array Kids-entries found in a NameTree/NumberTree.
Finally, to improve overall consistency and to hopefully prevent future bugs, the patch also updates a couple of other `NameTree` call-sites to correctly handle encoded keys. (Note that the `Catalog.attachments`-getter was already doing this.)
Searching through all of the files in the `web/`-folder has no *other* hits for the "textButton" string. Hence it's clear that there are no DOM elements actually using this class, and it's thus dead code.
This re-factors the various toolbar separators to *explicitly* specify both their dimensions and margins. Also, for the `horizontalToolbarSeparator`-class we can just set the `background-color` rather than using `border-top`.
Note that the `splitToolbarButtonSeparator`-class currently sets a number of unnecessary CSS rules, since as mentioned by the Firefox Devtools both the `display`- and `z-index`-properties are being ignored because `float` is used.
Finally, there's also no need to set a `z-index` for some of the `:hover`-rules. It's possible that this *was* necessary before the re-design, since the buttons had borders then.
Note how both of the openFile-buttons are always hidden during viewer initialization in the MOZCENTRAL build, i.e. the *built-in* Firefox PDF Viewer. Despite that we still include HTML, CSS, and JavaScript code for these buttons in the build.
This patch *reduces* the size of the `gulp mozcentral` output by `1679` bytes, which isn't a lot but still cannot hurt.
While working on PR 14825, I couldn't help noticing that the code to increment the `count` for cached ImageMasks was repeated multiple times. Hence it makes sense, as far as I'm concerned, to move this into a helper function instead.
These rules became unnecessary with PR 7697, over five years ago, since printing is now done from a `printContainer`-element rather than "directly" using the viewer.
Note how the *entire* `outerContainer`, which contains all of the DOM elements that were being manually hidden, is now being hidden during printing. Furthermore, note also how the print-canvases/images and their containers are using custom CSS-classes[1] rather than re-using the ones from the viewer.
---
[1] See the `printedPage` respectively `xfaPrintedPage` classes.
Rather than modifying the leading/trailing `margin` on the actual toolbar buttons, to achieve appropriate spacing at the left/right edges of the toolbar(s), it seems much more appropriate (and simpler) to just specify an explicit `padding` for the relevant toolbar containers.
Also, for toolbar buttons placed in `splitToolbarButton`-classes we can reduce some complexity around setting the `margin` (since it should always be zero now).
With these changes, we're thus able to get rid of a couple more `!important`-rules.
Currently we only insert optionalContent-data into the operatorList the first time that an image is parsed, which will (in hindsight) obviously cause problems for cached images.
Hence we also need to insert the optionalContent-data in the various worker-thread image caches, such that it can be accessed in the fast-paths that are used to skip re-parsing of images.
In order to reduce the amount of repeated code, this patch also adds a new `OperatorList`-method that takes care of inserting the necessary data in the operatorList.
In the referenced PDF document the fonts have /Encoding-entries that are Streams (containing completely bogus data), which are thus obviously not valid here.
Hence, only when `ignoreErrors` is set, we'll now ignore these corrupt /Encoding-entries and fallback to the existing code to try and infer a usable encoding.
Given that this is *clearly* a case of corrupt PDF documents, there's no guarantee that this will "fix" all such cases, however it's the best that we do here and shouldn't really be worse than ignoring an entire font.
- Remove a redundant `margin-top` rule for the `.dropdownToolbarButton`. After the (somewhat) recent UI-refresh all buttons now use `margin: 2px 1px;`, which renders the override unnecessary (and getting rid of an `!important`-rule can't hurt).
- Combine two `.toggled::before` rules, since they're identical.
Note that both the `errorWrapper` HTML and JavaScript code is being ignored in the MOZCENTRAL build, i.e. the *built-in* Firefox PDF Viewer, however the CSS rules are still being included.
That seems totally unnecessary, and while we currently don't have full build-target support in the CSS pre-processor we can actually improve things quite easily anyway. By (ab)using the existing CSS pre-processor, which will remove any non-Firefox CSS rules for the MOZCENTRAL build, it's possible to easily stop bundling any CSS rules by using comments that include a `-webkit`-string.
*Please note:* To easily test that this doesn't break the `errorWrapper` in GENERIC builds, try running e.g. `PDFViewerApplication._otherError("test");` in the web-console.
This special-case was added because the original PresentationMode-implementation used some CSS-tricks to hide everything except the current page. With the changes in PR 14112, which added a PAGE scroll-mode, many of the old PresentationMode-specific hacks could thus be removed from both the JS and CSS code.
This patch is yet another (small) clean-up step, to reduce the number of PresentationMode special-cases used throughout the `BaseViewer`. Note in particular that `BaseViewer.update` now works just fine in PresentationMode[1], and that we only need to ensure that the active page won't *accidentally* change because of the PresentationMode-specific zooming that occurs during page-switching.
---
[1] In the event that we ever want to try and support spread-modes in PresentationMode, which I'm really not keen on doing since documents with varying page sizes will be annoying to handle, these changes would be necessary as well.
The styling of the previous/next-buttons and the findInput, with the elements being "glued" together, was supposed to mimic the styling used in the Firefox *browser* findbar. However, after the most recent re-styling of the Firefox browser UI these elements are now visually separated.
Hence it makes sense, as far as I'm concerned, to try and follow this styling for the findbar used in the GENERIC viewer. One benefit of doing this is that we get more consistent styling, since the buttons now look/behave identically in both the main toolbar and the findbar. Additionally this also simplifies the CSS a bit, since a lot of the existing findbar-specific rules can be removed.
- most of the time the current transform is a scaling one (modulo translation),
hence it's possible to avoid to apply the transform on each bbox and then apply
it a posteriori;
- compute the bbox when it's possible in the worker.
The spread-mode code in `BaseViewer.#ensurePageViewVisible`-method was initially copied from the `BaseViewer._updateSpreadMode`-method, which means that it's slightly more complicated than actually necessary.
In particular, in the PAGE scroll-mode there can only be *one* spread active at a time and we thus don't need to handle insertion of multiple spread-divs.
Given that we're no longer, after PR 14560, bundling the `web-streams-polyfill`-code in the `legacy`-builds we shouldn't need to exclude it from Babel now.
- it's the second part of the fix for https://bugzilla.mozilla.org/show_bug.cgi?id=857031;
- some image masks can be used several times but at different positions;
- an image need to be pre-process before to be rendered:
* rescale it;
* use the fill color/pattern.
- the two operations above are time consuming so we can cache the generated canvas;
- the cache key is based on the current transform matrix (without the translation part)
and the current fill color when it isn't a pattern.
- the rendering of the pdf in the above bug is really faster than without this patch.
Given that no HTML element has used the `loadingBox`-id for many years, we obviously don't need to try and hide a non-existent element during printing.
Furthermore, we also shouldn't need to change the `overflow`-value for the `viewerContainer`-element during printing. Originally, many years ago now, we printed "directly" using the viewer and then this apparently made sense.
Given that none of these CSS rules are used at all, unless debugging is enabled, it seems completely unnecessary to load them *unconditionally* for all users.[1]
Note that if *both* the `textLayer` and `pdfBug` debugging hash-parameters are specified simultaneously, we'll now load the `PDFBug`-file *twice* (since the code is simpler that way). However, given first of all that none of this is enabled by default and secondly that using those parameters together isn't helpful[2], potentially loading that file twice is hopefully not an issue.
For the `gulp mozcentral` target, the size of the *built* `viewer.css` file is reduced `> 3%` with this patch.
---
[1] For the Firefox built-in PDF Viewer, in order to even be able to access the `PDFBug` functionality, you need to first of all set `pdfjs.pdfBugEnabled = true` manually in `about:config`. Secondly, you then also need to append the `pdfBug=...` hash-parameter to the URL when *initially* loading the document.
[2] Note how the `textLayer`-settings are already, since essentially forever, overriding the highlighting-features of the "FontInspector"-tab.
- avoid to call normalizeRect which clones the rectangles: it's useless
and time consuming;
- in profiling the pdf in bug 1135277, the time spent in intersect drops
from ~1s to ~30ms.
Because of a bug in previous `core-js` versions, which caused an Error to be thrown if its `structuredClone` polyfill was called with an *explicit* `null`/`undefined` transfer-parameter, the `LoopbackPort`-class contained a work-around.
In the latest `core-js` version this has been fixed, and we can thus simplify our code ever so slightly; please see https://github.com/zloirock/core-js/releases/tag/v3.22.0
This CSS variable is only used together with the `annotationCanvasMap`-functionality in the canvas-code, however its value can be *trivially* computed by using the older `--zoom-factor` CSS variable together with the `PixelsPerInch`-structure.
Rather than having *two different* CSS variables that are this closely linked, it seems better to simplify things by using just one CSS variable instead.
According to the CSS, there should be a visible "divider" after the "Page Width" zoom-option. However, this is being ignored in both Mozilla Firefox[1] and Google Chrome hence the rule is effectively useless now.
Furthermore, the "custom" zoom-option is already being `hidden` using the attribute (in the HTML code) and there should thus be no reason to duplicate this in the CSS as well.
---
[1] Support for *detailed* styling of `<select>`-elements was removed as part of the E10s project.
With just a couple of exceptions, for the `thumbnailView`, all of the sidebarViews share the same basic styling which thus allows for some simplification.
- For the findbar/secondaryToolbar case, the `min-width` rule doesn't really make sense since it's way too small to be useful. Furthermore, the findbar is already specifying its own `min-width` and the secondaryToolbar will (thanks to its buttons) receive a correct/useful width.
- The pageNumber-input already has an *explicit* `width` set, hence setting the `min-width` rule as well is completely unnecessary.
- The treeItem-links are supposed to *compute* their `min-width`, and the static value was only added as a fallback for older browsers without `calc()` support.
In a couple of spots, mostly related to the debugging tools, we're unnecessarily using a somewhat "complex" `background`-format only to specify a solid color. This can be simplified by using `background-color` instead, and the patch also removes a `color`-rule that's being ignored anyway.
After the changes in https://bugzilla.mozilla.org/show_bug.cgi?id=1761839, we no longer need this CSS work-around to prevent the entire `<dialog>` contents from becoming selected when the backdrop is clicked.
When the viewer becomes narrow enough that the sidebar is overlaying the document, which means that the `viewerContainer` is not moving when opening/closing the sidebar, we're currently not removing the `sidebarMoving` CSS class as intended.
While this doesn't cause any *visible* issues, it's nonetheless wrong and should be fixed.
With the changes in PR 8993, a number of the `@media`-related CSS rules became unnecessary. However, it appears that some of these rule were *accidentally* left behind despite being unused now.
Note that previously, when opening the sidebar shifted the position of the main toolbar, we had to take both the sidebar opened *and* closed cases into account in these `@media` rules.
- write some uint32 instead of uint8 to avoid the check before clamping;
- unroll the loop to write data in the buffer
- but keep a loop for the last element of a line: it likely doesn't hurt
that much since it's executed only for one time for each line;
- I tested on a macbook with an Apple chip, and on Firefox nightly the new
code is almost 3.5x faster than before (~1.8x with Chrome).
*This is yet another installment in a never-ending series of patches that attempt to simplify and improve old code.*
The `fileInput`-element is used to support the "Open file"-button in the `GENERIC` viewer, however this is very old code.
Rather than creating the element dynamically in JavaScript, we can simply define it conditionally in the HTML code thanks to the pre-processor. Furthermore, the `fileInput`-element currently has a number of unnecessary CSS rules, since the element is *purposely* never made visibly.
Note that with these changes, the `fileInput`-element will now *always* have `display: none;` set. This shouldn't matter, since we can still trigger the `click`-event of the element just fine (via JavaScript) and this patch has been successfully tested in both Mozilla Firefox and Google Chrome.
- it aims to partially fix performance issue reported: https://bugzilla.mozilla.org/show_bug.cgi?id=857031;
- the idea is too avoid to use byte arrays but use ImageBitmap which are a way faster to draw:
* an ImageBitmap is Transferable which means that it can be built in the worker instead of in the main thread:
- this is achieved in using an OffscreenCanvas when it's available, there is a bug to enable them
for pdf.js: https://bugzilla.mozilla.org/show_bug.cgi?id=1763330;
- or in using createImageBitmap: in Firefox a task is sent to the main thread to build the bitmap so
it's slightly slower than using an OffscreenCanvas.
* it's transfered from the worker to the main thread by "reference";
* the byte buffers used to create the image data have a very short lifetime and ergo the memory used is globally
less than before.
- Use the localImageCache for the mask;
- Fix the pdf issue4436r.pdf: it was expected to have a binary stream for the image;
- Move the singlePixel trick from operator_list to image: this way we can use this trick even if it isn't in a set
as defined in operator_list.
With the changes in the previous patch, we can simplify the state-tracking by using the `PresentationModeState`-values directly in the `PDFPresentationMode` class.
We don't need to first check if the Dictionary contains the key, since trying to get a non-existent key simply returns `undefined` and we're already ensuring that the value is a boolean.
Furthermore, we shouldn't need to worry about the `Object.prototype` containing enumerable properties since the checks (in `src/core/worker.js`) done for `Array.prototype` *indirectly* also cover `Object`s. (Keep in mind that an `Array` is just a special kind of `Object` in JavaScript.)
The `pdfOpenParams` parameter has never really made sense in PresentationMode, since e.g. the zoom-value doesn't (generally) agree with the value chosen by the user prior to entering PresentationMode.
This has never mattered all that much, since the `viewBookmark`-button isn't visible in PresentationMode (nor is any other toolbar button for that matter). However, in the `PDFHistory`-implementation we're currently forced to handle this case specifically since we don't want to populate the browser history with nonsensical state.
Hence it makes overall sense, as far as I'm concerned, to tweak the "updateviewarea" event to include a *simplified* `pdfOpenParams` parameter when PresentationMode is active. Given that the `viewer components` don't include PresentationMode functionality, this change thus shouldn't matter for third-party users.
This method was originally added specifically to work-around bugs/issues related to PresentationMode in Google Chrome. Note that prior to PR 14112 we were using some CSS hacks to only show the current page in PresentationMode, and that could lead to the `getVisibleElements` function not always finding the correct elements.
However, after the changes in PR 14112 we're now using the Page-scrolling mode in PresentationMode and consequently there'll only be *a single* page visible at a time. Hence then `BaseViewer._getCurrentVisiblePage` helper method should no longer be needed, and when testing (locally) in Google Chrome everything seems to work correctly now.
The various functionality in `web/debugger.js` is currently *indirectly* added to the global scope, since that's how `var` works when it's used outside of any functions/closures.
Given how this functionality is being accessed/used, not just in the viewer but also in the API and textLayer, simply converting the entire file to a module isn't trivial[1]. However, we can at least export the `PDFBug`-part properly and then `import` that dynamically in the viewer.
Also, to improve the code a little bit, we're now *explicitly* exporting the necessary functionality globally.
According to MDN, see https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Statements/import#browser_compatibility, all the browsers that we now support have dynamic `imports` implementations.
---
[1] We could probably pass around references to the necessary functionality, but given how this is being used I'm just not sure it's worth the effort. Also, adding *official* support for these viewer-specific debugging tools in the API feels both unnecessary and unfortunate.
*Please note:* This is possibly bad/wrong in general, but I figured that submitting it for review wouldn't hurt.
It seems that even Adobe Reader doesn't handle the non-ASCII characters that appear in some of the fields correctly, however it should be pretty easy to improve things on the PDF.js side.
- it aims to fix#14685;
- add a basic object to get values from the parsed datasets;
- these annotations don't have an appearance so we must create one when printing or saving.
Please note that this patch is purposely quite basic, e.g. it doesn't add the polyfill-CSS in order to simplify the build process, and things such as `::backdrop` thus isn't working.
However, this patch does ensure that older browsers can at least still *access* all of the previous overlays and that things like e.g. opening of password-protected documents respectively printing still works.
This will hopefully improve the a11y a little bit in these dialogs, however there's most definately more things that can be done here (by someone more knowledgeable about a11y).
This way we're able to store the `<dialog>` elements directly, which removes the need to use manually specified name-strings thus simplifying both the `OverlayManager` itself and its calling code.
This replaces our *custom* overlays with standard `<dialog>` DOM elements, see https://developer.mozilla.org/en-US/docs/Web/HTML/Element/dialog, thus simplifying the related CSS, HTML, and JavaScript code.
With these changes, some of the functionality of the `OverlayManager` class is now handled natively (e.g. `Esc` to close the dialog). However, since we still need to be able to prevent dialogs from overlaying one another, it still makes sense to keep this functionality (as far as I'm concerned).
PDFScriptingManager uses the `mousedown` and `mouseup` listeners to keep
track of whether the mouse pointer is pressed in the `isDown` flag.
These listeners were registered to run during the bubbling phase of the
event dispatch, which can be interrupted if any of the previous event
listeners stopped the event propagation. An example of that is by
`GrabToPan` in web/grab_to_pan.js.
Since the mousedown (and mouseup) listeners of PDFScriptingManager are
free of side effects, and the intention is to always run them, it makes
most sense to register them with the capture flag.
- it aims to fix issue #14627;
- the basic idea of the recent text refactoring was to only consider the rendered visible whitespaces.
But sometimes, the heuristics aren't correct and although some whitespaces are in the text stream
they weren't in the text chunks because they were too small. Hence we added some exceptions, for example,
we always add a whitespace when it is between two non-whitespace chars but only when in the same Tj.
So basically, this patch removes the constraint to have the chars in the same Tj
(in using a circular buffer to save the two last chars) but don't add a space when the visible space is really
too small (hence `NOT_A_SPACE_FACTOR`).
Given that the textLayer-code has been using a `DocumentFragment` ever since PR 3356 (back in 2013), simply updating the type of the `container` property should be fine.
This patch also tries to, ever so slightly, improve the grammar of a couple of other properties in the typedef.
After the recent round of patches, I figured that we'd gone as far as possible in replacing `dir`-dependent CSS rules for the viewer.
However, it occurred that me that we could actually use a bit of CSS-trickery to get rid of the remaining ones. More specifically, this was done by defining a CSS variable whose value depends on the document direction and then using that variable together with `calc()` in the affected rules.
*Please note:* I suppose that this could perhaps be seen as a bit too "magical", hence I understand if this patch is ultimately rejected, however this is probably the only simple way to get rid of the remaining `dir`-dependent CSS rules.
The old code would allow an overlay to force close *itself*, before immediately re-opening itself, which actually isn't very helpful in practice since that won't re-run any overlay-specific initialization code.
Given how the overlays are being used this really shouldn't have caused any issues, but it's a bug that we should fix nonetheless.
There's a couple of `getDocument` parameters that should be numbers, but which are currently not *fully* validated to prevent issues elsewhere in the code-base.
Also, improves validation of the `ownerDocument` parameter since we currently accept more-or-less anything here.
Note that the Prettier update made it possible to move a couple of comments after `default:`-cases back to their original/intended positions, please see https://prettier.io/blog/2022/03/16/2.6.0.html
*Please note:* This is another step in what will, time permitting, become a series of patches to simplify/modernize the viewer CSS.
Rather than having to manually specify ltr/rtl-specific float-values in the CSS, we can use logical `inline-start`/`inline-end` instead (and similar for some related left/right occurrences).
These logical properties depend on, among other things, the direction of the HTML document which we *always* specify in the viewer.
Given that most of these logical CSS properties are fairly new, and that cross-browser support is thus somewhat limited (see below), we rely on PostCSS plugins in order to support this in the GENERIC viewer.
- https://developer.mozilla.org/en-US/docs/Web/CSS/float#browser_compatibility
- https://developer.mozilla.org/en-US/docs/Web/CSS/inset-inline-end#browser_compatibility
Currently we're resolving the Promises in the `_extractTextPromises` Array with the page-index, despite that not really being necessary since the Promises in the Array are explicitly inserted in the correct order.
Furthermore, we can replace the standard `for`-loop with a `for...of`-loop which results in ever so slightly more compact code.
This patch removes the existing `forEach` methods, in favor of making the classes properly iterable instead. Given that the classes are using a `Set` respectively a `Map` internally, implementing this is very easy/efficient and allows us to simplify some existing code.
Rather than *manually* specifying a "mode", we can simply use the regular `defines` directly instead. To improve consistency, in the `external/builder/builder.js` file, a couple of parameters are also re-named.
Given that we now only use Workers when `postMessage` transfers are supported, there's really no point in trying to send a "test" message *without* transfers present.
Hence, if `postMessage` transfers are not supported by the browser, we'll now fallback to "fake" Workers immediately instead. The comment about Opera is also removed, since it was originally added back in PR 983 and mentions Opera `11.60` [which was released in 2011](https://en.wikipedia.org/wiki/History_of_the_Opera_web_browser#Version_11).
Given that none of these methods were ever intended to be accessed directly from the outside, we can use modern ECMAScript features to ensure that they are indeed private.
This patch also makes `fieldData` private, to remove the old hack used to prevent it from being modified from the outside.
Given that we're now *building* the `web/viewer.css` file used in the development viewer, i.e. with `gulp server`, we no longer need to hard-code these `-webkit`-prefixed rules and can instead let Autoprefixer handle that for us.
To allow using modern CSS features that currently only Mozilla Firefox supports[1], while still enabling development/testing in recent Google Chrome versions, we'll have to start building the `web/viewer.css` file with `gulp server` as well.
In my testing, building the development CSS (and copying the images) takes *less than* `200 ms` on average which is hopefully an acceptable overhead for this sort of feature.
---
[1] In particular `float`, with `inline-start`/`inline-end` values.
Every single call-site has always passed in `true` for this parameter, ever since the function was first added back in PR 8023. Hence the parameter appears to be completely unnecessary, which is why it's removed and the function is updated to *unconditionally* strip out any license headers (in the middle of the file).
Apparently this unit-test works in Node.js now, hence it's *possible* that the reason it didn't work previously is that there were bugs in our old `structuredClone` polyfill.
This patch fixes an old inconsistency, when using `BasePreferences.{reset, set}`, where the internal Preference values would be kept even if writing them to storage failed.
Given that none of these fields were ever intended to be accessed directly from the *outside*, since that will lead to inconsistent/broken state, we can use modern ECMAScript features to ensure that they are indeed private.
These changes make sense for two reasons:
- Given that the parameters are potentially passed to the worker-thread, depending on the `useWorkerFetch` parameter, we need to prevent errors if the user provides values that aren't clonable.
- By ensuring that the default values are indeed `null`, we'll trigger main-thread fetching (of CMaps and Standard fonts) as intended in the `PartialEvaluator` and thus potentially provide better Error messages.
This function is currently placed in the `src/shared/util.js` file, which means that the code is duplicated in both of the *built* `pdf.js` and `pdf.worker.js` files. Furthermore, it only has a single call-site which is also specific to the `GENERIC`-build of the PDF.js library.
Hence this helper function is instead moved into the `src/display/api.js` file, in such a way that it's conditionally defined but still can be unit-tested.
Besides converting the `send` function to use the Fetch API, this patch also changes the method to return a `Promise` to get rid of the callback function. (Although, currently there's no call-site passing in a callback function.)
This is the final part in a series of patches that try to re-implement PR 14287 in smaller steps.
Besides converting `inlineImages` to use the Fetch API, this patch also combines the `inlineImages` and `resolveImages` functions since they are always used together.
Originally the code in the `src/`-folder was shared between the main/worker-threads, and back then it probably made sense that the `PDFDocument` constructor accepted different arguments.
However, for many years we've not been passing anything *except* Streams to `PDFDocument` and we should thus be able to slightly simplify that code. Note that for e.g. unit-tests of this code, using either a `NullStream` or a `StringStream` works just fine.
According to the MDN compatibility data, see https://developer.mozilla.org/en-US/docs/Web/API/DOMMatrix/DOMMatrix#browser_compatibility, all browsers that we support have native `DOMMatrix` implementations (since quite some time too).
Hence Node.js is the only environment that lack `DOMMatrix` support, which probably isn't that surprising given that it's browser functionality.
While the `DOMMatrix` polyfill isn't that large, it nonetheless seems completely unnecessary to bundle it in the `legacy` builds when it's not needed in browsers. However, we can avoid that by simply listing `dommatrix` as a dependency for the `pdfjs-dist` library.
This is another part in a series of patches that try to re-implement PR 14287 in smaller steps.
Besides converting `Driver._send` to use the Fetch API, this also changes the method to return a `Promise` to get rid of the callback function.
Please note that I *purposely* try to maintain the existing behaviour of re-sending the data on failure/unexpected response, including how/where the old callback function was invoked.
*Please note:* I don't really know anything about a11y, hence it's possible that this patch either doesn't work correctly or at least isn't a complete solution.
In both the SecondaryToolbar and the Sidebar we have "button groups" that functionally acts essentially like radio-buttons. Based on skimming through [this MDN article](https://developer.mozilla.org/en-US/docs/Web/Accessibility/ARIA/Roles/radio_role) it thus appears that we should tag them as such, using `role="radiogroup"` and `role="radio"`, and then utilize the `aria-checked` attribute to indicate to a11y software which button is currently active.
When there are *multiple* empty glyphs at the start of the data, ensure that the "first" glyph gets a correct `endOffset` to avoid skipping it during parsing in the `sanitizeGlyph` function.
The situation described in issue 14626 seems like a fairly special case, and it thus seem reasonable that we simply follow the same pattern as elsewhere in the `PartialEvaluator` when the `stopAtErrors` API-option is being used.
Unfortunately simply using `appearance: textfield;`, or even `-webkit-appearance: textfield;`, doesn't actually hide the "spinner" in number-input elements in e.g. Google Chrome.
Hence we need to use a work-around with the `-webkit-inner-spin-button` rule, however in our CSS code we also have `-webkit-outer-spin-button` currently. According to both [the MDN compatibility data](https://developer.mozilla.org/en-US/docs/Web/CSS/::-webkit-outer-spin-button#browser_compatibility) and also manual testing in Google Chrome Beta 99, the `-webkit-outer-spin-button` rule is no longer necessary and we can thus clean-up the CSS a tiny bit.
This was added in PR 4516 specifically for Safari on iOS devices, but according to MDN it should no longer be necessary now; see https://developer.mozilla.org/en-US/docs/Web/CSS/-webkit-overflow-scrolling#browser_compatibility
According to the MDN compatibility data, this CSS feature:
- Was never implemented anywhere *except* for Safari on iOS.
- Was never standardized and thus never existed in an *unprefixed* version.
- Has now been removed, starting with Safari version 13.
Given that [the FAQ](https://github.com/mozilla/pdf.js/wiki/Frequently-Asked-Questions#faq-support) already lists Safari as "Mostly" supported, and that the default viewer is written primarily for Mozilla Firefox, it ought to be fine to remove these CSS rules now.
Since the Autoprefixer plugin indirectly depends on this, it seems like a good idea to add this as a direct dependency in the PDF.js project to hopefully avoid having to manually update `caniuse-lite` in the future; see https://github.com/browserslist/browserslist#browsers-data-updating
Also, slightly tweaks the Autoprefixer config for GENERIC-builds of the PDF.js library; note that this change doesn't affect the contents of the *built* `web/viewer.js` file.
The "External: Promise"-page in the JSDocs pre-dates the introduction of `Promise`s, as a generally available standard JS feature, by a number of years. Hence it now longer seems necessary, as far as I can tell, to include this "special" page in the documentation.
Also, while unrelated to the rest of the patch, updates the `test/`-folder description in the documentation.
The call-sites are replaced by direct `typeof`-checks instead, which removes unnecessary function calls. Note that in the `src/`-folder we already had more `typeof`-cases than `isString`-calls.
This removes the `DocumentInfoValidators` structure, and thus (slightly) simplifies the code overall. With these changes we only have to iterate through, and validate, the actually available Dictionary entries.
- Scroll the selected reference into view (makes it easier to tell which pdf you're looking at)
- Show the keyboard shortcuts (easier for new people)
- Keep the test/ref controls visible (if you scroll you can now tell if you're looking at a test or ref)
Sometimes I get a "Unable to find target with id XXX closeTarget..." error
when running tests which happens when test.js tries to close all the
open pages. I haven't been able to fully verify since this is intermittent,
but I think this is coming from us closing the window in driver.js and also
trying to close it in test.js.
- it aims to fix:
- https://bugzilla.mozilla.org/show_bug.cgi?id=1753075;
- https://bugzilla.mozilla.org/show_bug.cgi?id=1743245;
- https://bugzilla.mozilla.org/show_bug.cgi?id=1710019;
- issue #13211;
- issue #14521.
- previously we were trying to adjust lineWidth to have something correct after the current transform is applied but this approach was not correct because finally the pixel is rescaled with the same factors in both directions.
And sometimes those factors must be different (see bug 1753075).
- So the idea of this patch is to apply a scale matrix to the current transform just before setting lineWidth and stroking. This scale matrix is computed in order to ensure that after transform, a pixel will have its two thickness greater than 1.
All call-sites that use `wrapReason` should be passing a (possibly cloned) `Error` to the helper function, hence we shouldn't need to have a fallback code-path for any other data.
Note that for the `cancel`/`error` methods on Streams, since PR 11115 we've been asserting that the argument is in fact an `Error` as intended.
When calling `wrapReason` from *rejected* Promises, we should also be guaranteed that an `Error` is provided thanks to the ESLint rules `no-throw-literal` and `prefer-promise-reject-errors`.
At this point in time, after recent rounds of clean-up, the `webkit`-prefixed Fullscreen API is the only remaining *browser-specific* compatibility hack in the `web/`-folder JavaScript code.
The standard, and thus unprefixed, Fullscreen API has been supported for *over three years* in both Mozilla Firefox and Google Chrome. [According to MDN](https://developer.mozilla.org/en-US/docs/Web/API/Fullscreen_API#browser_compatibility), the unprefixed Fullscreen API has been available since:
- Mozilla Firefox 64, released on 2018-12-11; see https://wiki.mozilla.org/Release_Management/Calendar#Past_branch_dates
- Google Chrome 71, released on 2018-12-04; see https://en.wikipedia.org/wiki/Google_Chrome_version_history
Hence *only* Safari now requires using a prefixed Fullscreen API, and it's thus (significantly) lagging behind other browsers in this regard.
Considering that the default viewer is written *specifically* to be the UI for the Firefox PDF Viewer, and that we ask users to not just use it as-is[1], I think that we should only support the standard Fullscreen API now.
Furthermore, note also that the FAQ already lists Safari as "Mostly" supported; see https://github.com/mozilla/pdf.js/wiki/Frequently-Asked-Questions#faq-support
---
[1] Note e.g. http://mozilla.github.io/pdf.js/getting_started/#introduction
> The viewer is built on the display layer and is the UI for PDF viewer in Firefox and the other browser extensions within the project. It can be a good starting point for building your own viewer. *However, we do ask if you plan to embed the viewer in your own site, that it not just be an unmodified version. Please re-skin it or build upon it.*
Now that the there's ECMAScript support for properly private methods on `class`es, we can use that instead and thus remove all of the `@private` JSDocs comments.
We can (and in my opinion should) use the standard `atob`/`btoa` functions, rather than manually re-implementing this functionality for the font-tests.
This will default to generating test images at the device pixel
ratio of the machine the tests are created on unless the
test explicitly defines and output scale using the
`outputScale` setting. This makes the test look visually
like they would on the machine they are running on. It
also allows us to test different output scales.
Currently we'll happily attempt to send any argument passed to this method over to the worker-thread, without doing any sort of validation.
That could obviously be quite bad, since there's first of all no protection against sending unclonable data. Secondly, it's also possible to pass data that will cause the `Ref.get` call in the worker-thread to fail immediately.
In order to address all of these issues, we'll now properly validate the argument passed to `PDFDocumentProxy.getPageIndex` and when necessary reject already on the main-thread instead.
Trying to use a non-string argument in either a `Cmd` or a `Name` is not intended, and would basically be an implementation error. Hence we can add a non-PRODUCTION check to enforce this, similar to the existing one used e.g. in the `Dict.set` method.
This removes the `ViewerPreferencesValidators` structure, and thus (slightly) simplifies the code overall. With these changes we only have to iterate through, and validate, the actually available Dictionary entries.
Trying to use a non-string `key` in a `Dict` is not intended, and would basically be an implementation error. Hence we can add a non-PRODUCTION check to enforce this, complementing the existing `value` check added in PR 11672.
The call-sites are replaced by direct `typeof`-checks instead, which removes unnecessary function calls. Note that in the `src/`-folder we already had more `typeof`-cases than `isNum`-calls.
These changes were *mostly* done using regular expression search-and-replace, with two exceptions:
- In `Font._charToGlyph` we no longer unconditionally update the `width`, since that seems completely unnecessary.
- In `PDFDocument.documentInfo`, when parsing custom entries, we now do the `typeof`-check once.
Unless you actually need to check that something is both a `Name` and also of the *correct* type, using `instanceof Name` directly should be a tiny bit more efficient since it avoids one function call and an unnecessary `undefined` check.
This patch uses ESLint to enforce this, since we obviously still want to keep the `isName` helper function for where it makes sense.
Unless you actually need to check that something is both a `Dict` and also of the *correct* type, using `instanceof Dict` directly should be a tiny bit more efficient since it avoids one function call and an unnecessary `undefined` check.
This patch uses ESLint to enforce this, since we obviously still want to keep the `isDict` helper function for where it makes sense.
Unless you actually need to check that something is both a `Cmd` and also of the *correct* type, using `instanceof Cmd` directly should be a tiny bit more efficient since it avoids one function call and an unnecessary `undefined` check.
This patch uses ESLint to enforce this, since we obviously still want to keep the `isCmd` helper function for where it makes sense.
Given that we expose `PDFObjects`-instances, via the `commonObjs` and `objs` properties, on the `PDFPageProxy`-instances this ought to help provide slightly better TypeScript definitions.
The manually tracked `resolved`-property is no longer necessary, since the same information is now directly available on all `PromiseCapability`-instances.
Furthermore, since the `PDFObjects.resolve` method is not documented as accepting e.g. only Object-data, we probably shouldn't resolve the `PromiseCapability` with the `data` and instead only store it on the `PDFObjects`-instance.[1]
---
[1] While Objects are passed by reference in JavaScript, other primitives such as e.g. strings are passed by value and the current implementation *could* thus lead to increased memory usage. Given how we're using `PDFObjects` in the PDF.js code-base none of this should be an issue, but it still cannot hurt to change this.
This ensures that the underlying data cannot be accessed directly, from the outside, since that's definately not intended here.
Note that we expose `PDFObjects`-instances, via the `commonObjs` and `objs` properties, on the `PDFPageProxy`-instances hence these changes really cannot hurt.
This helper function is not really needed, since it's just a wrapper around a simple `instanceof` check, and it only adds unnecessary indirection in the code.
*Please note:* I'm completely fine with this patch being rejected, and the issue instead closed as WONTFIX, since this is unfortunately a case where the TypeScript definitions dictate how we can/cannot write JavaScript code.
Apparently the TypeScript definitions generation converts the existing `PixelsPerInch` code into a `namespace` and simply ignores the getter; please see a7fc0d33a1/types/src/display/display_utils.d.ts (L223-L226)
Initially I tried tagging `PixelsPerInch` as en `@enum`, see https://jsdoc.app/tags-enum.html, however that unfortunately didn't help.
Hence the only good/simple solution, as far as I'm concerned, is to convert `PixelsPerInch` into a class with `static` properties. This patch results in the following diff, for the `gulp types` build target:
```diff
@@ -195,9 +195,10 @@
*/
static toDateObject(input: string): Date | null;
}
-export namespace PixelsPerInch {
- const CSS: number;
- const PDF: number;
+export class PixelsPerInch {
+ static CSS: number;
+ static PDF: number;
+ static PDF_TO_CSS_UNITS: number;
}
declare const RenderingCancelledException_base: any;
export class RenderingCancelledException extends RenderingCancelledException_base {
```
In some cases, in the `PDFPageView` implementation, we're modifying the `sx`/`sy` properties when CSS-only zooming is being used.
Currently this requires that you remember to *manually* update the `scaled` property to prevent issues, which doesn't feel all that nice and also seems error-prone. By replacing the `scaled` property with a getter, this is now handled automatically instead.
The `CanvasRenderingContext2D.backingStorePixelRatio` property was never standardized, and only Safari set (its prefixed version of) it to anything other than `1`.
Note that e.g. MDN doesn't contain any information about this property, and one of the few sources of information (at this point) is the following post: https://stackoverflow.com/questions/24332639/why-context2d-backingstorepixelratio-deprecated
Hence we can simplify the `getOutputScale` helper function, by removing some dead code, and now it no longer requires any parameters when called.
Soft masks can be enabled/disabled at anytime and at different
points in the save/restore stack. This can lead to
the amount of save/restores becoming unbalanced across the
two canvases. Instead of save/restoring on the temporary canvas
change it so we only track state on the main (suspended canvas).
I was also getting an out balance stack from patterns, so I've also
fixed that and added a warning that will at least show up on chrome.
It would be nice to add this so Firefox at some point too.
Fixes#11328, #14297 and bug 1755507
At this point all the various Stream-classes extends an abstract base-class, hence this helper function is no longer necessary and only adds unnecessary indirection in the code.
Unfortunately I don't have a test-case that breaks without this change, however the `stringToPDFString` helper function will fail if anything other than a string is passed to it.
The changes in this patch thus make this code more-or-less identical to that found in the `Catalog.{_collectJavaScript, parseDestDictionary}` methods.
Given that the `Navigator` interface has been available since "forever", please see https://developer.mozilla.org/en-US/docs/Web/API/Navigator#browser_compatibility, it's somewhat difficult to see why these checks are actually necessary since the viewer is only intended for usage in browsers.
Looking at the history of the code, this functionality was originally placed in the general `src/shared/compatibility.js` file which could thus run in e.g. worker-threads and Node.js environments (where the `Navigator` interface isn't available).
- it aims to fix#14562;
- 'X-\n' were not correctly positioned;
- when X is a diacritic (e.g. in "sä-\n", which is decomposed into "sa¨-\n") we must handle both things:
- diacritics on the one hand;
- "-\n" on the other hand.
According to the MDN compatibility data, see https://developer.mozilla.org/en-US/docs/Web/API/ReadableStream#browser_compatibility, all browsers that we support have native `ReadableStream` implementations (since quite some time too).
Hence only Node.js is now lagging behind w.r.t. `ReadableStream` support, and its experimental implementation doesn't really help us given the life-span of the LTS releases (see https://en.wikipedia.org/wiki/Node.js#Releases).
It seems quite unfortunate to bundle a `ReadableStream` polyfill in the `legacy` builds when it's unnecessary in browsers, given its overall size, but fortunately we can avoid that by simply listing `web-streams-polyfill` as a dependency for the `pdfjs-dist` library.
With recent changes, specifically PR 14515 *and* the previous patch, the `createObjectURL` helper function is now only used with the SVG back-end.
All other call-sites, throughout the code-base, are now using `URL.createObjectURL(...)` directly and it no longer seems necessary to keep exposing the helper function in the API.
Finally, the `createObjectURL` helper function is moved into the `src/display/svg.js` file to avoid unnecessarily duplicating this code on both the main- and worker-threads.
Given that most of the code-base is already using native functionality, we can update these unit-tests similarily as well.
- For the `blob:`-URL test, we simply use `URL.createObjectURL(...)` and `Blob` directly instead.
- For the `data:`-URL test, we simply use `btoa` to do the Base64 encoding and then build the final URL-string.
This is essentially a *continuation* of PR 7926, where we added support for rejecting the current `PDFDocumentLoadingTask`-promise by throwing inside of the `onPassword`-callback.
Hence the naive way to address [bug 1754421](https://bugzilla.mozilla.org/show_bug.cgi?id=1754421) would be to simply throw in the `onPassword`-callback used in the default viewer. However it unfortunately turns out to not work, since the password input/validation is asynchronous, and we thus need another approach.
The simplest solution that I can come up with here, is thus to *extend* the `onPassword`-callback to also reject the current `PDFDocumentLoadingTask`-instance if an `Error` is explicitly passed as the input to the callback function. (This doesn't feel great, but I cannot see a better solution that isn't really complicated.)
While it's obviously fine to use the same PDF document in different reference-tests, note how we e.g. have both `eq` and `text` tests for one document, we should always avoid adding *duplicate* files in the `test/pdfs/` folder.
The file used in this test-case is *identical* to, i.e. the md5 entry perfectly matches, the file used with the `xfa_bug1716380` test-case.
While it's obviously fine to use the same PDF document in different reference-tests, note how we e.g. have both `eq` and `text` tests for one document, we should always avoid adding *duplicate* files in the `test/pdfs/` folder.
This appears to be consistent with the behaviour in both Adobe Reader and PDFium (in Google Chrome); this is essentially the same approach as used for a single decimal point in PR 9827.
Please note that while we "support" some (by now) fairly old browsers, that essentially means that the library (and viewer) will load and that the basic functionality will work as intended.[1]
However, in older browsers, some functionality may not be available and generally we'll ask users to update to a modern browser when bugs (specific to old browsers) are reported.[2]
There's always a question of just how old browsers the PDF.js contributors can realistically support, and here I'm suggesting that we place the cut-off point at approximately *three* years.
With that in mind, this patch updates the *minimum* supported browsers (and environments) as follows:
- Chrome 73, which was released on 2019-03-12; see https://en.wikipedia.org/wiki/Google_Chrome_version_history
- Firefox ESR (as before); see https://wiki.mozilla.org/Release_Management/Calendar
- Safari 12.1, which was released on 2019-03-25; see https://en.wikipedia.org/wiki/Safari_version_history#Safari_12
- Node.js 12, which was release on 2019-04-23 (and will soon reach EOL); see https://en.wikipedia.org/wiki/Node.js#Releases
---
[1] Assuming a `legacy`-build is being used, of course.
[2] In general it's never a good idea to use an old/outdated browser, since those may contain *known* security vulnerabilities.
Note that the *browser* findbar in Firefox uses "Title Case" for the labels, and it thus seem like a good idea to ensure that `PDFFindBar` in consistent with that.
Furthermore, the new label added in PR #13261 uses the "Title Case" format which means that currently the default viewer findbar looks inconsistent.
*Please note:* Based on the official Firefox localization docs, see https://firefox-source-docs.mozilla.org/l10n/overview.html#string-updates, changing only the casing should *not* require updating the key:
> 1) If the change is minor, like fixing a spelling error or case, the developer should update the en-US translation without changing the l10n-id.
When the viewer becomes narrow, the `PDFFindBar` will (forcibly) wrap its elements to prevent it from extending to the full window width.
Currently, after PR 13261, this now leads to the `findResultsCount` span taking up vertical space *unconditionally* when the findbar is wrapped. To avoid a blank space being shown in this case, before searching has begun, place the `findResultsCount` span in a "message" rather than an "options" container.
- get original index in using a dichotomic seach instead of a linear one;
- normalize the text in using NFD;
- convert the query string into a RegExp;
- replace whitespaces in the query with \s+;
- handle hyphens at eol use to break a word;
- add some \s* around punctuation signs
[api-minor] Remove the `normalizeWhitespace` option in the `PDFPageProxy.{getTextContent, streamTextContent}` methods (issue 14519, PR 14428 follow-up)
With these changes, we'll now *always* replace all whitespaces with standard spaces (0x20). This behaviour is already, since many years, the default in both the viewer and the browser-tests.
Either the latest Chromium update, the latest Puppeteer update, or a combination of them both are now causing the Windows bot to timeout during the browser-tests; please see PR 14392.
- it aims to fix#14502 and bug 1721335;
- Acrobat and Pdfium do the same;
- it'll avoid to have truncated data when printed;
- change the factor to compute font size in using field height: lineHeight = 1.35*fontSize
- this is the value used by Acrobat.
- in order to not have truncated strings on the bottom, add few basic metrics for standard fonts.
This `disableCreateObjectURL` option was originally introduced all the way back in PR 4103 (eight years ago), in order to work-around `URL.createObjectURL()`-bugs specific to Internet Explorer.
In PR 8081 (five years ago) the `disableCreateObjectURL` option was extended to cover Google Chrome on iOS-devices as well, since that configuration apparently also suffered from `URL.createObjectURL()`-bugs.[1]
At this point in time, I thus think that it makes sense to re-evaluate if we should still keep the `disableCreateObjectURL` option.
- For Internet Explorer, support was explicitly removed in PDF.js version `2.7.570` which was released one year ago and all IE-specific compatibility code (and polyfills) have since been removed.
- For Google Chrome on iOS-devices, while we still "support" such configurations, it's *not* the focus of any development and platform-specific bugs are thus often closed as WONTFIX.
Note here that at this point in time, the `disableCreateObjectURL` option is *only* being used in the viewer and any `URL.createObjectURL()`-bugs browser/platform bugs will thus not affect the main PDF.js library. Furthermore, given where the `disableCreateObjectURL` option is being used in the viewer the basic functionality should also remain unaffected by these changes.[2]
Furthermore, it's also possible that the `URL.createObjectURL()`-bugs have been fixed in *browser* itself since PR 8081 was submitted.[3]
Obviously you could argue that this isn't a lot of code, w.r.t. number of lines, and you'd be technically correct. However, it does add additional complexity in a few different viewer components which thus add overhead when reading and working with this code.
Finally, assuming the `URL.createObjectURL()`-bugs are still present in Google Chrome on iOS-devices, I think that we should ask ourselves if it's reasonable for the PDF.js project (and its contributors) to keep attempting to support a configuration if the *browser* developers still haven't fixed these kind of bugs!?
---
[1] According to https://groups.google.com/a/chromium.org/forum/#!topic/chromium-html5/RKQ0ZJIj7c4, which is linked in PR 8081, that bug was mentioned/reported as early as the 2014 (eight years ago).
[2] Viewer functionality such as e.g. downloading and printing may be affected.
[3] I don't have access to any iOS-devices to test with.
Either the latest Chromium update, the latest Puppeteer update, or a combination of them both are now causing the Windows bot to timeout during the browser-tests.
To unblock both the updates and other improvements (i.e. the `structuredClone` polyfill), let's simply disable the problematic configuration for now since this a Mozilla project after all.
This allows us to remove the manually implemented `structuredClone` polyfill, thus reducing the maintenance burden for the `LoopbackPort` class; refer to https://github.com/zloirock/core-js#structuredclone
*Please note:* While `structuredClone` support landed already in Firefox 94, Google Chrome only added it in version 98 (currently in Beta). However, given that the `LoopbackPort` will only be used together with *fake workers* in browsers this shouldn't be too much of a problem.[1]
For Node.js environments, where *fake workers* are unfortunately necessary, using a `legacy/`-build is already required which thus guarantees that the `structuredClone` polyfill is available.
Also, the patch updates core-js to the latest version since that one includes `structuredClone` improvements; please see https://github.com/zloirock/core-js/releases/tag/v3.20.3
---
[1] Given that we only support browsers with proper worker support, if *fake workers* are being used that essentially indicates a configuration problem/error.
- it aims to fix#14497;
- previously, only rotations with an angle 0, 90, 180 or 270 were taken into account;
- so generalize to any angle but keep the fast path for 0, 90, ... because they're likely more common than anything else.
So that Firefox doesn't switch to pixel mode for compat with other
browsers.
This should fix https://github.com/mozilla/pdf.js/issues/14476, in terms
of restoring the previous behavior.
We probably want to change the pixel-based scrolling code to not scroll
so much (the deltaMode stuff normalizes to +/-1 tick for each wheel
event, perhaps the pixel-based value should do the same).
This commit fixes Bug 1743245 (Grided PDF file lines rendered too thick) which was created by a fix for #12868 .
The lineWidth was set to round(1 * this._combinedScaleFactor) when the pixel is drawn as a parallelorgam with a height <1. This fix changes this to floor(1*this._combinedScaleFactor) .
This change shows a visual result comparable to Chrome and Acrobat.
Regarding the last PR 3 statements in canvas.js are affected and will change with this commit (stroke and paintChar).
renaming the reference files to naming comvention
Given that the regular expression has already become more complex (after the initial patch adding it), it seems to me that it probably cannot hurt to add a global cache to reduce unnecessary re-parsing.
Obviously the `Glyph`-instances are being cached *per* font, however in most documents multiple fonts are being used and in practice there's very often a fair amount of overlap between the /ToUnicode-data in different fonts[1].
Consider for example loading and rendering the entire `tracemonkey.pdf` document (from the test-suite), which isn't a particularily large document. In that case the `getCharUnicodeCategory` function is being called a total of `601` times, however there's only `106` *unique* unicode-chars being checked.
*Please note:* In practice I suppose that this won't have a *huge* effect on overall performance, however given the relative simplicity of this patch I figured that it'd not hurt to submit it for review.
---
[1] Consider e.g. how there's usually different fonts used for regular, bold, respectively italic text.
- it aims to fix issue #14307;
- this event has been added recently in Firefox and we can now use it;
- fix few bugs in aform.js or in annotation_layer.js;
- add some integration tests to test keystroke events (see `AFSpecial_Keystroke`);
- make dispatchEvent in the quickjs sandbox async.
- it aims to fix https://bugzilla.mozilla.org/show_bug.cgi?id=1749563;
- use some helper functions to get (u|i)int** values in buffer: it helps to have a clearer code;
- in composite glyphes the translations values with a transformations are signed so consequently get some int8 instead of uint8;
- add few TODOs.
Please refer to https://www.pdfa.org/norm-refs/Type1Fonts.pdf#page=15 for the expected format for the /CharStrings entries.
In the referenced PDF document the /CharStrings are missing the expected end-token, which causes us to swallow the start of the next glyph name.
After the changes in PR 14428 we can *directly*, and more efficiently, handle whitespace conversion in `PartialEvaluator.getTextContent` when the `normalizeWhitespace` option is being used.
This way we no longer need a separate helper function for this, and can avoid having to (again) iterate through the text and checking each character. Finally, this also removes the need for using a regular expression on e.g. all non-ASCII text.
Inlining the checks should be a *tiny bit* more efficient, since it avoids have to make *unconditional* function calls in these fairly commonly used helper functions.
*Please note:* This is a tentative patch, since I don't know if this is deemed important enough to fix.
The new event could be seen as a *supplement* to the existing "documentinit" and "documentloaded" events, but for the case when a PDF document fails to load.
To make the "documenterror" event generally useful, it'll include both the localized error message as well as the original reason for the error (when that exists).
This patch implements this by looking for the UTF-8 BOM, i.e. `\xEF\xBB\xBF`, in order to determine the encoding.[1]
The actual conversion is done using the `TextDecoder` interface, which should be available in all environments/browsers that we support; please see https://developer.mozilla.org/en-US/docs/Web/API/TextDecoder#browser_compatibility
---
[1] Assuming that everything lacking a UTF-16 BOM would have to be UTF-8 encoded really doesn't seem correct.
In corrupt PDF documents Type3 fonts may introduce circular dependencies, thus resulting in the affected font(s) never loading and parsing/rendering never completing.
Note that I've not seen any real-world examples of this kind of font corruption, but the attached PDF document was rather found in https://github.com/pdf-association/safedocs/tree/main/Miscellaneous%20Targeted%20Test%20PDFs
*Please note:* That repository contains a number of reduced test-cases that are specifically intended to test interoperability (between PDF viewer) and parsing/rendering for various kinds of strange/corrupt PDF documents.
Some of the test-cases found there may thus not make sense to try and "fix" upfront, in my opinion, unless the problems are also found in real-world PDF documents.
While `PageViewport` apparently makes sense in TypeScript environments, given that it's being returned by the `PDFPageProxy.getViewport`-method in the API, we really don't want to extend the *public* API by simply exporting the class directly in `src/pdf.js` since it should never be called/initialized manually.
Hence we follow the same pattern as in PR 14013, and also extend the API unit-tests to ensure that `PDFPageProxy.getViewport` always returns a `PageViewport`-instance as expected.
This prevents the `BaseSVGFactory.create`-method from throwing, and thus preventing any remaining Annotations (on the page) from rendering in corrupt documents.
This helper function has never been used in e.g. the worker-thread, hence its placement in `src/shared/util.js` led to a *small* amount of unnecessary duplication.
After the previous patches this helper function is now *only* used in the viewer, hence it no longer seems necessary to expose it through the official API.
*Please note:* It seems somewhat unlikely that third-party users were relying *directly* on this helper function, which is why it's not being exported as part of the viewer components. (If necessary, we can always change this later on.)
As part of the changes/improvement in PR 14092, we're no longer using the `addLinkAttributes` directly in e.g. the AnnotationLayer-code.
Given that the helper function is now *only* used in the viewer, hence it no longer seems necessary to expose it through the official API.
*Please note:* It seems somewhat unlikely that third-party users were relying *directly* on the helper function, which is why it's not being exported as part of the viewer components. (If necessary, we can always change this later on.)
This structure contains *almost* exclusively references to DOM elements (and a couple of simple strings), rather than complete classes/functions. Hence the `eventBus`-option sticks out a fair bit, and I'd guess that it's *mostly* unused in e.g. third-party implementations.
Given that we, in multiple places, mention that the default viewer shouldn't be used as-is I really don't think that we need to keep this special `eventBus`-option around. Furthermore, nowadays it's also a lot easier to (safely) access the existing `EventBus`-instance in the viewer; see https://github.com/mozilla/pdf.js/wiki/Third-party-viewer-usage#initialization-promise which shows how to listen for the default viewer being initialized (and its `eventBus` thus being available).
The patch in PR 14335 *essentially* re-introduced the old code from before PR 3848, however looking at this code a bit closer it should be possible to simplify it by making the method asynchronous.
While this method is currently only used as a *fallback* in corrupt documents, the way that `MissingDataException`s are handled is less than ideal. Note that if a `MissingDataException` is thrown, we're forced to re-parse the *entire* /Pages tree[1].
With this method now being asynchronous, we're able to handle fetching of References in a *much* easier/nicer way than before without having to throw `MissingDataException`s and re-parse anything.
These changes also let us simplify the call-site slightly, by calling the method *directly* instead of using the `PDFManager`-instance (since again it will no longer throw `MissingDataException`s).
Furthermore, this patch contains the following other changes:
- Reduce unnecessary duplication in the various `catch` handlers throughout the method, by simply moving the `XRefEntryException` handling into the `addPageError` helper function instead.
- Move the "circular references"-check to occur slightly earlier, since there's obviously no point in asynchronously fetching data just to then throw an Error *immediately* afterwards.
---
[1] Imagine e.g. a thousand page document, where there's a `MissingDataException` thrown when fetching/parsing page 900.
This method is now being used a lot more, compared to when it's added, since it's now used together with scripting as part of the `PDFDocument.fieldObjects` parsing (called during viewer initialization).
For /Page Dictionaries that we've already parsed, the `pageIndex` corresponding to a particular Reference is already known and we're thus able to skip *all* parsing in the `Catalog.getPageIndex` method for those cases.
I happened to notice that we didn't have *any* unit-tests for either `getFieldObjects` or `getCalculationOrderIds`, on the `PDFDocumentProxy` class, which seems unfortunate since it's API functionality that we depend on in e.g. the viewer.
Besides converting `Catalog.getPageDict` to an `async` method, thus simplifying the code, this patch also allows us to pro-actively fix a existing issue.
Note how we're looking up References in such a way that `MissingDataException`s won't cause trouble, however it's *technically possible* that the entries (i.e. /Count, /Kids, and /Type) in a /Pages Dictionary could actually be indirect objects as well. In the existing code this could lead to *some*, or even all, pages failing to load/render as intended.
In practice that doesn't *appear* to happen in real-world PDF documents, but given all the weird things that PDF software do I'd prefer to fix this pro-actively (rather than waiting for a bug report).
With `Catalog.getPageDict` being `async` this is now really simple to address, however I didn't want to introduce a bunch more *unconditional* asynchronicity in this method if it could be avoided (since that could slow things down). Hence we'll *synchronously* lookup the *raw* data in a /Pages Dictionary, and only fallback to asynchronous data lookup when a Reference was encountered.
In addition to the above, this patch also makes the following notable changes:
- Let `Catalog.getPageDict` *consistently* reject with the actual error, regardless of what data we're fetching. Previously we'd "swallow" the actual errors except when looking up Dictionary entries, which is inconsistent and thus seem unfortunate. As can be seen from the updated unit-tests this change is API-observable, hence why the patch is tagged `[api-minor]`.
- Improve the consistency of the Dictionary /Type-checks in both the `Catalog.getPageDict` and `Catalog.getAllPageDicts` methods.
In `Catalog.getPageDict` there's a fallback code-path where we're *incorrectly* checking the /Page Dictionary for a /Contents-entry, which is wrong since a /Page Dictionary doesn't need to have a /Contents-entry in order to be valid.
For consistency the `Catalog.getAllPageDicts` method is also updated to handle errors in the /Type-lookup correctly.
- Reduce the `PagesCountLimit.PAUSE_EAGER_PAGE_INIT` viewer constant, to further improve loading/rendering performance of the *second* page during initialization of very long documents; PR 14359 follow-up.
*This addresses the following case missing from the previous patch:*
The viewer is loaded in an *active* window/tab, and enough time is allowed to pass in order to allow rendering to start. However, if the user then switches to another tab (or another program) *before* rendering has finished, the "load" event also needs to be unblocked.
Given that `requestAnimationFrame` is being used, see the `src/diplay/api.js` file, an inactive window/tab means that rendering will not run and we'll thus not fetch all pages. The latter is a requirement for the "load" event to be unblocked, in the MOZCENTRAL-version of, the default viewer.
This patch is a *partial* solution, since it only addresses the following situations:
- A *background* tab (containing the viewer) is reloaded, e.g. via the tab-bar context menu.
- The viewer is loaded in a active tab, but the user switches away from it (or switches to another program window) *before* rendering has started.
Most likely this code predates our use of Node.js, but in Node.js asking
for user confirmation is a solved problem, so we can remove the custom
logic we have for this, which overall makes things much simpler.
This provides a work-around for badly generated PDF documents that contain *negative* /FitH parameters (in the referenced issue the value `-32768` is used).
This patch, first of all, removes circular dependencies in the TypeScript definitions. Secondly, it also moves `RenderingStates` into `web/ui_utils.js` to break another type-dependency and directly use the `XfaLayerBuilder` during XFA-printing.
Finally, note that this patch *slightly* reduces the size of the default viewer (e.g. in the `MOZCENTRAL` build) by not having to bundle code which is completely unused.
This patch circumvents the issues seen when trying to update TypeScript to version `4.5`, by "simply" fixing the broken/missing JSDocs and `typedef`s such that `gulp typestest` now passes.
As always, given that I don't really know anything about TypeScript, I cannot tell if this is a "correct" and/or proper way of doing things; we'll need TypeScript users to help out with testing!
*Please note:* I'm sorry about the size of this patch, but given how intertwined all of this unfortunately is it just didn't seem easy to split this into smaller parts.
However, one good thing about this TypeScript update is that it helped uncover a number of pre-existing bugs in our JSDocs comments.
The size of the `web/ui_utils.js` file has increased over time, as more code has been added to (or moved into) that file. To reduce its size slightly, this patch moves the event-related functionality into a separate file.
In PR 14114 this was only added to the default viewer, which means that in the viewer components the user would need to *manually* implement /Lang handling. This was (obviously) a bad choice, since the viewer components already support e.g. structTrees by default; sorry about overlooking this!
To avoid having to make *two* `getMetadata` API-calls[1] very early during initialization, in the default viewer, the API will now cache its result. This will also come in handy elsewhere in the default viewer, e.g. by reducing parsing when opening the "document properties" dialog.
---
[1] This not only includes a round-trip to the worker-thread, but also having to re-parse the /Metadata-entry when it exists.
After the changes in PR 14338, specifically in the `XRef.parse`-method, the /Pages-entry will now always have been fetched/validated when the `Catalog`-instance is created.
Hence we can directly access the /Pages-entry in `Catalog.getPageDict` and thus avoid *one* asynchronous data-lookup per page in the document. (In practice this is unlikely to show up in e.g. benchmarks, but it really cannot hurt.)
Finally, make sure that the `getPageDict`/`getAllPageDicts`-methods track the /Pages-tree reference correctly to prevent circular references in corrupt documents.
We use string arguments in all other places, so these two places are a
bit inconsistent in that sense. Moreover, it's just one argument now,
which makes it a bit easier to read and see what it does because we
don't have to pass the always-empty options argument anymore. Finally,
doing it like this ensures it works in all Puppeteer versions given
https://github.com/puppeteer/puppeteer/issues/7836.
Once the upstream bug is fixed it can be enabled again because it's
causing way too much noise now. This is tracked in issue #14293. Note
that I deliberately added a new block so we can easily remove it later
on and because the other block is about another bug.
By making this API-call *unconditionally*, we introduce a (slight) delay in the initialization of *all* documents.
That seems quite unfortunate, since `pdfjs.enablePermissions` is off by default, and it thus seem better only do the API-call when actually needed; sorry about this!
For encrypted PDF documents without the required permissions set, this patch adds support for disabling of form editing. However, please note that it also requires that the `pdfjs.enablePermissions` preference is set to `true`[1] (since PDF document permissions could be seen as user hostile).
Based on https://www.adobe.com/content/dam/acom/en/devnet/pdf/pdfs/PDF32000_2008.pdf#G6.1942134, this condition hopefully makes sense.
---
[1] Either manually with `about:config`, or using e.g. a [Group Policy](https://github.com/mozilla/policy-templates).
Version 14 that we used before is now in maintenance mode, so we should
upgrade to the most recent LTS version.
Moreover, use the most recent `setup-node` workflow version and syntax;
see https://github.com/actions/setup-node#usage.
This patch is essentially *another* continuation of PR 11263, which tried to improve loading/initialization performance of *very* large/long documents.
For most documents, unless they're *very* long, we'll eagerly initialize all of the pages in the viewer. For shorter documents having all pages loaded/initialized early provides overall better performance/UX in the viewer, however there's cases where it can instead *hurt* performance.
For documents with a couple of thousand pages[1], the parsing and pre-rendering of the *second* page of the document can be delayed (quite a bit). The reason for this is that we trigger `PDFDocumentProxy.getPage` for *all pages* early during the viewer initialization, which causes the worker-thread to be swamped with handling (potentially) thousands of `getPage`-calls and leaving very little time for other parsing (such as e.g. of operatorLists).
To address this situation, this patch thus proposes temporarily "pausing" the eager `PDFDocumentProxy.getPage`-calls once a threshold has been reached, to give the worker-thread a change to handle other requests.[2]
Obviously this may *slightly* delay the "pagesloaded" event in longer documents, but considering that it's already the result of asynchronous parsing that'll hopefully not be seen as a blocker for these changes.[3]
---
[1] A particularly problematic example is https://github.com/mozilla/pdf.js/files/876321/kjv.pdf (16 MB large), which is a document with 2236 pages and a /Pages-tree that's only *one* level deep.
[2] Please note that I initially considered simply chaining the `PDFDocumentProxy.getPage`-calls, however that'd slowed things down for all documents which didn't seem appropriate.
[3] This patch will *hopefully* also make it possible to re-visit PR 11312, since it seems that changing `Catalog.getPageDict` to an `async` method wasn't the problem in itself. Rather it appears that it leads to slightly different timings, thus exacerbating the already existing issues with the worker-thread being overloaded by `getPage`-calls.
Having recently worked with that method, there's a couple of (very old) issues that I'd also like to address and having `Catalog.getPageDict` be `async` would simplify things a great deal.
Rather than "swallowing" the actual Errors, when data fetching fails, ensure that they're always being propagated as intended to the call-site instead.
Note that we purposely handle `XRefEntryException` specially, to make it possible to fallback to indexing all XRef objects.
Rather than trying, and failing, to fetch the entire /Pages-tree for documents with corrupt XRef tables, let's fallback to indexing all objects *before* trying to invoke the `Catalog.getAllPageDicts` method.
Given that not all pages necessarily are being accessed, or that the pages may be accessed out of order, using a `Map` seems like a more appropriate data-structure here.
Finally, also changes the `pagePromises` to a *private* property since it's not supposed to be accessed from the "outside".
Given that not all pages necessarily are being accessed, or that the pages may be accessed out of order, using a `Map` seems like a more appropriate data-structure here.
For one thing, this simplifies iteration since we no longer have to worry about/check if `pageCache`-entries are undefined (which will happen for *sparse* `Array`s).
Of particular note is that we're no longer attempting to "null" the `pageCache`-entry from within the `PDFPageProxy._destroy`-method. Given that *synchronous* JavaScript will always run to completion[1] and that we're looping through all pages in `WorkerTransport.destroy` and immediately clear the cache afterwards, that code did/does not really make a lot of sense (as far as I can tell).
Finally, also changes the `pageCache` to a *private* property since it's not supposed to be accessed from the "outside".
---
[1] Unless there are errors, of course.
PR 8207 added caching to improve the performance of `Catalog.getPageDict`, by not having to repeatedly fetch the same data and also reducing the asynchronicity of that method.
However, because of *another* oversight on my part, we're only caching /Page references once we've found the correct page. As long as all pages are loaded *in order* this doesn't really matter (happens by default in the viewer), but when `disableAutoFetch` is used the pages may be fetched in a more random order (this patch reduces the asynchronicity of `Catalog.getPageDict` slightly in that case).
Given that [bug 1336591](https://bugzilla.mozilla.org/show_bug.cgi?id=1336591) was just closed as fixed, thus fixing issue 8022 in Firefox, let's add a test-case to enable us to catch any future regressions either in PDF.js or in browsers themselves.
PR 8207 added caching to improve the performance of `Catalog.getPageDict`, by not having to repeatedly fetch the same data and also reducing the asynchronicity of that method.
However, because of annoying off-by-one errors[1] the caching became less efficient than it could/should be.[2] Note here that the /Pages-tree is zero-indexed, and that e.g. `pageIndex = 5` thus correspond to the *sixth* page of the document.
---
[1] In particular the `currentPageIndex + count < pageIndex` part.
[2] For example, even when loading a relatively small/simple document such as `tracemonkey.pdf` in the viewer, the number of `xref.fetchAsync(currentNode)` calls are reduced from `56` to `44` with this patch.
These changes improves the consistency ever so slightly in the `PDFOutlineViewer._dispatchEvent` method, by making sure that we can tell the following two cases apart:
- The "pagesloaded" event has *not yet* been fired.
- The "pagesloaded" event has been fired, but no pages were available.
*This patch can be tested e.g. with the `poppler-85140-0.pdf` document from the test-suite.*
For some sufficiently corrupt documents the `getDocument` call will succeed, but fetching even the very first page fails. Currently we only print error messages (in the console) from the `{BaseViewer, PDFThumbnailViewer}.setDocument` methods, but don't actually provide these errors to allow the viewer to handle them properly.
In practice this means that the GENERIC viewer won't display the `errorWrapper`, and in the MOZCENTRAL viewer the *browser* loading indicator is never hidden (since we never unblock the "load" event).
Trying to shadow a non-existent property is always an implementation mistake, since it leads to the `shadow`-call not having any effect.
In PR 14152 I overlooked the fact that it's fairly easy to enforce this during development/testing, since that can help catch e.g. simple spelling bugs.
Currently the `Catalog.metadata` getter only handles errors during parsing, however in a *corrupt* PDF document fetching of the raw /Metadata can obviously fail as well.
Without this patch the `PDFDocumentProxy.getMetadata` method, in the API, can thus fail which it *never* should and this will cause the viewer to not initialize all state as expected.
Fixes one of the documents in issue 14305.
Given that [bug 1336572](https://bugzilla.mozilla.org/show_bug.cgi?id=1336572) was just closed as fixed, thus fixing issue 8019 in Firefox[1], let's add a test-case to enable us to catch any future regressions either in PDF.js or in browsers themselves.
---
[1] It also seems to be working in Google Chrome, although I'm having a slightly difficult time deciphering *exactly* what configurations were affected when looking through issue 8019.
*This patch improves handling of a couple of PDF documents from issue 14303.*
- Update `XRef.indexObjects` to actually clear *all* XRef-caches. Invalid XRef tables *usually* cause issues early enough during parsing that we've not populated the XRef-cache, however to prevent any issues we obviously need to clear that one as well.
- Improve the /Root dictionary validation in `XRef.parse` (PR 9827 follow-up). In addition to checking that a /Pages entry exists, we'll now also check that it can be successfully fetched *and* that it's of the correct type. There's really no point trying to use a /Root dictionary that e.g. `Catalog.toplevelPagesDict` will reject, and this way we'll be able to fallback to indexing the objects in corrupt documents.
- Throw an `InvalidPDFException`, rather than a general `FormatError`, in `XRef.parse` when no usable /Root dictionary could be found. That really seems more appropriate overall, since all attempts at parsing/recovery have failed. (This part of the patch is API-observable, hence the tag.)
With these changes, two existing test-cases are improved and the unit-tests are updated/re-factored to highlight that. In particular `GHOSTSCRIPT-698804-1-fuzzed.pdf` will now both load and "render" correctly, whereas `poppler-395-0-fuzzed.pdf` will now fail immediately upon loading (rather than *appearing* to work).
*Please note:* This is similar to the method that existed prior to PR 3848, but the new method will *only* be used as a fallback when parsing of corrupt PDF documents.
The implementation in PR 14311 unfortunately turned out to be *way* too simplistic, as evident by the recently added test-files in issue 14303, since it may *cause* infinite loops in `PDFDocument.checkLastPage` for some corrupt PDF documents.[1]
To avoid this, the easiest solution that I could come up with was to fallback to eagerly parsing the *entire* /Pages-tree when the /Count-entry validation fails during document initialization.
Fixes *at least* two of the issues listed in issue 14303, namely the `poppler-395-0.pdf...` and `GHOSTSCRIPT-698804-1.pdf...` documents.
---
[1] The whole point of PR 14311 was obviously to *get rid of* infinte loops during document initialization, not to introduce any more of those.
Given that we're able to "render" this document, let's extend the unit-test to actually check that we're able to obtain the operatorList; although given the overall issues in the document it'll be empty.
This was added in PR 14311, but given that I completely missed to update the `PDFDocument.getPage` signature accordingly it's completely unused.
Given that things work just as fine as-is, let's simply remove that optional parameter for now; sorry about the churn here!
This only applies to severely corrupt documents, where it's possible that the `Parser` throws when we try to access e.g. a /Kids-entry in the /Pages-tree.
Fixes two of the issues listed in issue 14303, namely the `poppler-742-0.pdf...` and `poppler-937-0.pdf...` documents.
Given that Node.js doesn't support Workers, general PDF.js performance will be worse when compared to browsers. In an attempt to improve at least memory usage a little bit, update the Node.js examples to release page resources once parsing is done for that page.
This patch is essentially a continuation of PR 11263, which tried to improve loading/initialization performance of *very* large/long documents.
Note that browsers, in general, don't handle a huge amount of DOM-elements very well, with really poor (e.g. sluggish scrolling) performance once the number gets "large". Furthermore, at least in Firefox, it seems that DOM-elements towards the bottom of a HTML-page can effectively be ignored; for the PDF.js viewer that means that pages at the end of the document can become impossible to access.
Hence, in order to improve things for these *very* large/long documents, this patch will now enforce usage of the (recently added) PAGE-scrolling mode for these documents. As implemented, this will only happen once the number of pages *exceed* 15000 (which is hopefully rare in practice).
While this might feel a bit jarring to users being *forced* to use PAGE-scrolling, it seems all things considered like a better idea to ensure that the entire document actually remains accessible and with (hopefully) more acceptable performance.
Fixes [bug 1588435](https://bugzilla.mozilla.org/show_bug.cgi?id=1588435), to the extent that doing so is possible since the document contains 25560 pages (and is 197 MB large).
This was added on the assumption that the viewer would (eventually) start using the `PDFSinglePageViewer` for e.g. PAGE-scrolling mode and PresentationMode. However, having both a `PDFViewer` and a `PDFSinglePageViewer` side-by-side in the viewer would've been tricky to implement well, which is why PR 14112 implemented PAGE-scrolling for the general `BaseViewer` instead.
Given that the default viewer is no longer (potentially) going to use `PDFSinglePageViewer`, there's code in the `SecondaryToolbar` (and related CSS rules) which is now unnecessary.
Since NPM 7, which is over a year old now since it released in October
2020, NPM automatically transforms lock files from version 1 to version
2. In the NPM 7 release notes they reported:
"One change to take note of is the new lockfile format, which is
backwards compatible with npm 6 users. The lockfile v2 unlocks the
ability to do deterministic and reproducible builds to produce a
package tree."
Not only is this change backwards compatible (so older versions of NPM
will still be able to install everything as expected), reproducability
is also a nice property to have and modern NPM versions will otherwise
constantly do the conversion anyway, causing contributors to explicitly
have to revert the change. Therefore, I believe we should do this now
since it doesn't break backwards compatibility for consumers of this
file. It only means that producers of this file (i.e., us contributors)
need to use at least NPM 7 or higher (as of writing NPM 8 is even
available). According to https://nodejs.org/en/download/releases/ this
means contributors should at least run Node.js 15.0.0, while 17.1.0 is
the most recent as of writing, so to me that sounds reasonable to ask.
In Puppeteer 11 we noticed that Firefox doesn't shut down once the tests
are done anymore. I tracked this down to the `page.goto` call, in
`startBrowser`, never resolving anymore. I can only assume that
something changed in Puppeteer, possibly in combination with recent
Firefox Nightly versions, that caused this, but haven't been able to
fully track it down.
However, I did find that the problem is that the `load` event no longer
triggers, so fortunately we can fix the problem by explicitly waiting
for the `domcontentloaded` event instead. In general this change might
even be better since we now wait until the test framework is fully
loaded before we continue. Note that this also still works for the
current Puppeteer version.
I did find two upstream references that appear to track this issue, both
on the Puppeteer side and on the Firefox side, making me further suspect
that the issue is partly on both sides:
- https://github.com/puppeteer/puppeteer/issues/5806
- https://bugzilla.mozilla.org/show_bug.cgi?id=1706353
If a browser cannot be started, we currently get the following log:
`Error while starting firefox: [object Object]`. This is simply an
oversight from the initial Puppeteer integration work since we never got
into this code path before. With this fix the error log becomes more
useful: `Error while starting firefox: connect ECONNREFUSED ::1:45387`
*Please note:* While this patch on its own is sufficient to prevent the worker-thread from hanging, however in combination with PR 14311 these PDF documents will both load *and* render correctly.
Rather than focusing on the particular structure of these PDF documents, it seemed (at least to me) to make sense to try and prevent all circular references when fetching/looking-up data using the XRef table.
To avoid a solution that required tracking the references manually everywhere, the implementation settled on here instead handles that internally in the `XRef.fetch`-method. This should work, since that method *and* the `Parser`/`Lexer`-implementations are completely synchronous.
Note also that the existing `XRef`-caching, used for all data-types *except* Streams, should hopefully help to lessen the performance impact of these changes.
One *potential* problem with these changes could be certain *browser* exceptions, since those are generally not catchable in JavaScript code, however those would most likely "stop" worker-thread parsing anyway (at least I hope so).
Finally, note that I settled on returning dummy-data rather than throwing an exception. This was done to allow parsing, for the rest of the document, to continue such that *one* bad reference doesn't prevent an entire document from loading.
Fixes two of the issues listed in issue 14303, namely the `poppler-91414-0.zip-2.gz-53.pdf` and `poppler-91414-0.zip-2.gz-54.pdf` documents.
*This patch basically extends the approach from PR 10392, by also checking the last page.*
Currently, in e.g. the `Catalog.numPages`-getter, we're simply assuming that if the /Pages-tree has an *integer* /Count entry it must also be correct/valid.
As can be seen in the referenced PDF documents, that entry may be completely bogus which causes general parsing to breaking down elsewhere in the worker-thread (and hanging the browser).
Rather than hoping that the /Count entry is correct, similar to all other data found in PDF documents, we obviously need to validate it. This turns out to be a little less straightforward than one would like, since the only way to do this (as far as I know) is to parse the *entire* /Pages-tree and essentially counting the pages.
To avoid doing that for all documents, this patch tries to take a short-cut by checking if the last page (based on the /Count entry) can be successfully fetched. If so, we assume that the /Count entry is correct and use it as-is, otherwise we'll iterate through (potentially) the *entire* /Pages-tree to determine the number of pages.
Unfortunately these changes will have a number of *somewhat* negative side-effects, please see a possibly incomplete list below, however I cannot see a better way to address this bug.
- This will slow down initial loading/rendering of all documents, at least by some amount, since we now need to fetch/parse more of the /Pages-tree in order to be able to access the *last* page of the PDF documents.
- For poorly generated PDF documents, where the entire /Pages-tree only has *one* level, we'll unfortunately need to fetch/parse the *entire* /Pages-tree to get to the last page. While there's a cache to help reduce repeated data lookups, this will affect initial loading/rendering of *some* long PDF documents,
- This will affect the `disableAutoFetch = true` mode negatively, since we now need to fetch/parse more data during document initialization. While the `disableAutoFetch = true` mode should still be helpful in larger/longer PDF documents, for smaller ones the effect/usefulness may unfortunately be lost.
As one *small* additional bonus, we should now also be able to support opening PDF documents where the /Pages-tree /Count entry is completely invalid (e.g. contains a non-integer value).
Fixes two of the issues listed in issue 14303, namely the `poppler-67295-0.pdf` and `poppler-85140-0.pdf` documents.
Given that not all pages necessarily are being accessed, or that the pages may be accessed out of order, using a `Map` seems like a more appropriate data-structure here.
Furthermore, this patch also adds (currently missing) caching for XFA-documents. Loading a couple of such documents in the viewer, with logging added, shows that we're currently re-creating `Page`-instances unnecessarily for XFA-documents.
For this particular PDF document, we have `/W [1 2 166666666666666666666666666]` which obviously makes no sense.
While this patch makes no attempt at actually validating the entries in the /W-array, we'll now simply abort all processing when the end of the PDF document has been reached (thus preventing hanging the browser).
Please note that this patch doesn't enable the PDF document to be loaded/rendered, but at least it fails "correctly" now.
Fixes one of the issues listed in issue 14303, namely the `REDHAT-1531897-0.pdf`document.
This bug was surprisingly difficult to track down, since it didn't just depend on range-requests being used but also on how quickly the document was loaded. To even be able to reproduce this locally, I had to use a very small `rangeChunkSize`-value (note the unit-test).
The cause of this bug is a bogus entry in the XRef-table, causing us to attempt to request data from *beyond* the actual document size and thus getting into an infinite loop.
Fixes *one* of the issues listed in issue 14303, namely the `PDFBOX-4352-0.pdf` document.
*This patch can be tested e.g. with the `sizes.pdf` document in the test-suite.*
While this patch isn't necessarily the best solution, e.g. it might be possible to solve this with *only* CSS, it's what I was able to come up with to address an old issue.
The solution here re-uses the `spread`-class in PresentationMode, since that one already takes care of centering pages *vertically*, together with a dummy-page that takes up the entire height of the window.
Finally, some PresentationMode-related CSS-rules are also simplified slightly, since the changes in PR 14112 (using Page-scrolling) allows some clean-up here.
The issue that this patch fixes has existed ever since the viewer was first re-factored into components, however it only really affects the `disableAutoFetch = true` mode.
By default we're fetching all pages in `BaseViewer.setDocument`, and as part of the parsing/initialization we're also populating the `PDFLinkService`-pagesRefCache. The purpose of that cache is to make navigating to any internal destinations faster, by not having to (asynchronously) lookup the pageNumber via the API when handling the destination.
In comparison, when the `disableAutoFetch = true` mode is being used we're instead *lazily* initializing the pages in the `BaseViewer.#ensurePdfPageLoaded`-method. For some reason, that I can only assume is a simple oversight, we're not attempting to update the `PDFLinkService`-pagesRefCache in that case.
In the `BaseViewer` this cache is mostly relevant in the `disableAutoFetch = true` mode, since the pages are being initialized *lazily* in that case.
In the `PDFThumbnailViewer` this cache is mostly used for thumbnails that are actually being rendered, as opposed to those created directly from the "regular" pages.
Please note that I'm not suggesting that we remove these caches because they're only used in some situations, but rather because they're for all intents and purposes actually *redundant*. In the API itself, we're already caching both the page-promises and the actual pages themselves on the `WorkerTransport`-instance.
Hence these viewer-caches aren't really necessary in practice, and adds what to me mostly seems like an unnecessary level of indirection.[1]
Given that the viewer now relies on caching in the API itself, this patch also adds a new unit-test to ensure that page-caching works (and keep working) as expected.
---
[1] In the `WorkerTransport.getPage`-method the parameter is being validated on every call, but that's hardly enough code to warrant keeping the "duplicate" caches in the viewer in my opinion.
*Please note:* These changes will primarily benefit longer documents, somewhat at the expense of e.g. one-page documents.
The existing `PDFDocumentProxy.getStats` function, which in the default viewer is called for each rendered page, requires a round-trip to the worker-thread in order to obtain the current document stats. In the default viewer, we currently make one such API-call for *every rendered* page.
This patch proposes replacing that method with a *synchronous* `PDFDocumentProxy.stats` getter instead, combined with re-factoring the worker-thread code by adding a `DocStats`-class to track Stream/Font-types and *only send* them to the main-thread *the first time* that a type is encountered.
Note that in practice most PDF documents only use a fairly limited number of Stream/Font-types, which means that in longer documents most of the `PDFDocumentProxy.getStats`-calls will return the same data.[1]
This re-factoring will obviously benefit longer document the most[2], and could actually be seen as a regression for one-page documents, since in practice there'll usually be a couple of "DocStats" messages sent during the parsing of the first page. However, if the user zooms/rotates the document (which causes re-rendering), note that even a one-page document would start to benefit from these changes.
Another benefit of having the data available/cached in the API is that unless the document stats change during parsing, repeated `PDFDocumentProxy.stats`-calls will return *the same identical* object.
This is something that we can easily take advantage of in the default viewer, by now *only* reporting "documentStats" telemetry[3] when the data actually have changed rather than once per rendered page (again beneficial in longer documents).
---
[1] Furthermore, the maximium number of `StreamType`/`FontType` are `10` respectively `12`, which means that regardless of the complexity and page count in a PDF document there'll never be more than twenty-two "DocStats" messages sent; see 41ac3f0c07/src/shared/util.js (L206-L232)
[2] One example is the `pdf.pdf` document in the test-suite, where rendering all of its 1310 pages only result in a total of seven "DocStats" messages being sent from the worker-thread.
[3] Reporting telemetry, in Firefox, includes using `JSON.stringify` on the data and then sending an event to the `PdfStreamConverter.jsm`-code.
In that code the event is handled and `JSON.parse` is used to retrieve the data, and in the "documentStats"-case we'll then iterate through the data to avoid double-reporting telemetry; see https://searchfox.org/mozilla-central/rev/8f4c180b87e52f3345ef8a3432d6e54bd1eb18dc/toolkit/components/pdfjs/content/PdfStreamConverter.jsm#515-549
Given that all modern browsers now support `postMessage` transfers, and have for years, it no longer seems necessary for the PDF.js library to support using Workers unless the `postMessage` transfers functionality is available.
This patch is a follow-up to PR 11123, which made it impossible to *manually* disable `postMessage` transfers for performance reasons (since it increases memory usage), which hasn't caused any bug reports as far as I know.[1]
Hence we'll now only support *proper* Worker implementations, with fully working `postMessage` transfers, and fallback to using "fake" Workers otherwise.
---
[1] At the time of that PR we still "supported" IE, which is why this code was left intact.
On the main thread call `driver.log` and the message will output in the
terminal with the pdf id and the message.
I've been using this a lot when trying to find certain PDFs or logging
stats.
It shouldn't be necessary to iterate through *all* pages when using a non-default `spreadMode`, since we already know which page(s) should become visible.
This code is a left-over from the initial (local) implementation that resulted in PR 14112, however I forgot to clean-up some things such as e.g. this loop.
Also fixes an outdated comment, see PR 14204 which removed the mentioned data-structure.
This patch preserves the old behaviour of appending a `loadingIcon`-div to all pages that are not yet loaded/rendered. However, the actual `loadingIcon`-spinner (i.e. the `loading-icon.gif` image) will only be displayed on *visible* pages to improve performance.
To avoid having to iterate through all pages in the document, which doesn't seem like a good idea for a PDF document with thousands of pages, we use a combination of the currently visible *and* cached pages to toggle the `loadingIcon`-spinner.
This code is the last piece[1] of the viewer that's not using standard `class`es, and by converting this code we get rid of some now unneeded boilerplate code (slightly reducing the size of the *built* `web/viewer.js` file).
Note that while this code was originally imported from a separate repository, it was last sync-ed with upstream *five years* ago which is why this re-factoring should be OK as far as I'm concerned (and we've done some other clean-up since then as well).
---
[1] Technically the `web/debugger.js` file is left as well, however that code is first of all not bundled in the *built* `web/viewer.js` file and secondly it's not even loaded by default either.
There's obviously no guarantee that this will work in general, if the document is sufficiently corrupt, but it should hopefully be better than just throwing `InvalidPDFException` as currently happens.
Please note that, as is often the case with corrupt documents, it's somewhat difficult to know if we're rendering the document "correctly" with this patch[1]. In this case even Adobe Reader cannot open the document, which is always a good sign that it's *really* corrupt, however we're at least able to render *something* with this patch.
---
[1] Whatever "correct" even means when dealing with corrupt PDF documents, where often times different PDF viewers won't agree completely.
- it aims to fix https://bugzilla.mozilla.org/show_bug.cgi?id=931481;
- real space chars are pushed in the chunk but when there is an extra spacing, the next char position must be compared with the previous one;
- for example, an extra spacing can cancel a space so visually there are no space.
- First step to fix https://bugzilla.mozilla.org/show_bug.cgi?id=1737260;
- several interactive pdfs use the possibility to hide/show buttons to show different icons;
- render pushbuttons on their own canvas and then insert it the annotation_layer;
- update test/driver.js in order to convert canvases for pushbuttons into images.
Reporting telemetry, in Firefox, includes using `JSON.stringify` on the data and then sending an event to the `PdfStreamConverter.jsm`-code.
In that code the event is handled and `JSON.parse` is used to retrieve the data, and in the "pageInfo"-case we'll then proceed to ignore everything except *the first* such event; see https://searchfox.org/mozilla-central/rev/24fac1ad31fb9c6e9c4c767c6a7ff45d226078f3/toolkit/components/pdfjs/content/PdfStreamConverter.jsm#509-514
All-in-all, sending the "pageInfo" telemetry for each rendered page is thus unnecessary and this patch makes the viewer send it only *once* instead.
*This is a tentative patch, since we unfortunately cannot easily test it (as far as I can tell).*
In Firefox this (obviously) works as-is, but in Google Chrome the `markedContent` spans are inserted within the regular text-content (in the DOM) and with non-zero heights.
*This is a tentative patch, since I don't have the necessary hardware to test it.*
See https://developer.mozilla.org/en-US/docs/Web/CSS/text-size-adjust, which is currently ignored in Firefox.
It seems overall safer, and more future-proof, to simply add this to the *entire* `textLayer` rather than its individual elements.
With the previous patch, this helper function is no longer used and keeping it around will simply increase the size of the builds.
This removal is purposely done *separately*, to make it easy to revert the patch in the future if this helper function would become useful again.
This relies on the fact that `Set`s preserve the insertion order[1], which means that we can utilize an iterator to access the *first* stored view.
Note that in the `resize`-method, we can now move the visible pages to the back of the buffer using a single loop (hence we don't need to use the `moveToEndOfArray` helper function any more).
---
[1] This applies to `Map`s as well, although that's not entirely relevant here.
This small change will help validate an important part of the upcoming re-factoring, regarding the *correct* iteration of the `Set` in the `PDFPageViewBuffer.resize` method in particular.
Previously, when we created a shading pattern canvas we created it
as the same size as the page. This was good for caching if the same
pattern was used over and over again, but when lots of different
shadings are created that caused us to create many full page
canvases.
Instead of creating the full page canvses, create the canvas
as the same size as the current path bounding box. This reduces memory
consumption by a lot since most paths are pretty small. Also, in real world
PDFs it's rare for a shading (non shading fill) to be reused over and over again.
Bug 1721949 is an example where the same pattern is reused and it will be slightly
slower than before.
In `beginGroup` we create a new canvas that is the size of the
bounding box and we translate it to the offset. This means we don't need to
also apply the bounding box during `paintFormXObjectBegin`.
This improves #6961 quite a bit, but it still is missing the indention
in the ruler.
The `PDFPageViewBuffer`-code is very important for the correct function of the viewer, but it's currently not tested at all.
While the `PDFPageViewBuffer` is obviously intended to be used with `PDFPageView`-instances, it only accesses a couple of `PDFPageView` properties/methods and consequently it's fairly easy to unit-test this code with dummy-data.
These unit-tests should help improve our confidence in this code, and will also come in handy with other changes that I'm working on (regarding modernizing and re-factoring the `PDFPageViewBuffer`-code).
- it aims to fix https://bugzilla.mozilla.org/show_bug.cgi?id=1739502;
- when the target area was the current content area, everything was pushed in it instead of creating a new one (and consequently a new pageArea is created).
- the pdf shows an alignment issue on page 4:
- the hAlign is "center" but the subform was the width of its parent, so compute the real width of the subform with tb layout;
- there is an extra empty page at the end of the pdf:
- there is a subform with some hidden elements which are not rendered for now (since there is no plugged JS engine it isn't possible to draw them in changing their visibility).
- so in case a subform is empty and has no real dimensions (at least one is 0), we just consider it as empty.
We were incorrectly using the transform in the pattern before it had been
adjusted causing the pattern to be misplaced relative to the page.
Fixes: ShowText-ShadingPattern.pdf (already in corpus)
Fixes: #8111Fixes: #9243
Subfrom nomin displays even though it's subform is set to <occur max=-1 min=0>
If we look through specs of XFA 3.3 : https://www.pdfa.org/norm-refs/XFA-3_3.pdf
- The min attribute is used when processing a form that contains data. Regardless of the data at least this number of instances is included. It is permissible to set this value to zero, in which case the container is entirely excluded if there is no data for it.
However, in our case it doesn't happen, because we let our empty dataNode get through. Though by setting a clause:
- eliminate unmatched data with occur min=0
we are checking our empty data and sending it to uselessNode array where at the end it gets removed;
The way that we're currently handling the last-`id` is very old, and there's no longer any good reason to special-case things when only one thumbnail is visible.
Furthermore, we can also modernize the loop slightly by using `for...of` instead of `Array.prototype.some()` when checking for fully visible thumbnails.
Note how in `PDFPageViewBuffer.resize` we're manually iterating through the visible pages in order to build a Set of the visible page `id`s. By instead moving the building of this Set into the `getVisibleElements` helper function, as part of the existing parsing, this code becomes *ever so slightly* more efficient.
Furthermore, more direct access to the visible page `id`s also come in handy in other parts of the viewer as well.
In the `BaseViewer.isPageVisible` method we no longer need to loop through the visible pages, but can instead directly check if the pageNumber is visible.
In the `PDFRenderingQueue.getHighestPriority` method, when checking for "holes" in the page layout, we can also avoid some unnecessary look-ups this way.
Very short strings can narrowly miss the existing Bidi-detection threshold, leading to incorrect text-selection and copying behaviour.
In my testing, neither Adobe Reader or PDFium seem to handle copying "correctly" for this document. Hence it's not entirely clear to me that we actually want to fix this, since tweaking these heuristics can *obviously* cause regressions elsewhere (and our test coverage for RTL-text isn't exactly great).
Starting a new path will wipe out any of the current subpaths in the
current graphics state, so we should reset the min/maxes.
This makes a number of the bounding boxes smaller and reduces the number
of composed pixels. For the smask tests in the corpus, the number of
composed pixesl goes from 19,872,109 to 19,676,905. The difference is much
larger on other PDFs though.
*Sometimes I'll hopefully learn to optimize my code directly when writing it, rather than having to do multiple clean-up passes; sorry about the churn here!*
For most page layouts there won't be any "holes" in the visible pages (or thumbnails), and in those cases it'd obviously be preferable not having to repeat any checks of already rendered pages.
Rather than only checking the "distance" between the first/last pages, we can instead compare the theoretical number of pages (between first/last) with the actually visible number of pages instead. This way, we're able to better detect the "holes"-case and can skip unnecessary parsing in the common case.
I missed this one spot in PR 12870, when converting the other cases in the "keydown" event handler. However, given that it only matters in PresentationMode and/or when "page-fit" zooming is enabled, this oversight shouldn't have had any user-observable impact (but we should fix it nonetheless).
This is a follow-up to PR 10675, since there I completely overlooked that we also need to handle the case where a PDF document has *failed* to load when the "supportsRangedLoading" message is sent to the viewer.
It seems that issue 10301 was fixed by PR 13424, by combining the spans, however given that we don't have a lot of test coverage for RTL-text I figured that adding a simple reference test wouldn't hurt (rather than just closing the issue as WORKSFORME).
There were some links not working in some XFA files,I realized that the anchor tag that contains the link has an inline display and couldn't receive any height, solved this by adding a "position: absolute". Tested with two different files in Firefox Nightly and Chrome and now all links are working perfectly fine. Added reftest to avoid future regressions
Embedded JS in PDF keep throwing alert reagdring specific version of Acrobat (Spanish and version 5.0 or greater).
This happens because:
- JS in pdf is enabled
- PDF contains some unsupported features (e.g. XFA)
Alert come when app.formVersion = undefined || app.formVersion < 5.0
In pdf.js we were using FORM_VERSION = undefined. After researching based on https://opensource.adobe.com/dc-acrobat-sdk-docs/acrobatsdk/pdfs/acrobatsdk_jsapiref.pdf\#G4.1993509 and Acrobat DC we decided to go with the larger number to avoid unnecessary popups.
Through investigation we realise that VIEWER_VERSION should have same value - a number.
Due to all that, we implemented 21.00720099 as a value for both FORMS_VERSION and VIEWER_VERSION
The only reason for using a `DocumentFragment` in the first place, originally added in PR 8724, was to prevent errors in the `PDFPageView`-constructor. However, we should be able to simply make its `container`-option *optional* instead, since it's not being used for anything else in the class.
Note that pre-rendering still works correctly in my testing, and given that the `BaseViewer` keeps references to all `PDFPageView`-instances (via its `_pages` Array) it also shouldn't be possible to "lose" any pages/canvases this way.
Unfortunately there exist PDF documents where all pageLabels are empty strings, see e.g. http://www.cs.cornell.edu/~ragarwal/pubs/blk-switch.pdf (taken from an old issue), which result in the pageNumber-input being completely blank. That doesn't seem very helpful, and this patch simply extends the approach used to ignore pageLabels that are identical to standard page numbering.
In the GENERIC viewer, e.g. when dragging-and-dropping a new PDF document which automatically opens the outline, there can now be breaking errors in the `{BaseViewer, PDFThumbnailViewer}.#getScrollAhead` methods since there's no visible pages/thumbs during loading; sorry about the breakage!
This was a stupid oversight on my part, since the first/last visible pages have obviously already been rendered at the point when we're checking for any potential "holes" in the page layout.
While this will obviously not have any measurable effect on performance, we should nonetheless avoid doing completely unnecessary checks here.
This is a very old "issue", which has existed since essentially forever, and it affects all of the available scrollModes. However, in the recently added Page-mode it's particularily noticeable since we use a *simulated* scroll direction there.
When deciding what page(s) to pre-render, we only consider the current scroll direction. This works well in most cases, but can break down at the start/end of the document by trying to pre-render a page *outside* of the existing ones. To improve this, we'll thus *force* the scroll direction at the start/end of the document.
*Steps to reproduce:*
0. Open the viewer, e.g. https://mozilla.github.io/pdf.js/web/viewer.html
1. Enable vertical scrolling.
2. Press the <kbd>End</kbd> key.
3. Open the devtools and, using the DOM Inspector, notice how page 13 is *not* being pre-rendered.
This allows us to compose much smaller regions of soft
mask making them much faster. This should also allow
for further optimizations in the pattern code.
For example locally I see issue #6573 go from 55s
to 5s with this change.
Fixes#6573
The old method of handling soft masks had a number of issues where the temporary
drawing canvas and the suspended main canvas could get out of sync
(e.g. mismatched save/restores or clip state) or we could end up compositing at
the wrong time. A good example of things getting out sync is the reduced test
case in #9017.
To fix this I've changed two big things:
1) Duplicate all the needed graphics state from the temporary canvas to the
suspended main canvas. This ensure the canvases stay in sync so that when we
switch back to the main canvas the graphics state stack is the same
(e.g. transforms, clip paths).
2) Immediately composite after each drawing operation. This ensures that if
there's an active clip region that we'll still be able to composite the correct
portions of the canvas. Note: This solution could be avoided by using
getImageData and putImageData since those ignore clipping region, but this is
very very slow. Note2: I also think the old way of only compositing at the end
of the soft mask is incorrect and can lead to wrong colors if drawing over the
same region, but in practice this doesn't seem to matter much.
Fixes: #5781Fixes: #5853Fixes: #7267Fixes: #7891Fixes: #8403Fixes: #8624Fixes: #12798Fixes: #13891Fixes: #9017 (reduced test case)
Fixes: https://bugzilla.mozilla.org/show_bug.cgi?id=1703683
With ESLint 8 we should now finally be able to start using modern `class` features, see https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Classes/Public_class_fields and https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Classes/Private_class_fields
However, while both ESLint and Acorn now support this, it unfortunately turns out that Escodegen (which we use during building) still lack the necessary support. Looking at https://github.com/estools/escodegen there's not been any updates since last year, and there's also open PRs adding support for these new `class` features.
To avoid blocking usage of these `class` features in the PDF.js code-base, in particular *private* fields/methods, this patch thus proposes that we (hopefully temporarily) switch to an `escodegen` fork that has the necessary support; please see https://www.npmjs.com/package/@javascript-obfuscator/escodegen
While I have no reason to doubt the security of the `escodegen` fork, this patch nonetheless pins the version number. Furthermore, I've also diffed the output of the two `.js`-files in this forked package against the original files without finding anything that looks immediately "dangerous".
With ResetForm-action support added in PR 14083, there's a regression in the `issue12716` test-case. More specifically the border around the "Clear Form"-link is now rendered *twice*, once in the canvas via the appearance-stream and once in the annotationLayer via the border-data.
This looks slightly weird, and was most likely not intended, which is why this patch suggests that we ignore the border in the annotationLayer when an appearance-stream exists.
Apparently Node.js has added *global* `URL.createObjectURL` support, but not done the same thing for `Blob`. Hence we also need to check for the availability of `Blob` in the `createObjectURL` helper function, and it's probably a good idea to also update `examples/node/pdf2svg.js` to work-around this until these changes reach an official PDF.js release.
Trying to render these Annotation-types, when the borderWidth is `0`, causes a "hairline" border to appear. If these Annotations included an appearance stream, as they are supposed to, this wouldn't have happened and the simplest solution here seem to be to just ignore these particular Annotations.
- PR #13257 fixed a lot of issues but not all and this patch aims to fix almost all remaining issues.
- the idea in this new patch is to compare position of new glyph with the last position where a glyph has been drawn;
- no space are "drawn": it just moves the cursor but they aren't added in the chunk;
- so this way a space followed by a cursor move can be treated as only one space: it helps to merge all spaces into one.
- to make difference between real spaces and tracking ones, we used a factor of the space width (from the font)
- it was a pretty good idea in general but it fails with some fonts where space was too big:
- in Poppler, they're using a factor of the font size: this is an excellent idea (<= 0.1 * fontSize implies tracking space).
*Please note:* This is a tentative patch, since I don't have the necessary a11y-software to actually test it.
To avoid having to add a new API-method just for a single string, I figured that adding the new property to the existing `documentInfo`-data (accessed via `PDFDocumentProxy.getMetadata` in the API) will hopefully be deemed acceptable.
In these cases there's no good reason, in my opinion, to duplicate the `shadow`-lines since that unnecessarily increases the risk of simple typos (see the previous patch).
With this typo the shadowing doesn't actually work, which causes these checks to be unnecessarily repeated. In this particular case it didn't have a significant performance impact, however we should definately fix this nonetheless.
Looking at the code, I do have to agree with the point made in issue 12731 about it being unexpected/unhelpful that the `PDFFindController.executeCommand`-method isn't directly usable with the "find"-event.
The reason for it being this way is, as so often, for historical reasons: The `executeCommand`-method was added (just) prior to the introduction of the `EventBus` in the viewer.
Obviously we cannot simply change the existing `PDFFindController.executeCommand`-method, since that'd be a breaking change in code which has existed for over five years.
Initially I figured that we could simply add a new method in `PDFFindController` that'd accept the state from the "find"-event, however after thinking about this and looking through the use-cases in the default viewer I settled on a slightly different approach: Let the `PDFFindController` just listen for the "find"-event (on the `EventBus`-instance) directly instead, which also removes one level of (unneeded) indirection during searching in the default viewer.
For GENERIC builds of the PDF.js library, the old `PDFFindController.executeCommand`-method is still available with a deprecation warning.
Many years ago now there were some `Promise` implementations that had issues resolving with an *implicitly* `undefined` value. That should no longer be the case, and we've not been using the `Promise.resolve(undefined)` format for a long time, hence this patch fixes the few remaining cases.
Having recently worked with this code, in PR 14096 (and indirectly in PR 14112), I happened to notice a pre-existing issue with spreadModes at higher zoom levels.
The `PDFRenderingQueue` code was written back when the viewer only supported "normal" vertical scrolling, and some edge-cases related to spreadModes are thus not perfectly supported. Depending on the zoom level, it's possible that there are "holes" in the currently visible page layout, and those pages will not be pre-rendered as you'd expect.
*Steps to reproduce:*
0. Open the viewer, e.g. https://mozilla.github.io/pdf.js/web/viewer.html
1. Enable vertical scrolling.
2. Enable the ODD spreadMode.
3. Scroll down, such that both pages 1 and 3 are visible.
4. Zoom-in until *only* page 1 and 3 are visible.
5. Open the devtools and, using the DOM Inspector, notice how page 2 is *not* being pre-rendered despite all surrounding pages being rendered.
If a PDF included an embedded TrueType font whose preferred character
map (cmap) was in "format 2", the code would select that character map
and then refuse to read it because of an unsupported format, thus
causing the characters not to be rendered. This commit implements
support for format 2 as described at the link below.
https://developer.apple.com/fonts/TrueType-Reference-Manual/RM06/Chap6cmap.html
With the previous commit, both of the `PDFViewer` and `PDFSinglePageViewer` clases are now small/simple enough that it no longer seems necessary to keep them in separate files.
This implements a new Page scrolling mode, essentially bringing (and extending) the functionality from `PDFSinglePageViewer` into the regular `PDFViewer`-class. Compared to `PDFSinglePageViewer`, which as its name suggests will only display one page at a time, in the `PDFViewer`-implementation this new Page scrolling mode also support spreadModes properly (somewhat similar to e.g. Adobe Reader).
Given the size and scope of these changes, I've tried to focus on implementing the basic functionality. Hence there's room for further clean-up and/or improvements, including e.g. simplifying the CSS/JS related to PresentationMode and implementing easier page-switching with the mouse-wheel/arrow-keys.
This patch (slightly) simplifies a couple of `onProgress` and `onUnsupportedFeature` call-sites.
Finally, while unrelated, also removes some unnecessary `return undefined;` statements (PR 11601 follow-up).
For Circle, Square, and Polygon Annotations it's currently only possible to toggle the associated PopupAnnotation by clicking on its border. Depending on the border width, and also the current zoom-level in the viewer, that can make interacting with certain Annotations *practically* impossible (which is the case in issue 14107).
Hence, in order to improve this, change the "fill"-property of the SVG element in the annotationLayer to make the *entire* element part of the click/mouse-over target.
*Please note:* Given that this is a viewer-related issue, there's no simple way to test this as far as I can tell.
This is unfortunately *yet another* bug in the `preEvaluateFont`-implementation, and I've lost count of the number of times I've had to tweak this code over the years :-(
I really cannot help thinking that PR 4423 was way too simplistic, since it missed a bunch of cases that leads to broken font rendering in many PDF documents.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1734802
- the exact sentence from the spec:
"The token SOLIDUS (a slash followed by no regular characters) introduces a unique valid name defined by the empty sequence of characters."
- so just remove the warning.
Note that PR 14049 removed this, since https://github.com/mozilla/pdf.js/pull/14049#discussion_r716245518 claimed that it's not necessary anymore. Unfortunately that, in my testing on Windows, actually re-introduced exactly the issue described in the comment; more specifically once the *last* character has been entered in the comb-field it's now again incorrectly scrolled (with the first character being invisible) until focus is lost.
This can be tested with e.g. `f1040.pdf`, see page 2, from the test-suite.
With a recent addition to the HTML specification, the internal structured clone algorithm used in browsers is (or will be, once it's implemented) *directly* accessible to JavaScript; please see https://developer.mozilla.org/en-US/docs/Web/API/WindowOrWorkerGlobalScope/structuredClone
Hence we'll *eventually* not need to maintain our own structured clone functionality in the `LoopbackPort`-class in the API, however for the time being we'll feature detect `structuredClone` and fallback to the existing PDF.js implementation.
Given that https://bugzilla.mozilla.org/show_bug.cgi?id=1722576 has landed in Firefox 94, we should no longer need the manually implemented `cloneValue`-functionality in MOZCENTRAL builds. Note also that in the Firefox built-in PDF Viewer it's not possible for users to *easily* disable workers, which should further reduce the risk of these changes.
This patch helps reduce some duplication, given that we now have a few essentially identical `addLinkAttributes` call-sites in the code-base.
To prevent runtime errors in the Annotation/XFA-layer code, we'll warn if a custom/incomplete `PDFLinkService` is being used (limited to GENERIC builds).
Please note that we (obviously) don't want to unconditionally pre-render more than one page all the time, since that could very easily lead to overall worse performance in some documents.[1]
However, when spreadModes are enabled it does make sense to attempt to pre-render both of the pages of the next/previous spread.
---
[1] Since it may cause pre-rendering to unnecessarily compete for parsing resources, on the worker-thread, with "regular" rendering.
Note how both the annotationLayer and the document outline will apply various URL-related options when creating the link-elements.
For consistency the `xfaLayer`-rendering should obviously use the same options, to ensure that the existing options are indeed applied to all URLs regardless of where they originate.
Given that `NodeList`s can be iterated using `for..of` we can use that instead, since it's a little bit nicer and easier to read than the `Array.prototype.forEach` format.
- it aims to fix#12721.
- Thanks to PR #14023, we've now the fieldObjects in the annotation layer so we can easily map fields names on their id if needed.
- Reset values in the storage, in the JS sandbox and in the visible html elements.
In the referenced bug, the embedded fonts contain custom CMap-data that only include strings. Note how for embedded composite TrueType fonts we're using the CMap-data when building the glyph mapping, and currently we end up with a completely empty map because the code expects only CID *numbers*.
Furthermore, just fixing the glyph mapping alone isn't sufficient to fully address the bug, since we also need to consider this "special" kind of CMap-data when looking up glyph widths.
Having recently worked with, and reviewed patches touching, this code it seemed that it's probably not a bad idea to move that functionality into `createValidAbsoluteUrl` as new options instead.
For the `addDefaultProtocolToUrl` functionality in particular, the existing helper function was not only moved but slightly improved as well. Looking at the code, I realized that there's a small risk that it would incorrectly match a *relative* URL-string too.
With these changes, the `createValidAbsoluteUrl` call-sites in the `src/core/`-code can be simplified a little bit.
*Please note:* This patch may, indirectly, change the format of the `unsafeUrl`-property returned with relevant Annotations and OutlineItems; hence the `api-minor` tag.
However, I'd argue that it's actually more correct this way since the whole purpose of `unsafeUrl` is/was to return the URL data as-is without any parsing done.
- it aims to fix https://bugzilla.mozilla.org/show_bug.cgi?id=1716758;
- some buttons have a JS action with the pattern `app.launchURL(...)` (or similar) so extract when it's possible the url and generate a <a> element with the href equals to the found url;
- pdf.js already had some code to handle that so this patch slightly refactor that.
- it aims to fix#14021;
- the N dict is empty here so just create a default one;
- it implies that the checked checkbox has no appearance so create a default one too in order to print it;
- in the pdf in the issue, a checked box is not printed because it has no default appearance so we need to guess its appearance from its state.
In order to implement this, we utilize the existing `bidi` function to infer the text-direction of /T and /Contents entries. While this may not be perfect in cases where one PopupAnnotation mixes LTR and RTL languages, it should work well enough in most cases.
To avoid having to add *two new* properties in lots of annotations, supplementing the existing `title`/`contents`-properties, this patch instead re-factors the existing code such that the properties are replaced by Objects (containing `str` and `dir`).
*Please note:* In order avoid breaking existing third-party implementations, `GENERIC`-builds of the PDF.js library will still provide the old `title`/`contents`-properties on annotations returned by `PDFPageProxy.getAnnotations`.
- In the pdf in issue #14071, some select fields don't contain any values;
- the corresponding node has a bindItems and a bind elements and _bindItems function was just not called.
In particular the `_processStreamMessage`-method is a bit cumbersome to read, given the way that the current streamController/streamSink is accessed, which we can improve with a couple of local variables.
After PR 11601, the `paintJpegXObject` operator is no longer used for anything. While I don't think we can just remove it, and essentially leave a "hole" in the `OPS` structure, we should at least mark it as explicitly unused to aid readability/maintainability of the code.
This replaces direct `document.getElementsByName` lookups with a helper method which:
- Lets the AnnotationLayer use the data returned by the `PDFDocumentProxy.getFieldObjects` API-method, such that we can directly lookup only the necessary DOM elements.
- Fallback to using `document.getElementsByName` as before, such that e.g. the standalone viewer components still work.
Finally, to fix the problems reported in issue 14003, regardless of the code-path we now also enforce that the DOM elements found were actually created by the AnnotationLayer code.
With these changes we'll thus be able to update form elements on all visible pages just as before, but we'll additionally update the AnnotationStorage for not-yet-rendered elements thus fixing a pre-existing bug.
*This is similar to the "isSymbolicFont"-property, which is no longer exported by default after PR 11777.*
Both "isMonospace" and "isSerifFont" are internal properties, used during font parsing and building of the glyph mapping on the worker-thread.
However both of these properties are completely unused on the main-thread and/or in the API, and accessing them they will now require setting the `fontExtraProperties`-option when calling `getDocument`.
Currently any AppOptions set using e.g. the "webviewerloaded" event listener can/will by default be overridden when the Preferences are read.
To avoid that happening the "disablePreferences"-option can be used, however unless it's been explicitly set all non-default AppOptions will be silently ignored. This patch thus attempts to improve the current situation somewhat, for third-party implementations, by logging a warning in the console when this happens.
With this patch we'll ensure that only valid absolute URLs can be used in XFA documents, similar to the existing validation done for "regular" PDF documents.
Furthermore, we'll also attempt to add a default protocol (i.e. `http`) to URLs beginning with "www." in XFA documents as well; this on its own is enough to fix https://bugzilla.mozilla.org/show_bug.cgi?id=1731240
Rather than re-computing this value in a number of different places throughout the code-base[1], we can expose this in the API via the existing `PixelsPerInch`-structure instead.
There's also been feature requests asking for the old `CSS_UNITS` viewer constant to be made accessible, such that it could be used in third-party implementations.
I suppose that it could be argued that it's somewhat confusing to place a unitless property in `PixelsPerInch`, however given that the `PDF_TO_CSS_UNITS`-property is defined strictly in terms of the existing properties this is hopefully deemed reasonable.
---
[1] These include:
- The viewer, with the `CSS_UNITS` name.
- The reference-tests.
- The display-layer, when rendering images; see PR 13991.
Given the simplicity of this functionality, we can move it from the default viewer and into the `BaseViewer` class instead. This way, it's possible to support more scripting functionality in the standalone viewer components; please see PR 14038.
Please note that I purposely went with `increaseScale`/`decreaseScale`-method names, rather than using "zoom", to better match the existing `currentScale`/`currentScaleValue` getters/setters that's being used in the `BaseViewer` class.
Currently we only exclude /Encoding entries that also contains a /Differences array, which is the cause of the text-selection problem in the referenced issue.
In order to address this we'll now also exclude /Encoding entries that contain one of the predefined *named* encodings, and no longer require that it also contains a /Differences array.
*Please note:* This patch cases a small "regression" in the `bug1130815-text` test-case, however this is actually an improvement when compared with Adobe Reader and PDFium (in Google Chrome).
- it aims to fix https://bugzilla.mozilla.org/show_bug.cgi?id=1719148;
- JS can set a property for a non-rendered annotation using the annotationStorage but the other unset default properties must be used when the annotation is finally rendered;
- so this patch just adds the properties already set in the annotationStorage to the default value.
This reduces some unnecessary duplication, since we currently have essentially the same code in a handful of places in the `Font.fallbackToSystemFont`-method.
Rather than forcing the "regular" `EventBus` to check and handle `isInAutomation` for every `dispatch` call, we can take advantage of subclassing instead.
Hence this PR introduces a new `AutomationEventBus` class, which extends `EventBus`, and is used by the default viewer when `isInAutomation === true`.
In this particular case the `CMap`-data that we create contains only numbers, but no strings, which causes `PartialEvaluator.readToUnicode` to create a ToUnicode-map with only empty strings.
*Please note:* This is yet another case where I don't know if it's necessarily the best and most correct solution, but it does fix the referenced issue.
- it aims to fix https://bugzilla.mozilla.org/show_bug.cgi?id=1722036.
- AS and V should share the same value for checkbox: it's at least what the specs say;
- the pdf in the above bug opens correctly in Acrobat so it likely means that AS is chosen over V.
Originally the library/viewer didn't support forms, and the hiding of the Download-buttons (when new documents are opened) didn't really matter all that much. Hence the simplest solution, at the time, was to hide the Download-buttons since we use the URL as a fallback when downloading data in the GENERIC `DownloadManager`.
Nowadays we obviously want to support saving of forms in the GENERIC viewer, regardless of how the document was opened, which could thus *potentially* lead to the fallback download-URL being wrong.
In order to be able to show the Download-buttons unconditionally, this patch slightly re-factors the viewer to track the download-URL *separately* to prevent any issues there.
*Please note:* As mentioned in the issue, the ViewBookmark-buttons are specific to the initial URL when the viewer is first opened. Hence they (still) don't make sense when a new document has been opened, since it's then impossible to obtain a usable link to the *currently* active document.
By using CSS variables to set the width of the zoom dropdown, we can simplify both the relevant CSS and JS code. This will not only improve overall maintainability of this code, but should also make it (slightly) easier for third-party users that want to customize the width.
Note in particular that by having the code in `Toolbar._adjustScaleWidth` lookup the values of the CSS variables, we no longer need to worry about keeping hard-coded values up-to-date with the CSS rules.
- it aims to fix#14024.
- this patch adds an attribute `acroformExportValue` to the HTML input in order to set the checked attribute in taking into account the exportValue for the checkboxes with the same name.
*Please note:* All of this feels very handwavy, but at least it passes all tests locally. Hopefully we have enough tests for this part of the font code.
For non-embedded composite standard fonts with an "incomplete" /CIDToGIDMap, we'll now fallback to an *explicitly defined* /ToUnicode map even when that one happens to be an /Identity-H or /Identity-V map.
The `Font.fallbackToSystemFont` method is unfortunately getting more and more special-cases, however that might be unavoidable given all the weird non-embedded fonts found in the wild :-(
While these types apparently makes sense in TypeScript environments, we really don't want to extend the *public* API by simply exporting the relevant classes directly in `src/pdf.js` (since they should never be called/initialized manually).
Please see e.g. issue 12384 where this was first requested, and note that a possible work-around was also provided there. This patch simply implements that work-around[1], which will hopefully be helpful to TypeScript users.
---
[1] Based on the discussion in PR 13957, the two previous patches appear to be necessary for this to actually work.
*This fixes something that I noticed, having recently looked at both the `Lexer.getObj` and `writeValue` code.*
Please note that I unfortunately don't have an example of a form where saving fails without this patch. However, given its overall simplicity and that unit-tests are added, it's hopefully deemed useful to fix this potential issue pro-actively rather than waiting for a bug report.
At this point one might, and rightly so, wonder if there's actually any real-world PDF documents where a `null` value is being used?
Unfortunately the answer is *yes*, and we have a couple of examples in the test-suite (although none of those are related to forms); please see: `issue1015`, `issue2642`, `issue10402`, `issue12823`, `issue13823`, and `pr12564`.
While some of the output looks worse to my eye, this behavior more
closely matches what I see when I open the PDFs in Adobe acrobat.
Fixes: #4706, #9713, #8245, #1344
Currently it's possible to accidentally, e.g. by simply copy-and-pasting from an existing test-case, add an unnecessary `"link": true`-entry for locally available PDF files.
This leads to inconsistencies in the manifest file, and doesn't feel like a great developer experience. However we can easily fix it by having `verifyManifestFiles` fail in this situation, and doing so actually turned up a couple of existing cases.
Similar to other viewer components, e.g. the `PDFFindBar` and `PDFPresentationMode`, there's no need to create a `PDFHistory`-instance when it's not going to be used.
The `MessageHandler`-implementation already handles either of these callbacks being undefined, hence there's no particular reason (as far as I can tell) to add no-op functions here.
Also, in a couple of `MessageHandler`-methods, utilize an already existing local variable more.
Following the STR in the issue, this patch reduces the number of `PartialEvaluator.getTextContent`-related `postMessage`-calls by approximately 78 percent.[1]
Note that by enforcing a relatively low value when batching text chunks, we should thus improve worst-case scenarios while not negatively affect all `textLayer` building.
While working on these changes I noticed, thanks to our unit-tests, that the implementation of the `appendEOL` function unfortunately means that the number and content of the textItems could actually be affected by the particular chunking used.
That seems *extremely* unfortunate, since in practice this means that the particular chunking used is thus observable through the API. Obviously that should be a completely internal implementation detail, which is why this patch also modifies `appendEOL` to mitigate that.[2]
Given that this patch adds a *minimum* batch size in `enqueueChunk`, there's obviously nothing preventing it from becoming a lot larger then the limit (depending e.g. on the PDF structure and the CPU load/speed).
While sending more text chunks at once isn't an issue in itself, it could become problematic at the main-thread during `textLayer` building. Note how both the `PartialEvaluator` and `CanvasGraphics` implementations utilize `Date.now()`-checks, to prevent long-running parsing/rendering from "hanging" the respective thread. In the `textLayer` building we don't utilize such a construction[3], and streaming of textContent is thus essentially acting as a *simple* stand-in for that functionality.
Hence why we want to avoid choosing a too large minimum batch size, since that could thus indirectly affect main-thread performance negatively.
---
[1] While it'd be possible to go even lower, that'd likely require more invasive re-factoring/changes to the `PartialEvaluator.getTextContent`-code to ensure that the batches don't become too large.
[2] This should also, as far as I can tell, explain some of the regressions observed in the "enhance" text-selection tests back in PR 13257.
Looking closer at the `appendEOL` function it should potentially be changed even more, however that should probably not be done here.
[3] I'd really like to avoid implementing something like that for the `textLayer` building as well, given that it'd require adding a fair bit of complexity.
This is essentially a simplified version of the code that's used in `PDFPageView`, which will hopefully reduce the number of issues opened specifically about blurry rendering.
However, note that *ideally* users should base their implementations on the `components/` examples rather than using the API directly (the "viewer components" already support HiDPI-screens).
While these changes will obviously not have a significant effect on overall memory usage, it cannot hurt as far as I'm concerned. This patch makes the following changes:
- Clear out `_textDivProperties` once rendering is done, since those properties are only necessary to keep alive when *enhanced* text-selection is being used.
- Reduce the size of the `_textDivProperties`-entries by default, since a majority of the properties are only relevant when *enhanced* text-selection is being used.
The old `update`-signature started to annoy me back when I added optional content support to the viewer, since we're (often) forced to pass in a bunch of arguments that we don't care about whenever these methods are called.
This is tagged `api-minor` since `PDFPageView` is being used in the `pageviewer` component example, and it's thus possible that these changes could affect some users; the next commit adds fallback handling for the old format.
In the referenced PDF document the /Contents stream contains MarkedContent-operators, however no optional content dictionary exists; according to [the specification](https://www.adobe.com/content/dam/acom/en/devnet/pdf/pdfs/PDF32000_2008.pdf#G7.3883825):
> Null values or references to deleted objects shall be ignored. If this entry is
not present, is an empty array, or contains references only to null or deleted
objects, the membership dictionary shall have no effect on the visibility of
any content.
In the PDF document some of the glyphs have bogus `differences`-entries[1] that cannot be resolved to valid glyph names, thus causing the glyph mapping to fail.
My initial idea was to use a similar approach as in the `PartialEvaluator._simpleFontToUnicode`-method, to extract the charCodes from those entries, however it turned out that that didn't actually help in this case (the mapping was still wrong).
To fix this I'm thus proposing that we fallback to the /ToUnicode map when no other useable data exists (e.g. no post-table), since it *hopefully* shouldn't make things any worse than leaving parts of the glyph map empty (which currently happens).
---
[1] As can be seem below, some of the entries are completely normal while others are non-standard:
```
Differences (array)
0 = 65
1 = /g5167
2 = /space
3 = /g11927
4 = /g17737
5 = /g11540
6 = /g2180
7 = /K
8 = /P
9 = /two
10 = /zero
11 = /one
12 = /five
13 = /four
14 = /g6932
15 = /g7246
16 = /g1691
17 = /g2343
18 = /g14792
19 = /g3325
20 = /g4280
21 = /g20383
22 = /g18166
23 = /g16988
24 = /g17943
25 = /g19223
26 = /g10830
27 = 97
28 = /g982
29 = /g1226
30 = /g5059
31 = /g2677
32 = /g1042
33 = /g11568
34 = /L
35 = /three
36 = /seven
37 = /g2364
38 = /g12063
39 = /g5356
40 = /g2173
41 = /g17877
42 = /g7273
43 = /g7647
44 = /g7224
45 = /g19327
46 = /g5054
47 = /g2342
48 = /g10136
49 = /g6856
50 = /g13381
51 = /g7257
52 = /g12093
53 = /g2359
```
- aims to fix https://bugzilla.mozilla.org/show_bug.cgi?id=1720179
- in some pdfs the XFA array in AcroForm dictionary doesn't contain an entry for 'datasets' (which contains saved data), so basically this patch allows to overwrite the AcroForm dictionary with an updated XFA array when doing an incremental update.
- when some named nodes in the template don't have their counterpart in datasets we create some nodes: the main node mustn't belong to the datasets namespace because it doesn't make sense and Acrobat Reader isn't able to read pdf with such nodes.
- so created nodes under a datasets node have a namespaceId set to -1 and consequently when serialized no namespace prefix will appear.
There's a fair number of regular expressions througout the code-base which are slightly more verbose than strictly necessary, in particular:
- We have a lot of regular expressions that use `[0-9]` explicitly, and those can be simplified to use `\d` instead.
- We have one instance of a regular expression containing a `A-Za-z0-9_` sequence, which can be simplified to use `\w` instead.
Despite its name, the fonts in ItcSymbol-family are "regular" fonts and not Symbol ones. However, given that the font name contains the word "Symbol" we ended up picking the wrong code-path in the `Font.fallbackToSystemFont`-method.
*Please note:* While this patch ensures that the text becomes readable, by falling back a standard font, the rendering will obviously not be perfect. However, that's the PDF generators "fault" since non-embedded fonts cannot be guaranteed to render correctly in all environments.
Given that the Fetch API is normally being used now, these changes are probably less important now than they used to be. However, given that it's simple enough to implement this I figured why not just fix issue 9883 (better late than never I suppose).
Testing:
- delete the pdf file while the initial request is inflight
- delete the pdf file after the initial request has finished
Repeat for a small file and large file, exercising both one-off and
chunked transports.
This patch changes the `PDFDocumentLoadingTask.destroy`-method and the `_fetchDocument`-function to be `async`, which slightly simplifies the relevant code.
Furthermore, remove the catch-handler from the `WorkerTransport.getPageIndex`-method since it's no longer needed. Given that the `MessageHandler` is nowadays wrapping every possible Exception, it's no longer necessary to try and re-wrap the reason here.
While e.g. the `simpleviewer` and `singlepageviewer` examples work, since they're based on the `BaseViewer`-class, the standalone `pageviewer` example currently doesn't support either XFA- or StructTree-layers. This seems like an obvious oversight, which can be easily addressed simply by exporting the necessary functionality through `pdf_viewer.component.js`, similar to the existing Text/Annotation-layers.
While working on, and testing, these changes I happened to notice a number of smaller things that's also fixed in this patch:
- Ensure that `XfaLayerBuilder.render` always have a *consistent* return type, to prevent possible run-time failures in `PDFPageView`; PR 13908 follow-up.
- Change the order of the options in the `XfaLayerBuilder`-constructor to agree with the parameter order in the `DefaultXfaLayerFactory.createXfaLayerBuilder`-method.
- Add a missing `textHighlighterFactory`-option, in the JSDocs for the `PDFPageView`-class.
- A couple of small tweaks in the `TextLayerBuilder.render`-method: Re-use an existing Array rather than creating a new one, and replace an `if` with optional chaining instead.
*Please note:* For now XFA-support is currently disabled by default, similar to the regular viewer.
While running the unit-tests with some logging statements added to this code, I noticed that `PasswordException` was missing from the list of potential Errors that could be passed to the `wrapReason` function.
A lot of the code in this getter has existed ever since the initial PresentationMode-implementation was first added all the way back in PR 1938 (which is nine years ago now).
At this point in time however, there's now a simpler way detect if a browser supports the FullScreen API and we should thus be able to simplify this getter; please refer to https://developer.mozilla.org/en-US/docs/Web/API/Document/fullscreenEnabled#browser_compatibility
This command was added all the way back when basic CI-support was first introduced (using Travis at the time), however it's never really intended to be used e.g. for local development.
By having a `npm test`-command listed in the `package.json` file, there's a very real risk that someone unfamiliar with the code-base would only run that one and thus miss all the other (more important) test-suites[1].
Hence this patch which removes the `npm test`-command, and instead simply calls the relevant gulp-task[2] directly in the GitHub Actions configuration.
---
[1] Which consist of the unit-tests (run in browsers), the font-tests (potentially), the reference-tests, and the integration-tests.
[2] Which is also renamed slightly, to better fit its current usage.
Something that I *just* realized is that while PR 13854 fixed an issue as reported, it could still cause bugs in other similarily broken documents since we'll not insert a matching endMarkedContent-operator in the operatorList.
Currently, in the `PartialEvaluator`, we only support Optional Content in Form-/XObjects. Hence this patch adds support for Image-/XObjects as well, which looks like a simple oversight in PR 12095 since the canvas-implementation already contains the necessary code to support this.
The Viewer API definitions do not compile because of missing imports and
anonymous objects are typed as `Object`. These issues were not caught
during CI because the test project was not compiling anything from the
Viewer API.
As an example of the first problem:
```
/**
* @implements MyInterface
*/
export class MyClass {
...
}
```
will generate a broken definition that doesn’t import MyInterface:
```
/**
* @implements MyInterface
*/
export class MyClass implements MyInterface {
...
}
```
This can be fixed by adding a typedef jsdoc to specify the import:
```
/** @typedef {import("./otherFile").MyInterface} MyInterface */
```
See https://github.com/jsdoc/jsdoc/issues/1537 and
https://github.com/microsoft/TypeScript/issues/22160 for more details.
As an example of the second problem:
```
/**
* Gets the size of the specified page, converted from PDF units to inches.
* @param {Object} An Object containing the properties: {Array} `view`,
* {number} `userUnit`, and {number} `rotate`.
*/
function getPageSizeInches({ view, userUnit, rotate }) {
...
}
```
generates the broken definition:
```
function getPageSizeInches({ view, userUnit, rotate }: Object) {
...
}
```
The jsdoc should specify the type of each nested property:
```
/**
* Gets the size of the specified page, converted from PDF units to inches.
* @param {Object} options An object containing the properties: {Array} `view`,
* {number} `userUnit`, and {number} `rotate`.
* @param {number[]} options.view
* @param {number} options.userUnit
* @param {number} options.rotate
*/
```
- Use `Node.TEXT_NODE` rather than a magical constant, in `TextHighlighter._convertMatches`, to improve readability. According to MDN, this has been supported since "forever": https://developer.mozilla.org/en-US/docs/Web/API/Node/nodeType#browser_compatibility
- Remove the `pageIdx`-property, on `TextLayerBuilder`-instances, since the re-factoring in PR 13908 meant that it's now unused.
- Remove the `matches`-property, on `TextLayerBuilder`-instances, since the re-factoring in PR 13908 meant that it's now unused.
Generalizing, and documenting, the `PDFHistory`-implementation as part of the web-interfaces doesn't seem entirely necessary and in hindsight I'm not entirely sure why we need it since:
- The `PDFHistory` implementation is/was written specifically for the default viewer use-case, which is why e.g. the `simpleviewer` component example isn't using it. (While it *could* be used there, it'd need to be manually created/initialized correctly.)
- There's only *one* `PDFHistory`-implementation present (and no other viewer-component will fail without it being available), as opposed to the other web-interfaces documented in this file.
- The `PDFHistory` implementation is not even usable with e.g. the `pageviewer` component example, since it (obviously) requires a complete viewer to work. (This is in contrast to e.g. `IPDFTextLayerFactory` and `IPDFAnnotationLayerFactory`.)
*This is done separately from the previous patch, to make it easier to revert these changes once they've been included in a couple of releases.*
Please note that because these two options are mutually exclusive, which is a large part of the reason for the previous patch, it's not guaranteed that the fallback-values will always be correct in every situation (but it's the best that we can do).
*This is a follow-up to PRs 13867 and 13899.*
This patch is tagged `api-minor` for the following reasons:
- It replaces the `renderInteractiveForms`/`includeAnnotationStorage`-options, in the `PDFPageProxy.render`-method, with the single `annotationMode`-option that controls which annotations are being rendered and how. Note that the old options were mutually exclusive, and setting both to `true` would result in undefined behaviour.
- For improved consistency in the API, the `annotationMode`-option will also work together with the `PDFPageProxy.getOperatorList`-method.
- It's now also possible to disable *all* annotation rendering in both the API and the Viewer, since the other changes meant that this could now be supported with a single added line on the worker-thread[1]; fixes 7282.
---
[1] Please note that in order to simplify the overall implementation, we'll purposely only support disabling of *all* annotations and that the option is being shared between the API and the Viewer. For any more "specialized" use-cases, where e.g. only some annotation-types are being rendered and/or the API and Viewer render different sets of annotations, that'll have to be handled in third-party implementations/forks of the PDF.js code-base.
Moves the logic out of TextLayerBuilder to handle
highlighting matches into a new separate class `TextHighlighter`
that can be used with regular PDFs and XFA PDFs.
To mimic the current find functionality in XFA, two arrays
from the XFA rendering are created to get the text content
and map those to DOM nodes.
Fixes#13878
Current pdf_viewer definitions result in errors like the following when
trying to use them in a ts project:
[error] TypeScript error
node_modules/.pnpm/pdfjs-dist@2.10.377/node_modules/pdfjs-dist/web/pdf_viewer.d.ts:1:15
- error TS2691: An import path cannot end with a '.d.ts' extension.
Consider importing 'pdfjs-dist/types/web/pdf_viewer.component.js'
instead.
1 export * from "pdfjs-dist/types/web/pdf_viewer.component.d.ts";
Import/export statements in typescript should not include file extensions.
Currently a `TESTING = true` environment variable will *always* take precedence in the various build-tasks, and there's no way to explicitly disable it for a particular build. That's clearly an oversight on my part, however it's easy enough to fix this; sorry about breaking this!
The `loadAndEnablePDFBug` helper function, in `web/app.js`, can be simplified a little bit by making it `async`. Furthermore, given how `PDFBug` is being used, we can also (slightly) re-factor `PDFBug.init` such that the `PDFBug.enable`-call is done internally rather than having to handle that manually at the call-site.
(Finally, utilize `await` more in the `loadFakeWorker` helper function.)
This way there cannot be any *incorrect* cache hits, since Refs are guaranteed to be unique.
Please note that the reason for caching by Ref rather than doing something along the lines of the `localShadingPatternCache` (which uses a `Map` directly), is that TilingPatterns are streams and those cannot be cached on the `XRef`-instance (this way we avoid unnecessary parsing).
*This patch is very similar to the recently fixed `renderInteractiveForms`-options, see PR 13867.*
As far as I can tell, this *subtle* bug has existed ever since `AnnotationStorage`-support was first added in PR 12106 (a little over a year ago).
The value of the `includeAnnotationStorage`-option, as passed to the `PDFPageProxy.render` method, will (potentially) affect the size/content of the operatorList that's returned from the worker (for documents with forms).
Given that operatorLists will generally, unless they contain huge images, be cached in the API, repeated `PDFPageProxy.render` calls where the form-data has been changed by the user in between, can thus *wrongly* return a cached operatorList.
In the viewer we're only using the `includeAnnotationStorage`-option when printing, which is probably why this has gone unnoticed for so long. Note that we, for performance reasons, don't cache printing-operatorLists in the API.
However, there's nothing stopping an API-user from using the `includeAnnotationStorage`-option during "normal" rendering, which could thus result in *subtle* (and difficult to understand) rendering bugs.
In order to handle this, we need to know if the `AnnotationStorage`-instance has been updated since the last `PDFPageProxy.render` call. The most "correct" solution would obviously be to create a hash of the `AnnotationStorage` contents, however that would require adding a bunch of code, complexity, and runtime overhead.
Given that operatorList caching in the API doesn't have to be perfect[1], but only have to avoid *false* cache-hits, we can simplify things significantly be only keeping track of the last time that the `AnnotationStorage`-data was modified.
*Please note:* While working on this patch, I also noticed that the `renderInteractiveForms`- and `includeAnnotationStorage`-options in the `PDFPageProxy.render` method are mutually exclusive.[2]
Given that the various Annotation-related options in `PDFPageProxy.render` have been added at different times, this has unfortunately led to the current "messy" situation.[3]
---
[1] Note how we're already not caching operatorLists for pages with *huge* images, in order to save memory, hence there's no guarantee that operatorLists will always be cached.
[2] Setting both to `true` will result in undefined behaviour, since trying to insert `AnnotationStorage`-values into fields that are being excluded from the operatorList-building will obviously not work, which isn't at all clear from the documentation.
[3] My intention is to try and fix this in a follow-up PR, and I've got a WIP patch locally, however it will result in a number of API-observable changes.
At this point in time, all of the supported browsers (in the PDF.js project) have native `ReadableStream` implementations; see https://developer.mozilla.org/en-US/docs/Web/API/ReadableStream#browser_compatibility
Hence the polyfill is *only* necessary in Node.js environments now, and we shouldn't need to do any detailed feature detection either (since that was only done for the non-Chromium versions of the MS Edge browser).
Finally, we can slightly reduce the size of the Chromium-extension since the polyfill shouldn't be needed there either.
By not adding any additional non-`Dict` entries to the list of candidates for merging of sub-dictionaries, we can very slightly reduce the amount of parsing required by not having to *again* iterate through unmergeable data.
The original idea behind including the class name, when logging errors, was to improve things in the *hypothetical case* where `PDFViewer`- and `PDFSinglePageViewer`-instances would be used side-by-side.
Given that all of the relevant methods are synchronous this seem unlikely to really be necessary, and furthermore it's probably best to avoid using `this.constructor.name` since that's not guaranteed to do what you intend (we've seen repeated issues with minifiers mangling function/class names).
Once we're finally able to get rid of SystemJS, which is unfortunately still blocked on [bug 1247687](https://bugzilla.mozilla.org/show_bug.cgi?id=1247687), we might also want to clean-up (or even completely remove) the `BaseException` abstraction and simply extend `Error` directly instead.
At that point we'd need to (explicitly) set the `name` on each class anyway, so this patch is essentially preparing for future clean-up. Furthermore, after the `BaseException` abstraction was added there's been *multiple* issues filed about third-party minification breaking our code since `this.constructor.name` is not guaranteed to always do what you intended.
While hard-coding the strings indeed feels quite unfortunate, it's likely the "best" solution to avoid the problem described above.
Rather than caching only the *last* `PDFPageProxy.getAnnotations` call, and having to handle the intent separately, we can instead implement the caching in exactly the same way as done in the `PDFPageProxy.{render, getOperatorList}` methods.
This patch removes the only remaining closure in the `src/display/api.js` file, utilizing a similar approach as used in lots of other parts of the code-base, which results in a small decrease in the size of the *build* `pdf.js` file.
Given that `PDFWorker` is exposed through the *public* API, this complicates things somewhat since there's a couple of worker-related properties that really should stay *private*. Initially, while working on PR 13813, I believed that we'd need support for private (static) class fields in order to get rid of this closure, however I've managed to come up with what's hopefully deemed an acceptable work-around here.
Furthermore, some helper functions were simply moved into the `PDFWorker` class as static methods, thus simplifying the overall implementation (e.g. we don't need to manually cache the Promise in the `PDFWorker._setupFakeWorkerGlobal`-method).
Finally, as part of this re-factoring a number of missing JSDoc-comments were added which *together* with the removal of the closure significantly improves the `gulp jsdoc` output for the `PDFWorker` class.
*Please note:* This patch is tagged with `api-minor` since it deprecates `PDFWorker.getWorkerSrc()` in favor of the shorter `PDFWorker.workerSrc`, with the fallback limited to `GENERIC` builds.
It shouldn't be necessary to assign these variables to the global scope (as far as I can tell), either explicitly with `window` or implicitly with `var`, and this way we don't need to disable the ESLint `no-undef` rule; fixes another small part of issue 13862.
*Please note:* I wasn't going to put additional work into this code after PR 13869, however these changes looked so simple that I figured trying to get rid of the few remaining "Code scanning alerts" wouldn't hurt.
However, this file would still very much benefit from additional clean-up and re-factoring work, since it's quite old and currently contains some dead code (commented out).
The value of the `renderInteractiveForms` parameter, as passed to the `PDFPageProxy.render` method, will (potentially) affect the size/content of the operatorList that's returned from the worker (for documents with forms).
Given that operatorLists will generally, unless they contain huge images, be cached in the API, repeated `PDFPageProxy.render` calls that *only* change the `renderInteractiveForms` parameter can thus return an incorrect operatorList.
As far as I can tell, this *subtle* bug has existed ever since `renderInteractiveForms`-support was first added in PR 7633 (which is almost five years ago).
With the previous patch, fixing this is now really simple by "encoding" the `renderInteractiveForms` parameter in the *internal* renderingIntent handling.
With the changes made in PR 13746 the *internal* renderingIntent handling became somewhat "messy", since we're now having to do string-matching in various spots in order to handle the "oplist"-intent correctly.
Hence this patch, which implements the idea from PR 13746 to convert the `intent`-strings, used in various API-methods, into an *internal* renderingIntent that's implemented using a bit-field instead. *Please note:* This part of the patch, in itself, does *not* change the public API (but see below).
This patch is tagged `api-minor` for the following reasons:
1. It changes the *default* value for the `intent` parameter, in the `PDFPageProxy.getAnnotations` method, to "display" in order to be consistent across the API.
2. In order to get *all* annotations, with the `PDFPageProxy.getAnnotations` method, you now need to explicitly set "any" as the `intent` parameter.
3. The `PDFPageProxy.getOperatorList` method will now also support the new "any" intent, to allow accessing the operatorList of all annotations (limited to those types that have one).
4. Finally, for consistency across the API, the `PDFPageProxy.render` method also support the new "any" intent (although I'm not sure how useful that'll be).
Points 1 and 2 above are the significant, and thus breaking, changes in *default* behaviour here. However, unfortunately I cannot see a good way to improve the overall API while also keeping `PDFPageProxy.getAnnotations` unchanged.
Without this patch, when using `PDFPageView` directly[1] this CSS variable won't be updated and consequently things won't work as intended.
This is purposely implemented such that when a `PDFPageView`-instance is part of a viewer, we don't repeatedly set the CSS variable for every single page.
---
[1] See e.g. the "pageviewer" example in the `examples/components/` folder.
Given that issue 13862 tracks updating/modernizing the code, this patch purposely limits the scope of the changes. In particular, the following things are still left to address:
- The ESLint `no-undef` errors; for now the rule is simply disabled globally in this file.
- A couple of unused variables are commented out for now, but could perhaps just be removed.
The new command is a *variation* of the standard `gulp test` command and will run all unit/font/integration-tests just as normal, while *only* running ref-tests for XFA-documents to speed up development.
Given that we currently have (some) unit-tests for XFA-documents, and that we may also (in the future) want to add integration-tests, it thus makes sense to run all test-suites in my opinion.
*Please note:* Once this patch has landed, I'll submit a follow-up patch to https://github.com/mozilla/botio-files-pdfjs such that we can also run the new command on the bots.
Given how trivial the `isEOF` function is, we can simply inline the check at the various call-sites and remove the function (which ought to be ever so slightly more efficient as well).
Furthermore, this patch also changes the `EOF` primitive itself to a `Symbol` instead of an Object since that has the nice benefit of making it unclonable (thus preventing *accidentally* trying to send `EOF` from the worker-thread).
Given that the GitHub Advanced Security workflow now covers everything that LGTM does, but generally faster and with better GitHub-integration, there's no longer much point in also running LGTM separately.
As a follow-up to this patch, we should also disable/remove the LGTM-integration from the PDF.js repository.
The `viewerCssTheme` option was not rendered because its entry in
`preferences_schema.json` did not have a `title`.
The order of keys in `preferences_schema.json` determines the order of the
rendered preferences in the options UI. Since `viewerCssTheme` affects the UI
very significantly, I have moved the option to the top.
The only purpose, according to the README and existing files, is to
parse an integer from those lines, so (\d+) is sufficient for that. This
avoids potential exponential backtracking as flagged by CodeQL. I have
compared the output of the script with and without these changes and the
resulting files are the same.
Even though the code as-is *should* be safe, given that we're using an Object with a `null` prototype, it cannot hurt to change this to a Map to prevent any issues (since we're parsing unknown and potentially unsafe data).
Overall I also think that these changes improve the `parseQueryString` call-sites, since we now have a proper way of checking for the existence of a particular key (and don't have to use `in` which stringifies the keys in the Object).
This patch also changes the default, when no `value` exists, from `null` to an empty string since the use of `decodeURIComponent` currently can modify the value in a somewhat surprising way (at least to me).
Note how `decodeURIComponent(null) === "null"` which is unlikely to be what you actually want, whereas `decodeURIComponent("") === ""` which seems much more helpful.
This allows us to get the quality checks that LGTM does into GitHub
Advanced Security. Since it not only runs security checks anymore, the
workflow is also renamed to CodeQL to make this more explicit (and this
matches the documentation better).
This makes it consistent with the GitHub Advanced Security file and,
more importantly, ensures that all steps have a proper name for better
visibility.
While fixing issue 13794, I noticed that cancelling the `ReadableStream` returned by the `PDFPageProxy.streamTextContent`-method could lead to "Uncaught promise" messages in the console.[1]
Generally speaking, we don't really care about errors when *cancelling* a `ReadableStream` and it thus seems reasonable to simply suppress any output in those cases.
---
[1] Although, after that issue was fixed you'd now need to set the API-option `stopAtErrors = true` to actually trigger this.
This patch utilizes the same approach as used in lots of other parts of the code-base, which thus *slightly* reduces the size of this code.
By removing some of the (current) indirection, we can also simplify the JSDocs a little bit. Looking at the `gulp jsdoc` output, this actually seem to *improve* the documentation for this class.
Prior to PR 13042, when scripting wasn't really possible to use outside of the full viewer, the `enableScripting` option made sense.
However, at this point in time having to both pass in a `PDFScriptingManager`-instance *and* set the `enableScripting`-boolean when creating a `BaseViewer`-instance feels redundant and (mostly) annoying. Hence this patch, which removes the *separate* boolean and always enables scripting when `scriptingManager` is provided.
The relevant "viewer component" examples are also updated (with a comment), but in such a way that scripting support won't just break when used with the current PDF.js releases.
The PDF in bug 1721949 uses many unique pattern objects
that references the same shading many times. This caused
a new canvas pattern to be created and cached many times
driving up memory use.
To fix, I've changed the cache in the worker to key off the
shading object and instead send the shading and matrix
separately. While that worked well to fix the above bug,
there could be PDFs that use many shading that could
cause memory issues, so I've also added a LRU cache
on the main thread for canvas patterns. This should prevent
memory use from getting too high.
- All the scale factors in for the substitution font were wrong because of different glyph positions between Liberation and the other ones:
- regenerate all the factors
- Text may have polish chars for example and in this case the glyph widths were wrong:
- treat substitution font as a composite one
- add a map glyphIndex to unicode for Liberation in order to generate width array for cid font
- In order to better compute text fields size, use line height with no gaps (and consequently guessed height for text are slightly better in general).
- Fix default background color in fields.
Given that we've over time been reducing the number of `compatibilityParams` in use, there's now few enough left that I think it makes sense to simply inline them directly in the `web/app_options.js` file.
Note that we recently inlined/removed the separate `src/display/api_compatibility.js` file, see PR 13525, and that it (in my opinion) thus makes sense to do the same in the `web/`-folder. This patch will also slightly reduce the size of *built* `web/viewer.js` file, which cannot hurt.
For code that's part of the core library, rather than e.g. the `web/`-folder, we should always be careful about *directly* accessing any DOM methods.
The `navigator` is one such structure, which shouldn't be assumed to always be available and we should thus check that it's actually present.[1]
Hence this patch re-factors the `navigator.platform` access, in the `AnnotationLayer`-code, to ensure that it's generally safe. Furthermore, to reduce unnecessary repeated string-matching to determine the current platform, we're now using a shadowed getter which is evaluated only once instead (at first access).
---
[1] Note e.g. the `isSyncFontLoadingSupported` getter, in the `src/display/font_loader.js` file.
- an Image element was created, attached to its parent but the $globalData property was not set and that led to an error.
- the pdf in bug 1722029 has 27 rendered rows (checked in Acrobat) when only one was displayed: this patch some binding issues around the occur element.
This patch makes use of the existing `ignoreErrors` option, thus allowing a page to continue parsing/rendering even if (some of) its sub-streams are corrupt. Obviously this may cause *part* of a page to be broken/missing, however it should be better than (potentially) rendering nothing.
Also, to the best of my knowledge, this is the first bug of its kind that we've encountered.
To avoid having to pass in a bunch of, for a `BaseStream`-instance, mostly unrelated parameters when initializing a `StreamsSequenceStream`-instance, I settled on utilizing a callback function instead to allow conditional Error-suppression.
Note that the `StreamsSequenceStream`-class is a *special* stream-implementation that we only use when the `/Contents`-entry, in the `/Page`-dictionary, consists of an Array with streams.
There's no good reason, as far as I can tell, to explicitly define a bunch of methods to be `undefined`, which the current unconditional "copying" of methods will do.
Note that of the `OPS` ~23 percent don't, for various reasons, have an associated method on the `CanvasGraphics.prototype`.
For e.g. the `gulp mozcentral` command, the *built* `pdf.js` file decreases from `304 607` to `301 295` bytes with this patch. The improvement comes mostly from having less overall indentation in the code.
Apparently I didn't put one of the disable comments on the *correct* line, since I didn't read the instructions carefully enough, so let's try again.
Note that, most unfortunately, disabling of warnings isn't applied until *after* a patch has been merged.
The former prevents unintended errors in regexes. The latter is
something I ran into a couple of times before editors had proper support
for syntax highlighting in template literals, and it's an easy mistake
to make, especially when converting older code to use template literals.
Nowadays most editors display template literals in different colors, but
nevertheless it seems like a good idea to enable this rule.
Fortunately both rules don't require any code changes, so it's more for
prevention.
Most of the warnings we don't really care about, and those are simply white-listed using inline comments; however two cases prompted actual code changes:
- In `src/display/pattern_helper.js` the branch in question is indeed unreachable, and should thus be safe to remove. (This code originated in PR 4192, which is now over seven years ago.)
- In `test/test.js`, the function in question indeed doesn't accept any arguments. (The patch also re-formats a string just above, which didn't seem worthy of a separated patch.)
This now leaves only *one* warning in the LGTM report, however that one is a false positive that we'll need to report upstream.
In cases where even the very *first* attempt at reading from an object will throw, simply ignoring such objects will help improve rendering of *some* corrupt documents.
Note that this will lead to more parsing in some cases, but considering that this only applies to *corrupt* documents that shouldn't be a big deal.
There's no good reason, as far as I can tell, to have `PDFPageView.reset` attempt to cancel `annotationLayer`/`xfaLayer` rendering in one special-case (this is mostly a leftover from older code). Previously cancelling was moved into the separate `PDFPageView.cancelRendering`-method, and by slightly tweaking the conditions there we're able to remove a bit of now unnecessary code from the `PDFPageView.reset`-method.
Originally the `xfaLayer` wasn't implemented in such a way that it supported being removed from the DOM when pages were evicted from the cache, however this limitation was lifted in PR 13427 and the `xfaLayer` should thus be handled similar to e.g. the `annotationLayer`.
In addition to removing the `xfaLayer` from the DOM, this patch *also* implements proper rendering/hiding-handling for it (mirroring the `annotationLayer`-code).
*Please note:* This patch is tagged API-minor just in case[1], since it changes the signatures of a couple of `PDFPageView`-methods to improve readability of the code.
---
[1] Although users are *hopefully* not directly accessing any of the affected methods, and are rather using e.g. `PDFViewer` in which case none of these changes will matter.
Bug 1721218 has a shading pattern that was used thousands of times.
To improve performance of this PDF:
- add a cache for patterns in the evaluator and only send the IR form once
to the main thread (this also makes caching in canvas easier)
- cache the created canvas radial/axial patterns
- for shading fill radial/axial use the pattern directly instead of creating temporary
canvas
- a rectangle can be defined after a field and so from a z-index pov, it's on top of the field, which means that the rectangle avoid to have some mouse events in the fields;
- so this patch just disable pointer events for all elements under svg (included).
Given that PR 13772 seem to have worked as intended, it probably cannot hurt to enable the LGTM report for *most* of the remaining code in the `test`-folder. (The one remaining exception is code which originated *outside* of the PDF.js project.)
*Please note:* We'll need to land this patch to actually see any difference in the LGTM results.
Currently the `!mergeSubDicts` code-path is essentially just duplicated code, which we can easily avoid by simply moving that check. (This may lead to ever so slightly more parsing for this case, but the difference ought to be negligible in practice.)
By default all files/folders classified as test-code are excluded in the LGTM report, however I do believe that it makes sense to include at least the `font`/`integration`/`unit`-test folders.
It seems pretty obvious that you want to avoid *accidentally* introducing any simple logical errors in the tests (and not only in the main code), to ensure that e.g. the unit-tests actually test the desired thing.
Hopefully I've interpreted the information at https://lgtm.com/help/lgtm/customizing-file-classification correctly, however note that we may need to land this patch to actually see any difference in the results.
- in real life some xfa contains xml like <xfa:data><xfa:Foo><xfa:Bar>...</xfa:data>
since there are no Foo or Bar in the xfa namespace the JS representation are empty
and that leads to errors.
- so the idea is to make all nodes under xfa:data namespace agnostic which means
that ns are removed from nodes in the parser but only xfa:data descendants.
*Please note:* The PDF document in issue 13751 is *dynamically* created (in e.g. Adobe Reader), with pages added when certain buttons are clicked, hence this patch simply fixes the breaking error and nothing more.
It looks like the current code contains a little bit too much copy-and-paste from the *similar* `index` branch above, since we cannot set the `startIndex` to a negative value. Note how it's being used to initialize the loop-variable, which is then used to lookup values in an Array and accessing the `-1`th element of an Array obviously makes no sense.
*This is yet another case where I've got no idea if the patch is correct, but it does at least fix a breaking error :-)*
Note how in the [`Binder._bindValue` method](683ce66a48/src/core/xfa/bind.js (L92-L93)), we're assuming that if a `data`-value exists then it'll also be possible to actually access it. For the `XFAAttribute`-implementation however, the second method is missing and that's what causes the breaking errors in issue 13748.
Please note that another possible way of "fixing" the error wouldn't been to simply change the exists-check to return `false`, and I could see that being a preferred solution.
However, the reason for submitting the current patch is that we get *fewer* warnings about Nodes with mis-matched types this way.
As can be seen in the code (see below), the `searchNode` helper function will return `null` in some cases and all of its call-sites should protect against that before attempting to access the returned data.
While only one of these changes were necessary to fix the breaking errors in issue 13756, in order to prevent future bugs I've added similar defensive code throughout this file.
- 07955fa1d3/src/core/xfa/som.js (L169)
- 07955fa1d3/src/core/xfa/som.js (L239)
- 07955fa1d3/src/core/xfa/som.js (L254)
For e.g. `gulp mozcentral`, the *built* `pdf.worker.js` file decreases from `1 837 608` to `1 834 907` bytes with this patch-series.
The improvement comes first of all from less overall indentation in `PDFFunction`, and secondly from the removal of (now) unnecessary indirection in the code.
*This follows the exact same princial as PR 12083, but for the `PDFFunction` parsing instead.*
Given that the IR format is completely unused now, all that the current code does is add a bunch of unnecessary indirection/overhead to the handling of PDF-functions.
With this patch, the `PDFPageProxy.getOperatorList` method will now return `PDFOperatorList`-instances that also include Annotation-operatorLists (when those exist). Hence this closes a small, but potentially confusing, gap between the `render` and `getOperatorList` methods.
Previously we've been somewhat reluctant to do this, as explained below, but given that there's actual use-cases where it's required probably means that we'll *have* to implement it now.
Since we still need the ability to separate "normal" rendering operations from direct `getOperatorList` calls in the worker-thread, this API-change unfortunately causes the *internal* renderingIntent to become a bit "messy" which is indeed unfortunate (note the `"oplist-"` strings in various spots). As-is I suppose that it's not all that bad, but we may want to consider changing the *internal* renderingIntent to e.g. a bitfield in the future.
Besides fixing issue 13704, this patch would also be necessary if someone ever tries to implement e.g. issue 10165 (since currently `PDFPageProxy.getOperatorList` doesn't include Annotation-operatorLists).
*Please note:* This patch is *also* tagged "api-minor" for a second reason, which is that we're now including the Annotation-id in the `beginAnnotation` argument. The reason for this is to allow correlating the Annotation-data returned by `PDFPageProxy.getAnnotations`, with its corresponding operatorList-data (for those Annotations that have it).
Given that Internet Explorer is, since some time now, no longer supported in the PDF.js library this `<meta>` tag can also be removed (added in PR 6374).
When working on PR 13612, I mostly prioritized a simple solution that didn't require touching a lot of code. However, while working on PR 13735 I started to realize that the static `Name.empty` construction really wasn't a good idea.
In particular, having a special `Name`-instance where the `name`-property isn't actually a String is confusing (to put it mildly) and can easily lead to issues elsewhere. The only reason for not simply allowing the `name`-property to be an *empty* string, in PR 13612, was to avoid having to touch a lot of existing code. However, it turns out that this is only limited to a few methods in the `PartialEvaluator` and a few of the `BaseLocalCache`-implementations, all of which can be easily re-factored to handle *empty* `Name`-instances.
All-in-all, I think that this patch is even an *overall* improvement since we're now validating (what should always be) `Name`-data better in the `PartialEvaluator`.
This is what I ought to have done from the start, sorry about the code churn here!
Currently it's possible to run e.g. `gulp unittest`, `gulp fonttest`, and `gulp integrationtest` *separately* on the bots; see https://github.com/mozilla/botio-files-pdfjs
However, it's not possible to run *only* the `gulp browsertest` command on the bots without also running the full test-suite. In some cases, e.g. if the "browsertest" times out, having a way to only re-run those would thus save some time and resources.
If/when this patch lands, I'll follow-up with a patch adding a new `on_cmd_browsertest.js` file to the https://github.com/mozilla/botio-files-pdfjs repository.
*Please note:* This patch doesn't fix rendering of (various) patterns in browsers/environments without full `CanvasPattern.setTransform()` support, but it at least prevents outright failures and thus allows the rest of the page to render.
This patch provides a temporary work-around for Firefox 78 ESR[1], and for Node.js environments (see issue 13724), where rendering is currently completely broken.
---
[1] Please note that the `createMatrix` helper function doesn't actually work as intended. The reason is that it's not `DOMMatrix` itself which is unsupported in older Firefox versions, but rather calling `CanvasPattern.setTransform(...)` with a `DOMMatrix`-argument.
Furthermore, the `createSVGMatrix` fallback won't actually help either since that method doesn't accept any parameters and would thus require *manually* specifying the matrix-state; see e.g. https://developer.mozilla.org/en-US/docs/Web/API/CanvasPattern/setTransform#examples
Finally, given that it's less than a month to the [Firefox 91 ESR release](https://wiki.mozilla.org/RapidRelease/Calendar) and that as-is all patterns are completely broken e.g. when using the latest viewer in Firefox 78 ESR, I'm just not convinced that it's worth the "hassle" of providing a more proper work-around.
This way we ensure that these BBox values are *always* defined as expected for every `case`-block, and we also don't need to duplicate the lookup in multiple places. (Also, the patch removes a couple of unnecessary line-breaks in existing comments.)
Fixes https://github.com/mozilla/pdf.js/pull/13691#pullrequestreview-702356627, which was flagged by LGTM.
- font line height is taken into account by acrobat when it isn't with masterpdfeditor: I extracted a font from a pdf, modified some ascent/descent properties thanks to ttx and the reinjected the font in the pdf: only Acrobat is taken it into account. So in this patch, line heights for some substituted fonts are added.
- it seems that Acrobat is using a line height of 1.2 when the line height in the font is not enough (it's the only way I found to fix correctly bug 1718741).
- don't use flex in wrapper container (which was causing an horizontal overflow in the above bug).
- consequently, the above fixes introduced a lot of small regressions, so in order to see real improvements on reftests, I fixed the regressions in this patch:
- replace margin by padding in some case where padding is a part of a container dimensions;
- remove some flex display: some containers are wrongly sized when rendered;
- set letter-spacing to 0.01px: it helps to be sure that text is not broken because of not enough width in Firefox.
While I don't know if it's technically correct to even do this, it could provide a slightly better out-of-the-box behaviour in browsers that specify (from the PDF.js `l10n`-folder perspective) "incomplete" language codes.
Rather than immediately falling back to English, we'll use a white-list to try and re-write a "partial" language code to a (hopefully) suitable one that matches an existing `l10n`-folder. The disadvantage of this solution is that the list needs to be kept *manually* up-to-date with any changes in the `l10n`-folder, however new locales are added infrequently enough that this should be acceptable.
Fixes 13689 (assuming we actually want/care to do so, otherwise we should just WONTFIX the issue).
With the changes in PR 13687 we're now checking if `target` is defined *twice* in a row, which shouldn't be necessary :-)
(I noticed this when glancing at the unofficial LGTM results; maybe we should re-evalute the decision to not integrate that into the CI.)
When XFA support was added, the size of the *built* `pdf.worker.js` file increased quite a bit. Hence I think that it makes sense to, where easily possible, do what we can to (slightly) reduce the size of the PDF.js library.
The supplemental font data files (added for XFA rendering), containing rescale-factors respectively widths, seem like an excellent candidate here since they're not particularly large in either line-count or file sizes.
In this patch these files are instead merged into a *single* file per font, rather than four different ones, and even with these changes the resulting source files don't become all that large.[1]
For e.g. the `gulp mozcentral` build, this reduces the size of the *built* `pdf.worker.js` file by more than `3 kB`. Given the overall simplicity of the patch, that kind of size decrease definitely seem worthwhile to me.
---
[1] Especially when compared to truly large files such as e.g. `glyphlist.js`, `metrics.js`, and `unicode.js`.
Previously, when we filled image masks we didn't copy over the current transformation,
this caused patterns to be misaligned when painted. Now we create a temporary
canvas with the mask and have the transform copied over and offset it relative to
where the mask would be painted. We also weren't properly offsetting tiling patterns.
This isn't usually noticeable since patters repeat, but in the case of #13561 the pattern
is only drawn once and has to be in the correct position to line up with the mask image.
These fixes broke #11473, but highlighted that we were drawing that correctly by
accident and not correctly handling negative bounding boxes on tiling patterns.
Fixes#6297, #13561, #13441
Partially fixes#1344 (still blurry but boxes are in correct position now)
- Fix a typo in order to open the pdf in issue #13679
- After fixing the fill default color there wer some regressions because of z-index
and when fixing z-index there were some regressions because of borders
- So fix the borders rendering.
*While I cannot guarantee that this will fix the recent intermittents, this patch really shouldn't hurt.*
By setting the Image `src` first, there's a small possibility that the Image is loaded *before* we've had a change to attach the `onload`/`onerror` callbacks which may cause the Promise to remain in a pending state.
Note that prior to PR 13641 we didn't correctly await all image resources to actually load, which could explain the very recent intermittent test-failures.
According to https://docs.npmjs.com/cli/v7/configuring-npm/package-json#version, the "version" field is not required[1]:
> If you don't plan to publish your package, the name and version fields are optional.
Hence it shouldn't be necessary to have a "dummy" `version` field in the `package.json` file, and it seems quite unfortunate to have an essentially meaningless entry in that file.[2]
Furthermore, I'd even go as far as suggesting that it's actually doing more harm than good in practice, since it's not uncommon for people to open issues where they simply quote the `package.json`-entry when filling out the ISSUE_TEMPLATE thus causing confusion as to which *exact* version is actually used.
Unless I'm misremembering, I believe that the only reason for adding the `version` field was that is was necessary in order for things to work back when testing was run on Travis.
Now we're using GitHub Actions instead, where things seem to work just find even without a `version` field; hence why I think it makes sense to remove this.
---
[1] Please note that this patch doesn't affect the `pdfjs-dist` package, since the `package.json` file used there is created in `gulpfile.js` during building.
[2] Trying to, automatically, update the `version` field on *every* commit really doesn't seem worth it to me.
The PDF.js API has only ever supported accessing the original file ID, however the second one that (should) exist in *modified* documents have thus far been completely inaccessible through the API.
That seems like a simple oversight, caused e.g. by the viewer not needing it, since it really shouldn't hurt to provide API-users with the ability to check if a PDF document has been modified since its creation.[1]
Please refer to https://www.adobe.com/content/dam/acom/en/devnet/pdf/pdfs/PDF32000_2008.pdf#G13.2261661 for additional information.
For an example of how to update existing code to use the new API, please see the changes in the `web/app.js` file included in this patch.
*Please note:* While I'm not sure if we'll ever be able to remove the old `PDFDocumentProxy.fingerprint` getter, given that it's existed since "forever", that probably isn't a big deal given that it's now limited to only `GENERIC`-builds.
---
[1] Although this obviously depends on the PDF software following the specification, by updating the second file ID as intended.
Please refer to https://developer.mozilla.org/en-US/docs/Web/API/Web_Workers_API/Structured_clone_algorithm
Based on that information, and manually testing our code, the implementation in `cloneValue` has the following shortcomings:
- Attempting to clone `function`s is only prevented when they're part of an Object, but is currently allowed when they occur standalone.
- Cloning of `Symbol`s is currently not prevented, which it should be since the native structured clone algorithm throws.
- Any disallowed types should be checked first, to reduce the risk of future changes accidentally allowing something that shouldn't be supported.
Using `instanceof Object` is generally problematic, since it's not guaranteed to always do the right thing for all Objects.
(I stumbled upon this while working on another patch, when I noticed that the `outlineView` was broken with workers disabled.)
- support paragraph margins, line height, letter spacing, ...
- compute missing dimensions from fields based almost on the dimensions of caption contents.
- it aims to fix#13583;
- fix the switch to breakBefore target;
- force the layout of an unsplittable element on an empty page;
- don't fail when there is horizontal overflow (except in lr-tb);
- handle correctly overflow in the same content area (bug 1717805, bug 1717668);
- fix a typo in radial gradient first argument.
- some pdf use some fonts which are not embedded or they don't have any width array or don't have any css info (e.g. for standard fonts or Arial).
- so add widths arrays for Liberation fonts in order to compute the ones for other fonts in using scale factors array.
- the break element has been deprecated in XFA 2.4 but some old documents can use it, so replace it with one (or more) of its possible substitutions:
- breakBefore;
- breakAfter;
- overflow.
- when binding (after parsing) we get a map between some template nodes and some data nodes;
- so set user data in input handlers in using data node uids in the annotation storage;
- to save the form, just put the value we have in the storage in the correct data nodes, serialize the xml as a string and then write the string at the end of the pdf using src/core/writer.js;
- fix few bugs around data bindings:
- the "Off" issue in Bug 1716980.
According to a comment in `readCmapTable`, we're assuming that the cmap tables (when more than one exist) are sorted in ascending order. If that's not the case, keep checking the following cmap tables in order to fix the referenced issue.
Given that this property is only used with password protected documents, and is consequently document-specific rather than viewer-specific, ensure that `IPDFLinkService.externalLinkEnabled` is actually being reset by `PDFViewerApplication.close`.
To make things less confusing/inconsistent, remove the *undocumented* `externalLinkEnabled` property from the `PDFLinkService` constructor and force it to always be manually set when needed.
- when the CSS line-height property is set to 'normal' then the value depends of the user agent. So use a line height based on the font itself and if for any reasons this value is not available use 1.2 as default.
- it's a partial fix for https://bugzilla.mozilla.org/show_bug.cgi?id=1717681.
This patch provides an overall simpler *and* more consistent way of handling the `viewport` parameter during printing of XFA forms, since it's now again guaranteed to always be an instance of `PageViewport`.
Furthermore, for anyone attempting to e.g. implement custom printing of XFA forms this probably cannot hurt either.
Given that the "print"-intent is special, and that we should always fallback to the "display"-intent, let's ensure that the code actually reflects that.
Also, ensure that the method always returns a `Promise` since that's what the documentation says.
The `web/ui_utils.js` file should be usable from basically anywhere in the `web/`-folder, hence it should ideally not have any dependecies on its own and particularily *not* onces that pull in entire (large) factories.
Apparently some really bad PDF software can create documents with *empty* `Name`-entries, which we thus need to somehow deal with.
While I don't know if this patch is necessarily the best solution, it should at least ensure that the *empty* `Name`-instance cannot accidentally match a proper `Name`-instance (and it doesn't require changes to a lot of existing code).[1]
---
[1] I briefly considered using a `Symbol` rather than an Object, but quickly decided against that since the former one [is not clonable](https://developer.mozilla.org/en-US/docs/Web/API/Web_Workers_API/Structured_clone_algorithm#supported_types) and `Name`-instances may be sent to the API.
This extends the approach in PRs 12848 and 13606 to also apply to the `xfaLayer`, since otherwise XFA forms will be similarly broken in most non-default scroll/spread modes.
- it aims to fix https://bugzilla.mozilla.org/show_bug.cgi?id=1716838.
- some fonts in the pdf in the bug where bold when they shouldn't so write the font properties in the html to avoid to use some wrong inherited ones.
Note that as far as I can tell, this is *not* a regression but rather a bug which has existed since basically "forever".
**In order to reproduce this easily:**
- Open the viewer.
- Set the zoom level to `400%`,
- Search for "expression".
The problem here is that when scrolling matches into view, we're scrolling to the start of the *containing* `textLayer` element rather than the start of the highlighted match itself.[1] When the entire width (or at least most) of the page is visible in the viewer, that doesn't really matter though which is likely why this bug has gone unnoticed for so long.[2]
Given that the highlighted match can be placed anywhere, e.g. even at the very end, within its `textLayer` element it's quite easy to see why the current implementation becomes a problem at higher zoom levels. All of this is then *further* exacerbated by `PDFFindController.scrollMatchIntoView` using a negative left offset, to ensure that the current match has some (visible) context available once scrolled into view.
In order to address this long-standing bug, we'll determine the (left) offset of the `selected` match and use that to modify the final position scrolled to in `PDFFindController.scrollMatchIntoView` such that the match is visible regardless of zoom level.
---
[1] Unfortunately we cannot directly scroll to the `selected` match, since it's not absolutely positioned and changing that would cause other bugs/regressions (note recent patches in that area).
[2] I did actually stumble upon this problem a little while ago, while working on PR 13482, but forgot to look into this again until I saw the new issue.
- partial fix for https://bugzilla.mozilla.org/show_bug.cgi?id=1716980;
- some pdf can contain an invalid font family (e.g. 'Windings 3') so in this case remove the space;
- the font family in typeface attribute doesn't always match the one defined in the FontDescriptor dictionary.
This patch is an attempt at closing an old, and seemingly trivial, issue and the SVG-files created by the `pdf2svg.js` examples still appear to work just fine when opened in browsers (tested with Firefox Nightly and Google Chrome Beta).
Given that we're not imposing any font-type restrictions[1] in the non-/FontDescriptor case, it's not really clear to me why we'd actually need to do that in the general case.
Please note that there's some *expected* movement, all of which should be improvements, in the `fips197.pdf` file with this patch.
---
[1] With the exception of Type3-fonts, of course.
Please note that while the `gulp types`/`gulp typestest` tasks (obviously) still work with this patch, I've got no idea if the output is first of all even useable and secondly if it's actually useful in practice.
However, in the interest of closing some (seemingly simple) issues, I suppose that this probably shouldn't hurt (and we'd need TypeScript users to help improve things here).
*Sorry about the churn here, since the change that I made in PR 13516 was not very smart.*
With the current code, it's now *impossible* for a user to actually control the `useSystemFonts` option manually. To prevent outright breakage we obviously still need to default to setting `useSystemFonts = false` when `disableFontFace === true`, however that should be possible for an API consumer to override.
With the introduction of `PDFScriptingManager._closeCapability` in PR 13074, the pre-existing `PDFScriptingManager._pageEventsReady` boolean essentially became redundant.
Given that you always want to avoid tracking closely related state *separately*, since it's easy to introduce subtle bugs that way, we should just remove `PDFScriptingManager._pageEventsReady` now.
Obviously I *should* have done this already back in PR 13074, sorry about the churn here!
- PR #13554 is buggy, so this patch aims to fix bugs.
- check if a component fits into its parent in taking into account the parent layout.
- introduce method isSplittable for template nodes to know if a component can be splitted in case of overflow.
- some containers doesn't always have their 2 dimensions and those dimensions re based on contents;
- so in order to measure text, we must get the glyph widths (for the xfa fonts) before starting the layout;
- implement a word-wrap algorithm;
- handle font change during text layout.
Note, this only really fixes Radial/Axial shading patterns with masks.
I'm guessing tiling patterns and mesh patterns would also be broken
if applied like the test pdf. Hopefully I'll have some time to make
test cases for the other shadings.
Fixes#13372
- and fix few bugs:
- avoid infinite loop when layout the document;
- avoid confusion between break and layout failure;
- don't add margin width in tb layout when getting available space.
By adding basic linting of JSON files, we can ensure that they're actually valid and prevent e.g. test-failures caused by *accidental* errors when editing the `test/test_manifest.json` file (something that I've done *many* times myself).
For now this simply uses the `recommended` configuration, but we can obviously tweak this later if/when needed. Please find additional information at https://github.com/azeemba/eslint-plugin-json
Rather than having to create and check a *separate* `ToUnicodeMap` to handle these cases, we can simply use the `fallbackToUnicode`-data (when it exists) to directly supplement *missing* /ToUnicode entires in the regular `ToUnicodeMap` instead.
Note that all standard Encodings have the same length (i.e. `256` elements) and that missing entries are always represented by empty strings, hence why a separate exists-check isn't necessary in the `baseEncoding` case.
*This is somewhat similiar to the recent changes, in PR 13277, for fonts with an /Encoding entry.*
Currently we're *completely* ignoring the `builtInEncoding`, from the font data itself, for fonts which have a built-in /ToUnicode map.
While it (obviously) doesn't seem like a good idea in general to simply overwrite existing built-in /ToUnicode entries, it should however not hurt to use the `builtInEncoding` to supplement *missing* /ToUnicode entires.
- it aims to avoid to loop forever when opening pdf in #13213;
- the idea is to consider subformSet as inexistent when running in the tree. So if we've subformA > subformSet > subformB then subformB will be visited as a direct child of subformA.
This is first of all consistent with all of the other (similar) factories, and secondly it will also simplify a future addition of a corresponding `NodeSVGFactory` (if that's ever deemed necessary).
Given that `DOMMatrix` is, unsurprisingly, not supported in Node.js the `createMatrix` helper function in `src/display/pattern_helper.js` is most likely broken in Node.js environments. It will obviously try to fallback to the `DOMSVGFactory`, however that isn't intended for Node.js usage and errors will be thrown.
Rather than trying to implement a `NodeSVGFactory`, this patch takes the easier route of just adding a `DOMMatrix` polyfill using: https://www.npmjs.com/package/dommatrix
This isn't done only for simplicity, but it'll become necessary anyway since the `createMatrix` helper function is only temporary and will be removed in the future.
- a checkbox or radio doesn't have to be rescaled when the container is large so give the extra space to the caption to avoid some word wrapping.
- when the caption is on the right, then put ui on the left as first element and so remove flex:row-reverse stuff.
Currently only the Foxit license-file is included, which is most likely just an oversight as far as I can tell.
Furthermore, to be able the tell the two license-files apart, the Foxit one is also renamed slightly.
We must force-fetch standard font data, when `disableFontFace = true` is set in the API, since otherwise rendering in e.g. the viewer is still broken (same as before PR 12726 landed).
*Please note:* We still need to also load standard font data for patterns and/or some text-rendering modes, however that will require larger changes so I figured that it cannot hurt to submit *this* patch right now.
*This implementation is basically a copy of the pre-existing `builtInCMapCache` implementation.*
For some, badly generated, PDF documents it's possible that we'll end up having to fetch the *same* standard font data over and over (which is obviously inefficient).
While not common, it's certainly possible that a PDF document uses *custom* font names where the actual font then references one of the standard fonts; see e.g. issue 11399 for one such example.
Note that I did suggest adding worker-thread caching of standard font data in PR 12726, however it wasn't deemed necessary at the time. Now that we have a real-world example that benefit from caching, I think that we should simply implement this now.
- Some js files contain scale factors for each glyph in order to rescale Liberation to have a final font with the correct width.
- A lot of XFA have some containers where their dimensions are based on their text content, so using default font from browser can lead to an almost unreadable pdf.
- a lot of xfa files are using Myriad pro or Arial fonts without embedding them and some containers have some dimensions based on those font metrics. So not having the exact same font leads to a wrong display.
- since it's pretty hard to find a replacement font with the exact same metrics, this patch gives the possibility to read glyf table, rescale each glyph and then write a new table.
- so once PR #12726 is merged we could rescale for example Helvetica to replace Myriad Pro.
Given that there's no fallback on the worker-thread, it shouldn't be necessary to initialize `CMapReaderFactory`/`StandardFontDataFactory` when `useWorkerFetch = true` is set.
Slightly unrelated, but this patch also ensures that the `useSystemFonts` default value only does the `isNodeJS` check in builds where that's actually necessary.
At this point in time, the `apiCompatibilityParams` is essentially unused with the sole exception of the `disableFontFace` handling for Node.js environments.
Given that `isNodeJS` is a constant now (originally it was a function), we can simply set the correct fallback value for `disableFontFace` directly in the API and clean-up the code a bit here.
This patch uses the new option added in PR 12726 to *also* allow fetching binary CMap data directly in the worker-thread in browsers.
Given that these changes remove the need to transfer data between threads for the default (browser) use-case, we can also revert the changes in PR 11118 since that simplifies the overall implementation.
Given that these factories are being used in *different* files, for Browser respectively Node.js implementations, it seems reasonable to move them into their own file instead.
- some elements weren't displayed because their rotation angle was not taken into account;
- fix box model (XFA concept):
- remove use of outline;
- position correctly border which isn't part of box dimensions;
- fix margins issues (see issue #13474).
- move border on button instead of having it on wrapping div;
Generally, in the `src/display/` folder, we utilize `DOMSVGFactory` rather than manually creating an SVG-element; hence let's do the same thing in `src/display/pattern_helper.js` as well.
While this prevents the error which is currently thrown by the `assert` in the `DOMSVGFactory.create` method, the pattern still doesn't actually render (visibly). However, in the interest of getting rid of some open issues, this patch should make (some) sense and there's already other issues about patterns in the SVG-backend,
Given that, as clearly [outlined in the FAQ](https://github.com/mozilla/pdf.js/wiki/Frequently-Asked-Questions#backends), the SVG-backend is *not* officially supported and that there's currently no development of it; this is probably the most that is reasonable to do here.
This is necessary now, since with the previous patch the /FontBBox potentially depends on the contents of the /CharProcs-streams.
Note that if `getOperatorList` is called *before* `getTextContent`, this patch doesn't matter since the font is already fully loaded/parsed. However, for e.g. the `text` test-cases this is necessary to ensure correct reference images.
For Type3 fonts where the /CharProcs-streams of the individual glyph starts with a `d1` operator, we can use that to build a fallback bounding box for the font and thus improve text-selection in some cases.
While these objects aren't exactly that big and/or complex, they are nonetheless *only* necessary for XFA documents.
However, currently these objects are initialized *eagerly* for all PDF documents. By using the same pattern as elsewhere in the code-base, it's very easy to make these lazily initialized; so let's just do that instead :-)
Looking at the history of the code, this `gulp` task was *accidentally* broken in the Gulp 4 upgrade in PR 10266 (all the way back in 2018).
Given that no one has either noticed and/or complained about this since, it's safe to assume that it's completely unused, hence why I suggest that we just remove this `gulp` task (rather than fixing it).
- attribute 'use' was already implemented but not usehref
- in general, usehref should make reference to current document
- add support for SOM expressions in use and usehref to search a node.
- get prototype for all nodes if any.
Most of the `gulp` tasks with a "-pre" suffix, especially those related to the main build-targets, are no longer needed and can simply be inlined in their corresponding tasks. This is likely a left-over from older Gulp-versions, but currently it only adds unnecessary indirection.
Furthermore, a number of the `gulp` tasks which are only used *internally* are converted to standard functions instead. This prevents them from being exposed "globally" and e.g. listed by `gulp default`.
This code was added in PR 3968, apparently in order to fix scrolling of search results in HiDPI-mode.
However, after PR 4570 nothing is setting these `dataset`-properties any more and this is thus dead code which should be removed. (If that change had broken scrolling of search results in HiDPI-mode, you'd really expect that it'd been reported and fixed a long time ago.)
I missed this during review, since some of the changes in `web/pdf_print_service.js` broke printing.
Also, as part of these changes the patch replaces what looks like unnecessary `setAttribute` usage with "regular" `className` assignment and finally updates a couple of the CSS-rules to be more consistent.
Given that the same `PartialEvaluator`-instance is used for a lot of these unit-tests, manually changing the options in any one test-case could lead to intermittently failing unit-tests since they're run in a random order.
To fix this, we simply have to use the existing method to clone the `PartialEvaluator`-instance but with the custom options.
This resolves an issue I had when loading pdf.js with webpack. `pdfjsLib` was always `undefined` after the import, and I dug deep into pdf.js's `webpack.js` file to sort out this issue, making all sorts of changes.
I figured out that there was just something wrong with the export/import procedure. After making this change, I now have the full pdfjs object in my webpack application.
For HighlightAnnotations with a built-in appearance stream, we still rely on it to specify the opacity correctly via a suitable blend mode. However, if the Annotation-drawing operators are placed *within* a /XObject of the /Form-type, the /ExtGState won't apply to the final rendering and the result is that the highlighting obscures the underlying text.
The more *correct* and general solution would likely be to somehow modify the implementation in `src/display/canvas.js`, to special-case handling of /Form-type /XObjects when rendering Annotations. Since we can very easily work-around this problem for now by using the "no appearance stream" code-path, doing *something* here ought to be preferable.
This patch is (obviously) merely a work-around, but given that the referenced issue is (as far as I know) the first case we've seen of this problem a simple solution will hopefully suffice for now.
This fixes the colours, by respecting the strokeAlpha/fillAlpha-values, for a couple of Annotations in the PDF document from issue 13447.[1]
---
[1] Some of the annotations still won't render at all, when compared with Adobe Reader, but that could/should probably be handled separately.
- for xfa rendering, fonts are loaded and used in html;
- when printed and saved in pdf, on linux, Firefox uses cairo backend
- when subsetting a font, cairo uses the font postscript name and when this one is empty that leads to a bug
(the append at 63f0d62684/src/cairo-cff-subset.c (L2049) is failing because of null length)
- so this patch adds a postscript name to the font to make cairo happy.
Currently `charsCache` is initialized *lazily*, which considering that it just contains a simple `Object` doesn't seem entirely necessary. This first of all forces us to do repeated exists-checks in the `Font.charsToGlyphs` method, and secondly the similar/related `glyphCache` is already initialized eagerly.
Furthermore, this patch also does a bit of clean-up in the `Font.charsToGlyphs` method since this code is quite old.
- the only goal of this patch is to be able to get synchronously the fake html when printing from firefox:
- in order to print we need to inject some html in beforeprint callback but we cannot block in waiting for all the pages.
- from a memory point of view: it doesn't change anything since the fake HTML is deleted in the worker;
- this way we don't break any assumptions.
- I thought it was possible to rely on browser layout engine to handle layout stuff but it isn't possible
- mainly because when a contentArea overflows, we must continue to layout in the next contentArea
- when no more contentArea is available then we must go to the next page...
- we must handle breakBefore and breakAfter which allows to "break" the layout to go to the next container
- Sometimes some containers don't provide their dimensions so we must compute them in order to know where to put
them in their parents but to compute those dimensions we need to layout the container itself...
- See top of file layout.js for more explanations about layout.
- fix few bugs in other places I met during my work on layout.
This code is old and predates the improvements we made to the test
manifest to only contain working URLs (either Web Archive or
GitHub/Bugzilla links), so the fallback logic to try the Web Archive is
no longer necessary. This greatly simplifies the function and also
makes sure that we fail directly in case a bad URL is added to the
manifest, instead of having it work "accidentally" because of this
logic, since we want the manifest to be correct at all times (and
otherwise fail loudly).
Given that `URL`s aren't supported by the structured clone algorithm, see https://developer.mozilla.org/en-US/docs/Web/API/Web_Workers_API/Structured_clone_algorithm, the document in issue 1773 will cause the browser to throw `DataCloneError: The object could not be cloned.`-errors and nothing will render.
To fix this, we'll instead simply send the stringified version of the `URL` to prevent these errors from occuring.
The building of glyph paths, in the `FontRendererFactory`, can fail in various ways for corrupt font data. However, we're currently not attempting to handle any such errors in the evaluator, which means that a single broken glyph *can* prevent an entire page from rendering.
To address this we simply have to pass along, and check, the existing `ignoreErrors` option in `PartialEvaluator.buildFontPaths` similar to the rest of the `PartialEvaluator` code.
According to the specification, see https://web.archive.org/web/20210404042322if_/https://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#G6.2384179, the keys of a NameTree/NumberTree should be ordered.
For corrupt PDF files, which violate this assumption, it's thus possible that trying to lookup a single entry fails.
Previously, in PR 10274, we implemented a fallback that only applies to the "bottom" node of a NameTree/NumberTree, which in general might not actually help for sufficiently corrupt NameTree/NumberTree data.
Instead we remove the current *limited* fallback from `NameOrNumberTree.get`, and defer to the call-site to handle this case explicitly e.g. by using `NameOrNumberTree.getAll` for data where that makes sense. For well-formed documents, these changes should *not* lead to any additional data fetching/parsing.
Finally, as part of these changes, the validation of named destination data is improved in the `Catalog` and a new unit-test is also added.
This patch fixes the referenced bugs/issues, in a way that won't interfere with keyboard users, assuming that we actually want to fix these old bugs/issues. (If not, we should close them as WONTFIX.)
To get the maximum benefit from something like Prettier, you obviously don't want to disable the automatic formatting unless absolutely necessary. When we added Prettier there were a number of cases, mostly involving larger Arrays, which required disabling of the automatic formatting for overall readability and/or to not break inline comments.
With changes in Prettier version `2.3.0`, see [the release notes](https://prettier.io/blog/2021/05/09/2.3.0.html#concise-formatting-of-number-only-arrays-10106httpsgithubcomprettierprettierpull10106-10160httpsgithubcomprettierprettierpull10160-by-thorn0httpsgithubcomthorn0), there's now better formatting support for Arrays containing only numbers. Hence we can now remove a number of `// prettier-ignore` comments, and thus get the benefit of automatic formatting in (slightly) more of the code-base.
- in charstring specs at page 21 (section 4.2): "Also, it may appear in the charstring as the difference from nominalWidthX" so the number we've on the stack doesn't have to be positive.
- currently this bug has probably no visible effect
- but when the font is loaded to be used with XFA, then the rendering is incorrect.
As can be seen in PR 13371, some of the `no-var` changes in the `PartialEvaluator.{getOperatorList, getTextContent}` methods caused errors in `gulp server`-mode.
However, there's a handful of instances of `var` in other methods which should be completely *safe* to convert since there's no strange scope-issues present in that code.
With modern JavaScript modules, where only *explicitly* exported properties are visible to the outside, the `QueueOptimizerClosure` should no longer be necessary.
Furthermore, to reduce the possibility of `NullOptimizer` and `QueueOptimizer` getting out of sync (note e.g. the inconsistency fixed in PR 10784), we now let the latter extend the former one.
This patch replaces the old structure with an abstract base-class, which the new ShadingPattern classes then inherit from.
The old `createMeshCanvasClosure` can now be removed, since it's not necessary any more with modern JavaScript, and the `createMeshCanvas` function is now instead a method on the new `MeshShadingPattern` class (avoids unnecessary parameter passing).
With the changes in PR 13361, we're now using the `CanvasPattern.setTransform()` method when rendering certain Shadings/Patterns.
Note that while `CanvasPattern` itself has been supported since basically "forever", its `setTransform` method is a slightly newer addition to the specification; please refer to https://developer.mozilla.org/en-US/docs/Web/API/CanvasPattern#browser_compatibility
Rather than trying to re-write PR 13361 to not use, or possibly spending time/effort (if possible) polyfilling, `CanvasPattern.setTransform()` this patch thus suggests that we simply update the *minimum* supported browser versions instead.
According to the compatibility data linked above, the *minimum* supported browser versions in the PDF.js library are now as follows:
- Chrome >= 68, which was released on 2018-07-24.[1]
- Firefox ESR, see https://wiki.mozilla.org/Release_Management/Calendar.
- Safari >= 11.1, which was release on 2018-03-29.[2]
(Given that the PDF.js contributors cannot realistically test a bunch of old browsers, it's not unimaginable that some older browser versions are already not working with the PDF.js library.)
Based on these changes, which we should ensure are reflected in the Wiki as well, we can also remove a number of now redundant polyfills. Furthermore we'll no longer "claim" to support Windows XP, note the `gulpfile.js` changes, which should definitely *not* be an issue given that it's no longer officially supported.[3]
---
[1] According to https://en.wikipedia.org/wiki/Google_Chrome_version_history
[2] According to https://en.wikipedia.org/wiki/Safari_version_history#Safari_11
[3] According to https://en.wikipedia.org/wiki/Windows_XP#End_of_support
This patch replaces the old structure with a abstract base-class, which the new RadialAxial/Mesh-shading classes then inherit from.[1]
The old `MeshClosure` can now be removed, since it's not necessary any more, and most of the functions inside of it are now instead methods on the new `MeshShading` class. This is particularly nice, in my opinion, since we previously were *manually* passing around a reference to the current `Mesh`-instance.
---
[1] If we want/need to, in the future, split e.g. the Mesh-handling into multiple classes that should now be easy to do.
Neither the `type` or the `cs` properties are used outside of the "constructors", and we can thus remove them.[1]
Note that a lot of this code is very old, and that it actually predates the main/worker-thread split before which the *same* file was used on both the main- *and* worker-threads.
---
[1] On the main-thread, a similar `type` property was removed in PR 12591.
- Right now, a glyph with an erroneous outline is replaced by an empty glyph
if the error is far enough from the start there's likely something to render
so the idea is to replace a command with args by an endchar when no args are
on the stack: this way OTS is likely happy (no remaining args on stack) and we
can draw something which is likely better than nothing.
First of all, by using `Dict.getArray` in the `Page.content` getter we remove the need to manually iterate through and fetch the sub-streams (when they exist) in the `Page.getContentStream` method.
Secondly, we can simplify the code in `Page.{getOperatorList, extractTextContent}` by letting `Page.getContentStream` ensure that `content` is available and returning a Promise instead.
Similar to the `get`/`getAsync` methods, this should be a *tiny* bit more efficient which cannot hurt considering that `getArray` is now used a lot more than when initially added.
This reverts commit 0ef9b5aafc88094f19fec793c174c622e7e15542, since it cases a lot of warnings (see below) *locally* with e.g. the document from issue 9627.
Strangely enough, this only occurs with `gulp server`-mode and the actual builds are apparently fine. It seems that this *may* be some unfortunate interaction with the old Babel-plugin that's used together with SystemJS.
```
Warning: getTextContent - ignoring ExtGState: "FormatError: ExtGState should be a dictionary.".
```
Rather than taking the risk that this could actually cover a more serious bug, and since I cannot immediately figure out what's wrong, it thus seem safest to revert this for now and we can (carefully) revisit this once SystemJS has been removed (see PR 12563).
This patch was tested using the PDF file from issue 2618, i.e. https://bug570667.bugzilla-attachments.gnome.org/attachment.cgi?id=226471, with the following manifest file:
```
[
{ "id": "issue2618",
"file": "../web/pdfs/issue2618.pdf",
"md5": "",
"rounds": 50,
"type": "eq"
}
]
```
which gave the following results when comparing this patch against the `master` branch:
```
-- Grouped By browser, stat --
browser | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ------------ | ----- | ------------ | ----------- | --- | ---- | -------------
firefox | Overall | 50 | 3417 | 3426 | 9 | 0.27 |
firefox | Page Request | 50 | 1 | 1 | 0 | 5.41 |
firefox | Rendering | 50 | 3416 | 3426 | 9 | 0.27 |
```
Based on these results, there's no significant performance regression from using standard classes and this patch should thus be OK.
Previously, we set the base transformation and pattern matrix
directly to the main rendering ctx of the page, however doing this
caused the current transform to be lost. This would cause issues
with things like shear missing so the pattern was misaligned or when
stroke was used the scale of the line width or dash would be wrong.
Instead we should leave the current transform and use setTransfrom
on the pattern so it is applied correctly. For axial and radial shadings I had
to create a temporary canvas to draw the shading so I could in turn
use setTransform.
Fixes: #13325, #6769, #7847, #11018, #11597, #11473
The following already in the corpus are improved:
issue8078-page1
issue1877-page1
This patch moves the `PDFJSDev`-checks *inline*, similar to the rest of the code-base, such that the code in question is actually being removed from the *built* files in e.g. the official releases.
After PR 13117 it's now (finally) possible for *different* build targets to specify individual options/preferences, and we can utilize that to only expose the `renderer`-preference in builds where `SVGGraphics` is actually defined.
Note that for e.g. `MOZCENTRAL`-builds, trying to enable SVG-rendering will throw immediately and the preference thus doesn't make sense to include there.
Also, update the dummy `SVGGraphics` to use a class, tweak the `PDFJSDev`-check in `src/display/svg.js` to agree fully with the option/preference, and remove an unnecessary `eslint-disable`.
Reasons for the removal include:
- This functionality was always somewhat experimental and has never been enabled by default, partly because of worries about rendering bugs caused by e.g. bad/outdated graphics drivers.
- After the initial implementation, in PR 4286 (back in 2014), no additional functionality has been added to the WebGL implementation.
- The vast majority of all documents do not benefit from WebGL rendering, since only a couple of *specific* features are supported (e.g. some Soft Masks and Patterns).
- There is, and has always been, *zero* test-coverage for the WebGL implementation.
- Overall performance, in the PDF.js library, has improved since the experimental WebGL implementation was added.
Rather than shipping unused *and* untested code, it seems reasonable to simply remove the WebGL implementation for now; thanks to version control it's always possible to bring back the code should the need ever arise.
This functionality was originally implemented in PR 7029; however it's not, nor has it ever been, used as far as I can tell.[1]
Note in particular that the default viewer does not expose either a preference or even an option with which `disableCanvasToImageConversion` can be toggled, and source-code modification is thus required.
Furthermore, note also that we have multiple other instances of `canvas`-data accesses in both the `src/display/canvas.js` and `src/display/text_layer.js` files. If any of those are blocked, by e.g. browser settings, there will be outright rendering bugs and non-working thumbnails thus seem like a very small issue in the grand scheme of things; hence why I'm suggesting that we remove the unused `disableCanvasToImageConversion` functionality.
---
[1] For the Tor use-case mentioned in issue 7026, I *believe* that the solution was to white-list `canvas`-data accesses for its built-in PDF Viewer.
Compared to other data-structures, such as e.g. `Dict`s, we're purposely *not* caching Streams on the `XRef`-instance.[1]
The, somewhat unfortunate, effect of Streams not being cached is that repeatedly getting the *same* Stream-data requires re-parsing/re-initializing of a bunch of data; see `XRef.fetch` and related methods.
For the font-parsing in particular we're currently fetching the `toUnicode`-data, which is very often a Stream, in `PartialEvaluator.preEvaluateFont` and then *again* in `PartialEvaluator.extractDataStructures` soon afterwards.
By instead letting `PartialEvaluator.preEvaluateFont` export the "raw" `toUnicode`-data, we can avoid *some* unnecessary re-parsing/re-initializing when handling fonts.
*Please note:* In this particular case, given that `PartialEvaluator.preEvaluateFont` only accesses the "raw" `toUnicode` data, exporting a Stream should be safe.
---
[1] The reasons for this include:
- Streams, especially `DecodeStream`-instances, can become *very* large once read. Hence caching them really isn't a good idea simply because of the (potential) memory impact of doing so.
- Attempting to read from the *same* Stream-instance more than once won't work, unless it's `reset` in between, since using any method such as e.g. `getBytes` always starts at the current data position.
- Given that parsing, even in the worker-thread, is now fairly asynchronous it's generally impossible to assert that any one Stream-instance isn't being accessed "concurrently" by e.g. different `getOperatorList` calls. Hence `reset`-ing a cached Stream-instance isn't going to work in the general case.
Rather than re-fetching/re-parsing these properties immediately in `PartialEvaluator.translateFont`, we can simply export them instead. (Obviously the effect will be really tiny, but there is less parsing overall this way.)
*Please note:* While I don't have a document that this patches fixes, the current code is however not entirely correct as far as I can tell.
Looking at how the `Widths` array is parsed in `PartialEvaluator.extractWidths`, it's clear that the implementation in `PartialEvaluator.preEvaluateFont` is a bit too simplistic. In particular, by only wrapping the data into a TypedArray, there's no attempt to handle *indirect* objects which could potentially lead to colliding `hash`es being computed.
To hopefully help prevent any future bugs, make sure that composite/non-composite fonts cannot accidentally get matching `hash`es. Given the differences between those font types, that's very unlikely to be useful or even correct in general.
Without this some *composite* fonts may incorrectly end up with matching `hash`es, thus breaking rendering since we'll not actually try to load/parse some of the fonts.
*Please note:* Given that the document, in the referenced issue, doesn't embed *any* of its fonts there's no guarantee that it renders correctly in all configurations even with this patch.
With modern JavaScript modules, where you explicitly list the properties that should be exported, it's no longer necessary to wrap *all* of the code within one file into a top-level closure.[1]
This patch reduces the size, of even the *built* `pdf.worker.js` file, since there's now a lot less unnecessary whitespace.
---
[1] For files which contain *different* functionality, some closures may however still make sense in order to separate the code.
It might be possible to remove some of those cases later, e.g. once private class fields becomes generally available/usable in browsers.
With modern JavaScript modules, where you explicitly list the properties that should be exported, it's no longer necessary to wrap all of the code in a closure.[1]
This patch also tries to clean-up/improve a couple of the existing JSDoc-comments.
---
[1] This reduces the size, even of the *built* `pdf.js` file, since there's now a lot less unnecessary whitespace.
*My apologies for breaking this; thankfully PR 13303 hasn't reach mozilla-central yet.*
It's (obviously) necessary to initialize a `PredictorStream`-instance fully, since otherwise breakage may occur if there's errors during the actual stream parsing.
To reproduce this issue, try opening the PDF document from issue 13051 locally and observe the following message in the console:
```
Warning: Invalid stream: "ReferenceError: this hasn't been initialised - super() hasn't been called"
```
- app.alert and few other function can use an object as parameter ({cMsg: ...});
- support app.alert with a question and a yes/no answer;
- update field siblings when one is changed in an action;
- stop calculation if calculate is set to false in the middle of calculations;
- get a boolean for checkboxes when they've been set through annotationStorage instead of a string.
Also, removes a couple of unnecessary local variables in the `Stepper.breakIt` method.
Finally, this patch also disables the ESLint `no-var` rule, in preparation for the next patch, for a couple of data-structures that need to remain globally available.
It shouldn't be possible for the `getBytes`-call to throw a `MissingDataException`, since all resources are loaded *before* e.g. font-parsing ever starts; see f0817015bd/src/core/object_loader.js (L111-L126)
Furthermore, even if we'd *somehow* re-throw a `MissingDataException` here that still won't help considering where the `Type1Font`-instance is created. Note how in the `Font`-constructor we simply catch any errors and fallback to a standard font, which means that a `MissingDataException` would just lead to rendering errors anyway; see f0817015bd/src/core/fonts.js (L648-L691)
All-in-all, it's not possible for a `MissingDataException` to be thrown in `getHeaderBlock` and this code-path can thus be removed.
Obviously the `Font`-class is still *very* large, given particularly how TrueType fonts are handled, however this patch-series at least improves things by moving a number of functions/classes into their own files.
As a follow-up it might make sense to try and re-factor/extract the TrueType parsing into its own file, since all of this code is quite old, however that's probably best left for another time.
For e.g. `gulp mozcentral`, the *built* `pdf.worker.js` files decreases from `1 620 332` to `1 617 466` bytes with this patch-series.
These changes were made automatically, using `gulp lint --fix`.
Given the large size of this patch, the manual fixes are done separately in the next commit.
These changes were made *mostly* automatically, using `gulp lint --fix`, with the following manual changes:
```diff
diff --git a/src/core/fonts_utils.js b/src/core/fonts_utils.js
index f88ce4a8c..c4b3f3808 100644
--- a/src/core/fonts_utils.js
+++ b/src/core/fonts_utils.js
@@ -167,8 +167,8 @@ function type1FontGlyphMapping(properties, builtInEncoding,
glyphNames) {
}
// Lastly, merge in the differences.
- let differences = properties.differences,
- glyphsUnicodeMap;
+ const differences = properties.differences;
+ let glyphsUnicodeMap;
if (differences) {
for (charCode in differences) {
const glyphName = differences[charCode];
```
- `FontFlags`, is used in both `src/core/fonts.js` and `src/core/evaluator.js`.
- `getFontType`, same as the above.
- `MacStandardGlyphOrdering`, is a fairly large data-structure and `src/core/fonts.js` is already a *very* large file.
- `recoverGlyphName`, a dependency of `type1FontGlyphMapping`; please see below.
- `SEAC_ANALYSIS_ENABLED`, is used by both `Type1Font`, `CFFFont`, and unit-tests; please see below.
- `type1FontGlyphMapping`, is used by both `Type1Font` and `CFFFont` which a later patch will move to their own files.
This is done automatically with `gulp lint --fix` and the following
manual changes:
```diff
diff --git a/src/core/function.js b/src/core/function.js
index 878001057..b7e3e6ccf 100644
--- a/src/core/function.js
+++ b/src/core/function.js
@@ -131,7 +131,7 @@ function toNumberArray(arr) {
return arr;
}
-var PDFFunction = (function PDFFunctionClosure() {
+const PDFFunction = (function PDFFunctionClosure() {
const CONSTRUCT_SAMPLED = 0;
const CONSTRUCT_INTERPOLATED = 2;
const CONSTRUCT_STICHED = 3;
@@ -484,7 +484,9 @@ var PDFFunction = (function PDFFunctionClosure() {
// clip to domain
const v = clip(src[srcOffset], domain[0], domain[1]);
// calculate which bound the value is in
- for (var i = 0, ii = bounds.length; i < ii; ++i) {
+ const length = bounds.length;
+ let i;
+ for (i = 0; i < length; ++i) {
if (v < bounds[i]) {
break;
}
@@ -673,23 +675,21 @@ const PostScriptStack = (function PostScriptStackClosure() {
roll(n, p) {
const stack = this.stack;
const l = stack.length - n;
- let r = stack.length - 1,
- c = l + (p - Math.floor(p / n) * n),
- i,
- j,
- t;
- for (i = l, j = r; i < j; i++, j--) {
- t = stack[i];
+ const r = stack.length - 1;
+ const c = l + (p - Math.floor(p / n) * n);
+
+ for (let i = l, j = r; i < j; i++, j--) {
+ const t = stack[i];
stack[i] = stack[j];
stack[j] = t;
}
- for (i = l, j = c - 1; i < j; i++, j--) {
- t = stack[i];
+ for (let i = l, j = c - 1; i < j; i++, j--) {
+ const t = stack[i];
stack[i] = stack[j];
stack[j] = t;
}
- for (i = c, j = r; i < j; i++, j--) {
- t = stack[i];
+ for (let i = c, j = r; i < j; i++, j--) {
+ const t = stack[i];
stack[i] = stack[j];
stack[j] = t;
}
@@ -939,7 +939,7 @@ class PostScriptEvaluator {
// We can compile most of such programs, and at the same moment, we can
// optimize some expressions using basic math properties. Keeping track of
// min/max values will allow us to avoid extra Math.min/Math.max calls.
-var PostScriptCompiler = (function PostScriptCompilerClosure() {
+const PostScriptCompiler = (function PostScriptCompilerClosure() {
class AstNode {
constructor(type) {
this.type = type;
```
The `this.data` property is, when defined, sent from the worker-thread as a `Uint8Array` and there's thus no reason to re-initialize the TypedArray here.
Note also the `FontFaceObject.createNativeFontFace` method just above, where we simply use `this.data` as-is.
The explanation for this code looking like it does is, as is often the case, for historical reasons. Originally we only supported `@font-face`, before the Font Loading API existed, and back then we also polyfilled TypedArrays (using regular Arrays) which should explain this particular line of code.
Given that the `bytesToString(BaseStream.getBytes(...))` pattern is somewhat common throughout the `src/core/` code, it cannot hurt to add a new `BaseStream`-method which handles that case internally.
- Improve chunking in order to fix some bugs where the spaces aren't here:
* track the last position where a glyph has been drawn;
* when a new glyph (first glyph in a chunk) is added then compare its position with the last saved one and add a space or break:
- there are multiple ways to move the glyphs and to avoid to have to deal with all the different possibilities it's a way easier to just compare positions;
- and so there is now one function (i.e. "compareWithLastPosition") where all the job is done.
- Add some breaks in order to get lines;
- Remove the multiple whites spaces:
* some spaces were filled with several whites spaces and so it makes harder to find some sequences of words using the search tool;
* other pdf readers replace spaces by one white space.
Update src/core/evaluator.js
Co-authored-by: Jonas Jenwald <jonas.jenwald@gmail.com>
Co-authored-by: Jonas Jenwald <jonas.jenwald@gmail.com>
Given that both the textLayer rendering *and* the structTree parsing is asynchronous, it's possible that we'll attempt to insert the structTree in a removed page. While there's thankfully no outright breakage caused by this, it will nonetheless lead to errors being printed in the console and we should obviously avoid this.
To reproduce this bug (without the patch), open http://localhost:8888/web/viewer.html?file=/test/pdfs/pdf.pdf#disableStream=true&disableAutoFetch=true and scroll *very quickly* through the document and notice the following error being (intermittently) printed in the console:
```
Uncaught (in promise) TypeError: can't access property "appendChild", this.canvas is undefined
```
Looking at the `ChunkedStream` implementation, it's basically a "regular" `Stream` but with added functionality in order to deal with fetching/loading of missing data.
Hence, by letting `ChunkedStream` extend `Stream`, we can remove some duplicate methods from the `ChunkedStream` class.
For all of the other `DecodeStream`s we're not passing in a `Dict`-instance manually, but instead get it from the `stream`-parameter. Hence there's no particularly good reason, as far as I can tell, to not do the same thing in `Jbig2Stream`/`JpegStream`/`JpxStream` as well.
The way that `getBaseStreams` is currently handled has bothered me from time to time, especially how we're checking if the method exists before calling it.
By adding a dummy `BaseStream.getBaseStreams` method, and having the call-sites simply check the return value, we can improve some of the relevant code.
Note in particular how the `ObjectLoader._walk` method didn't actually check that the data in question is a Stream instance, and instead only checked the `currentNode` (which could be anything) for the existence of a `getBaseStreams` property.
By having an abstract base-class, it becomes a lot clearer exactly which methods/getters are expected to exist on all Stream instances.
Furthermore, since a number of the methods are *identical* for all Stream implementations, this reduces unnecessary code duplication in the `Stream`, `DecodeStream`, and `ChunkedStream` classes.
For e.g. `gulp mozcentral`, the *built* `pdf.worker.js` files decreases from `1 619 329` to `1 616 115` bytes with this patch-series.
Given that we're using modules, meaning that only explicitly `export`ed things are visible to the outside, it's no longer necessary to wrap all of the code in a closure.
This gets rid of *a lot* of boilerplate that stems from our old way of simulating classes, and it actually reduces the filesize noticeably.
For e.g. `gulp mozcentral`, the *built* `pdf.js` files decreases from `318 404` to `314 722` bytes (~1 percent) with this patch.
This is done automatically with `gulp lint --fix` and the following
manual changes:
```diff
diff --git a/src/core/image.js b/src/core/image.js
index 35c06b8ab..e718b9937 100644
--- a/src/core/image.js
+++ b/src/core/image.js
@@ -97,7 +97,7 @@ class PDFImage {
if (isName(filter)) {
switch (filter.name) {
case "JPXDecode":
- var jpxImage = new JpxImage();
+ const jpxImage = new JpxImage();
jpxImage.parseImageProperties(image.stream);
image.stream.reset();
```
Since we really don't want to let a particular Webpack-mode dictate how we can/can't write code, let's add a note in the webpack-example README about minification instead.
Using `for...of` is a modern and generally much nicer pattern, since it gets rid of unnecessary callback-functions. (In a couple of spots, a "regular" `for` loop had to be used.)
- In case of large string the sandbox initialization failed because of an OOM
* so allocate a new string in the heap
* and free it after use.
- it requires a quickjs update since we need to export some symbols (stringToNewUTF8 and free).
This patch first of all moves all checking/validation into the `appendIfJavaScriptDict` function, to avoid duplicating it in multiple places. Secondly, also removes what's now an outdated/incorrect comment since we have implemented scripting support.
Given that we're (almost) always iterating through the result of the `getAll`-calls, using a `Map` seems nicer overall since it's more suited to iteration compared to a regular Object.
Also, add a couple of `Dict`-checks in existing code touched by this patch, since it really cannot hurt to prevent *potential* errors in a corrupt PDF document.
First of all, while it should be very unlikely that the /ID-entry is an *indirect* object, note how we're using `Dict.get` when parsing it e.g. in `PDFDocument.fingerprint`. Hence we definitely should be consistent here, since if the /ID-entry is an *indirect* object the existing code in `src/core/writer.js` would already fail.
Secondly, to fix the referenced issue, we also need to check that the /ID-entry actually is an Array before attempting to access its contents in `src/core/writer.js`.
*Drive-by change:* In the `xrefInfo` object passed to the `incrementalUpdate` function, re-name the `encrypt` property to `encryptRef` since its data is fetched using `Dict.getRaw` (given the names of the other properties fetched similarly).
For CFF fonts without proper `ToUnicode`/`Encoding` data, utilize the "charset"/"Encoding"-data from the font file to improve text-selection (issue 13260)
Keeps screen readers from pausing on every span so
paragraphs are read more naturally. Note: this only seems
to affect Firefox, Chrome automatically combines the spans.
By replacing `XMLHttpRequest` with a `fetch` call, the helper function can be modernized to use async/await instead.
Note that the headers doesn't seem necessary to set now, since:
- The Fetch API provides a method for accessing the response as *text*, which renders the "Content-type" header unnecessary.
- According to https://developer.mozilla.org/en-US/docs/Glossary/Forbidden_header_name, the "Content-length" header isn't necessary.
By not waiting for the /Properties to load, before parsing of the operatorList/textContent starts, there's a very real risk that a `MissingDataException` will be thrown when trying to access the data in the `PartialEvaluator.parseMarkedContentProps` method.
If this ever happens it will thus lead to incomplete and/or outright broken rendering, and with e.g. `disableAutoFetch=true` set the likelihood of this occuring would increase quite a bit.
*Please note:* While I've not yet seen this error in an actual PDF document, it can happen during loading if you're unlucky enough with e.g. the structure of the PDF document and/or the download speed offered by the server.
- Use `PDFManager.ensureDoc`, rather than `PDFManager.ensure`, in a couple of spots in the code. If there exists a short-hand format, we should obviously use it whenever possible.
- Fix a unit-test helper, to account for the previous changes. (Also, converts a function to be `async` instead.)
- Add one more exists-check in `PDFDocument.loadXfaFonts`, which I missed to suggest in PR 13146, to prevent any possible errors if the method is ever called in a situation where it shouldn't be.
Also, print a warning if the actual font-loading fails since that could help future debugging. (Finally, reduce overall indentation in the loop.)
- Slightly unrelated, but make a small tweak of a comment in `src/core/fonts.js` to reduce possible confusion.
While I wasn't able to figure out *exactly* why the old format didn't work, re-factoring the `parseOptions` function to use `yargs` differently "just worked" so that's hopefully good enough here.
With these changes everything related to a *particular* option now appears in one place, rather than being spread out, which aids readability in my opinion. Also, the options are now sorted alphabetically, to make it easier to find a particular one.
https://www.npmjs.com/package/yargs
According to a decision by UX and PM, please see https://bugzilla.mozilla.org/show_bug.cgi?id=1705060#c2 (and implemented in https://bugzilla.mozilla.org/show_bug.cgi?id=1705327), we no longer show the notification-bar in Firefox; hence the special `PDFViewerApplication._delayedFallback` functionality should no longer be necessary.
Furthermore, note that at this point in time *most* of the features which used the `PDFViewerApplication._delayedFallback` functionality is now enabled by default; hence that provides even less reason to keep this code around and existing calls are thus converted to "regular" `PDFViewerApplication.fallback` calls.
According to a decision by UX and PM, please see https://bugzilla.mozilla.org/show_bug.cgi?id=1705060#c2, in Firefox we should first of all *not* display the notification-bar for signatures. Secondly, as can also be seen there, we shouldn't display the notification-bar *at all* and it's thus disabled in https://bugzilla.mozilla.org/show_bug.cgi?id=1705327.
If we purposely don't display a notification, for documents with signatures, in the *built in* Firefox PDF Viewer then it cannot be necessary in the GENERIC viewer either.
- Different fonts can be used in xfa and some of them are embedded in the pdf.
- Load all the fonts in window.document.
Update src/core/document.js
Co-authored-by: Jonas Jenwald <jonas.jenwald@gmail.com>
Update src/core/worker.js
Co-authored-by: Jonas Jenwald <jonas.jenwald@gmail.com>
There is no asynchronous code involved here, so we can get rid of all
done callbacks here and simply use the fact that if the function call
ends without failed assertion that the test passed.
There is no asynchronous code involved here, so we can get rid of all
done callbacks here and simply use the fact that if the function call
ends without failed assertion that the test passed.
I've successfully run `gulp mozcentral` and `gulp generic` locally, with/without this patch and diff-ed the *built* `web/viewer.css` files. There were no changes, in the build-output, caused by this update.
https://www.npmjs.com/package/postcss-calc
The size of the `src/core/obj.js` file has increased slowly over the years, and it also contains a fair amount of *distinct* functionality.
In order to improve readability and make it easier to navigate through the code, this patch moves the `XRef` into its own file.
The size of the `src/core/obj.js` file has increased slowly over the years, and it also contains a fair amount of *distinct* functionality.
In order to improve readability and make it easier to navigate through the code, this patch moves `NameTree`/`NumberTree` into its own file.
The size of the `src/core/obj.js` file has increased slowly over the years, and it also contains a fair amount of *distinct* functionality.
In order to improve readability and make it easier to navigate through the code, this patch moves the `FileSpec` into its own file.
The size of the `src/core/obj.js` file has increased slowly over the years, and it also contains a fair amount of *distinct* functionality.
In order to improve readability and make it easier to navigate through the code, this patch moves the `ObjectLoader` into its own file.
`setSelectionRange(0, 0)` added in 44b24fcc29 for #12359, required only by Firefox ([bug](https://bugzilla.mozilla.org/show_bug.cgi?id=860329)), causes issues mozilla#13191, mozilla#12592 in Safari.
`scrollLeft = 0` is a fix that breaks the focus trap in Safari while **keeping Firefox behavior same for #12359**.
Remove the unused "GetIsPureXfa" message handler; and avoid unnecessary parsing when no structTree is available (PR 13069 follow-up, PR 13221 follow-up)
With the current implementation of `PDFDocument.hasJSActions`, in the worker-thread, we're not actually handling not-yet-loaded data correctly. This can thus fail in *two* different ways:
- The `PDFDocument.fieldObjects` getter (and its helper method), while it may *return* a Promise, still fetches all of its data synchronously and it can thus throw a `MissingDataException` during parsing.
- The `Catalog.jsActions` getter, which is completely synchronous, can obviously throw a `MissingDataException` during parsing.
If either of these cases occur currently, the `PDFDocumentProxy.hasJSActions` method in the API can either return a *rejected* Promise (which it never should) or possibly "hang" and never resolve.
*Please note:* While I've not *yet* seen this error in an actual PDF document, it can happen during loading if you're unlucky enough with e.g. the structure of the PDF document and/or the download speed offered by the server.
This patch is thus based on code-inspection *and* on manually throwing a `MissingDataException` on the first access of `Catalog.jsActions` to simulate this situation.
Finally, this patch adds a couple of *API* unit-tests for this (since none existed).
Given that this only an internal helper method, used by the `Catalog.{javaScript, jsActions}` getters, this change simplifies iteration of the returned data.
We can also (slightly) re-factor the code of the `jsActions` getter, and remove an obsolete[1] JSDoc-comment from the `openAction` getter.
---
[1] Not really relevant now that we've got proper scripting support.
Similar to all other data accesses, note e.g. the "GetDocJSActions" handler just above, we need to ensure that a `MissingDataException` isn't propagated to the main-thread if this data is accessed while the PDF document is still loading.
It's obviously (a bit) more efficient to return early in `Page.getStructTree`, rather than trying to first "parse" an *empty* structTree-root.
*Somehow I didn't think of this yesterday, but this feels like a much better solution overall; sorry about the churn here!*
Looking at the API, there's no code which actually sends this message. Most likely it's a left-over from a previous version of PR 13069, since the `isPureXfa` parameter is being included in the "GetDoc" message.
To simplify the overall implementation, given that it only applies to the GENERIC-viewer, this patch purposely re-uses the existing `errorWrapper`-functionality to display the message.
While that one is mostly intended for actual *errors*, by re-using it here we considerably reduce the amount of code/complexity necessary for supporting this new warning. It's obviously possible to re-factor/improve this later on, but the patch should do just fine here since it'll indeed inform users (of the GENERIC-viewer) about unverified signatures.
Finally this patch also tweaks the background-color of the `errorWrapper`, making it 20 percent lighter respectively darker (depending on the theme) to make it "stand out" a little bit *less*.[1] While it may perhaps be useful to re-style/re-factor the `errorWrapper`, this patch probably isn't the right place for doing that.
---
[1] Note how in the MOZCENTRAL-viewer, which instead uses the browser notification-bar, we're purposely using a neutral colour to not draw too much attention to the notification-bar.
This is first of all consistent with existing API-methods, where we return `null` when the data in question doesn't exist. Secondly, it should also be (slightly) more efficient since there's less dummy-data that we need to transfer between threads.
Finally, this prevents us from adding an empty/unnecessary span to *every* single page even in documents without any structure tree data.
I (unsurprisingly) managed to forget about handling the case where a "pagesloaded" event arrives *before* the outline has been parsed, in which case we'd not actually enable the `currentOutlineButton` as intended.
Also, in the "pagesloaded" event handler, we should ensure that there's actually any pages loaded since otherwise the "find current outlineItem"-feature doesn't make any sense.
The done callbacks are an outdated mechanism to signal Jasmine that a
unit test is done, mostly in cases where a unit test needed to wait for
an asynchronous operation to complete before doing its assertions.
Nowadays a much better mechanism is in place for that, namely simply
passing an asynchronous function to Jasmine, so we don't need callbacks
anymore (which require more code and may be more difficult to reason
about).
In these particular cases though the done callbacks never had any real
use since nothing asynchronous happens in these places. Synchronous
functions don't need to use done callbacks since Jasmine simply knows
it's done when the function reaches its normal end, so we can safely get
rid of these callbacks. The telltale sign is if the done callback is
used unconditionally at the end of the function.
This is all done in an effort to over time get rid of all callbacks in
the unit test code.
- but don't validate them for now;
- Firefox will display a bar to warn that the signature validation is not supported (see https://bugzilla.mozilla.org/show_bug.cgi?id=854315)
- almost all (all ?) pdf readers display signatures;
- validation is done in edge but for now it's behind a pref.
It's obviously better and more correct to handle the "pagesloaded" case within `PDFOutlineViewer` *itself*, rather than essentially splitting the logic in two parts and forcing `PDFSidebar` to deal with what should've been handled internally in `PDFOutlineViewer`.
This is what I *should* have done in PR 12777, but for some reason didn't figure out how to implement it well enough back then; sorry about the churn here!
*This patch fixes some technical debt in the viewer.*
Given that most API methods are (purposely) asynchronous, there's always a risk that the viewer could have been `close`d before the requested data arrives.
Lately we've started to check this case before using the data, to prevent errors and/or inconsistent state, however the outline/attachments/layers fetching and rendering is old enough that it pre-dates those checks.
When a PDF is "marked" we now generate a separate DOM that represents
the structure tree from the PDF. This DOM is inserted into the <canvas>
element and allows screen readers to walk the tree and have more
information about headings, images, links, etc. To link the structure
tree DOM (which is empty) to the text layer aria-owns is used. This
required modifying the text layer creation so that marked items are
now tracked.
Scripting, as implemented, requires access to a complete document/viewer in order to work. Hence it doesn't really make sense to keep the `enableScripting`-option on `PDFPageView`-instances.[1]
---
[1] Note that there's the `PDFSinglePageViewer`, which can be used in cases where you want access to all features/functionality of the viewer but only display *one* page at a time.
Note how we purposely don't expose the `AnnotationStorage`-class directly in the official API (see `src/pdf.js`), since trying to use *multiple* ones simultaneously doesn't really make sense (e.g. in the viewer).
Instead we lazily initialize, and cache, just *one* instance via `PDFDocumentProxy.annotationStorage` which should thus be available internally in the API itself without having to be manually passed to various methods.
To support these changes, the `AnnotationStorage`-instance initialization is moved into the `WorkerTransport`-class to allow both `PDFDocumentProxy` and `PDFPageProxy` to access it.
This patch implements the following simplifications:
- Remove the `annotationStorage`-parameter from `PDFDocumentProxy.saveDocument`, since it's already available internally.
Furthermore, while it's currently possible to call that method without an `AnnotationStorage`-instance, that really does *not* make any sense at all. In this case you're effectively reducing `PDFDocumentProxy.saveDocument` to a "regular" `PDFDocumentProxy.getData` call, but with *a lot* more overhead, which was obviously not the intention of the `PDFDocumentProxy.saveDocument`-method.
- Try to discourage third-party users from calling `PDFDocumentProxy.saveDocument` unconditionally, as a replacement for `PDFDocumentProxy.getData` (note the previous point).
- Replace the `annotationStorage`-parameter, in `PDFPageProxy.render`, with a boolean `includeAnnotationStorage`-parameter which simply indicates if the (internally available) `AnnotationStorage`-instance should be used during rendering (e.g. for printing).
- By removing the need to *manually* provide `annotationStorage`-parameters to various API-methods, using the API should become simpler (e.g. for third-parties) since you no longer need to worry about manually fetching and passing around this data.
The reason for the fairly large discrepancy, in the thumbnail quality, between the `draw`/`setImage`-methods is that in the former case we *directly* render the thumbnails at the final size that they'll appear at in the sidebar. In the latter case, we instead downsize the (generally) much larger "regular" pages.
To address this, I'm thus proposing that we let `PDFThumbnailView.draw` render thumbnails at *twice* their intended size and then downsize them to the final size.
Obviously this will increase *peak* memory usage during thumbnail rendering in `PDFThumbnailView.draw`, since doubling the width/height of a `canvas` will lead to its pixel-count increasing by a factor of `4`. Furthermore, since you need four components per pixel (given that it's RGBA-data), this will thus lead to the *temporary* thumbnail `canvas`-sizes increasing by a factor of `16` during rendering. Hence why rendering thumbnails at their "original" scale, i.e. using something like `PDFPageProxy.getViewport({ scale: 1 });`, would be an absolutely terrible idea!
To reduce the size and scope of these changes, I've tried to re-factor and re-use as much of the existing downsizing-implementation already present in `PDFThumbnailView` as possible.
While this will generally *not* make thumbnails rendered by `PDFThumbnailView.draw` look *identical* to those based on the rendered pages (via `PDFThumbnailView.setImage`), it's a considerable improvement as far as I'm concerned and enough to call the issue fixed.
*Please note:* This patch will not lead to *any* additional overhead, in either memory usage or parsing, for thumbnails which are based on the rendered pages.
A loop is less efficient than just overwriting the content, which is what we've generally been using (for years) in other parts of the code-base (see e.g. `BaseViewer` and `PDFThumbnailViewer`).
These properties are always updated/used together, and there's no other methods which depend on just one of them, hence they're changed into local variables instead.
Looking through the history of this code, it seems they were converted *from* local variables and to properties all the way back in PR 2914; however as far as I can tell from that diff it doesn't seem to have been necessary even back then!?
The fontName, as defined in the PDF document, cannot be found in *any* of the "name"-tables in the TrueType Collection font. To work-around that, this patch adds a *fallback* code-path to allow using an approximately matching fontName rather than outright failing.
While this method has only been deprecated in one releases now, the `AnnotationStorage`-functionality is new enough that third-party implementations hopefully don't rely heavily on it just yet. (And removing this quickly should help reduce the likelihood that someone starts using it.)
When removing tasks we're currently forced to *indirectly* iterate through the array, which can be avoided by using a Set instead.
Furthermore, we can also (slightly) modernize the code responsible for initializing the `renderTasks`.
Given that what we actually want is only to keep track of the loadedFont-names, rather than storing any actual data, using an object isn't really necessary here. Furthermore, in the current code, we're also using `in` when checking if the data exists, which is generally less efficient than just checking for the value directly.
As discussed in the issue, this is a small/simple patch that should help to prevent *outright* data loss in forms when a new document is opened in the GENERIC viewer.
While the implementation is perhaps a bit "simplistic", it does seem to work and should be fine given that this is an edge-case only relevant for the GENERIC viewer.
In the next patch we'll need to be able to actually wait for saving to complete, hence it's necessary to slightly re-factor the `save`-method.
As part of these changes, we can reduce some duplication in the `save`-method and slightly improve the overall code. For consistency, the `download`-method is updated similarily to improve the code (this functionality is *very* old, even pre-dating the introduction of Promises in the code-base).
As mentioned in the JSDoc comment, this should not be used unless you know what you're doing, since it will lead to increased memory usage. However, in some situations (e.g. SVG-rendering), we still want to be able to run general clean-up on both the main/worker-thread while keeping loaded fonts attached to the DOM.[1]
As part of these changes, `WorkerTransport.startCleanup` is converted to an async method and we'll also skip clean-up when destruction has started (since it's redundant).
---
[1] The SVG-rendering mode is obviously not officially supported, since it's both rather incomplete and inherently slower. However with recent changes, whereby we cache repeated images on the document rather than the page level, memory usage can be *a lot* worse than before if we never attempt to release e.g. cached image-data when the viewer is in SVG-rendering mode.
These two properties were *never* intended to be anything but "private", hence it really cannot hurt to actually indicate that they're *not* part of any official API.
Currently the `fontName`-property contains an actual /Name-instance, which is a problem given that its fallback value is an empty string; see ca7f546828/src/core/default_appearance.js (L35)
The reason that this is a problem can be seen in ca7f546828/src/core/primitives.js (L30-L34), since an empty string short-circuits the cache. Essentially, in PDF documents, a /Name-instance cannot be empty and the way that the `DefaultAppearanceEvaluator` does things is unfortunately not entirely correct.
Hence the `fontName`-property is changed to instead contain a string, rather than a /Name-instance, which simplifies the code overall.
*Please note:* I'm tagging this patch with "[api-minor]", since PR 12831 is included in the current pre-release (although we're not using the `fontName`-property in the display-layer).
Fixes#13107
In the issue, some TrueType glyph names have the format `uniXXXX`.
Font's `Encoding` dictionary has the entry `Differences` but no
`BaseEncoding`. `uniXXXX` names are converted to glyph indices
using font's `post` table but currently that is done only when
`BaseEncoding` exists. We must enable the conversion also when only
`Differences` exists.
Currently only URL-strings are officially supported by `getDocument`, however at this point in time I cannot really see any compelling reason to not support `URL`-objects as well.
Most likely the reason that we've don't already support `URL`-objects, in `getDocument`, is that historically `URL` wasn't fully implemented across browsers and our old polyfill wasn't perfect; see https://developer.mozilla.org/en-US/docs/Web/API/URL/URL#browser_compatibility
*Please note:* Because of how the `url` parameter is currently handled, there's actually *some* cases where passing a `URL`-object to `getDocument` already works. That, in my opinion, provides additional motivation for supporting `URL`-objects officially, since it makes the API more consistent.
The following is an attempt to summarize the *current* situation, based on the actual code rather than the JSDocs:
- `getDocument("url string")` works and is documented.[1]
- `getDocument({ url: "url string", })` works and is documented.[1]
- `getDocument(new URL(...))` throws immediately, since no supported parameters are found.
- `getDocument({ url: new URL(...), })` actually works even though it's not documented.[1] Originally, when data was fetched on the worker-thread, this would likely have thrown since `URL` isn't clonable.[2]
- `getDocument({ url: { abc: 123, }, })`, or some similarily meaningless input, will be "accepted" by `getDocument` and then throw a `MissingPDFException` when attempting to fetch the bogus data.
With the changes in this patch, not only is `URL`-objects now officially supported and documented when calling `getDocument`, but we'll also do a much better job at actually validating any URL-data passed to `getDocument` (and instead fail early).
---
[1] In *browsers*, we create a valid URL thus indirectly validating the input. In Node.js environments, on the other hand, no validation is done since obtaining a baseUrl is more difficult (and PDF.js is primarily written for browsers anyway).
[2] https://developer.mozilla.org/en-US/docs/Web/API/Web_Workers_API/Structured_clone_algorithm#supported_types
Given the number of parameters that we now need to parse here, this code is no longer as readable as one would like. Hence this re-factoring, which will improve overall readability and also help with the next patch.
The reasons for making this change are:
- This property is not, nor has it ever been, used anywhere in the PDF.js display-layer.
- Related to the previous point, the format of the `defaultAppearance`-string is such that it'd be difficult to use it as-is in the display-layer anyway.
- It (usually) contains the "raw" appearance-string, from the PDF document, which is neither parsed nor validated and could thus be bogus.
- We now expose a `defaultAppearanceData`-property, which is first of all used in the display-layer and secondly contains actually parsed/validated data.
- In the event that a third-party implementation needs the `defaultAppearance`-string, it could be easily constructed from the recently added `defaultAppearanceData`-property.
All-in-all, I'm thus suggesting that we stop exposing an unused and unnecessary property on all Annotation-instances.
* JS - Handle correctly hierarchy of fields
- it aims to fix#13132;
- annotations can inherit their actions from the parent field;
- there are some fields which act as a container for other fields:
- they can be access through js so need to add them with an empty type (nothing in the spec about that but checked in Acrobat);
- calculation order list (CO) can reference them so need make them through this.getField;
- getArray method must return kids.
- field values are number, string, ... depending of their type but nothing in the spec on how to know what's the type:
- according to the comment for Canonical Format: https://www.adobe.com/content/dam/acom/en/devnet/pdf/pdfs/PDF32000_2008.pdf#page=461
- it seems that this "type" can be guessed from js action Format (when setting a type in Acrobat DC, the only affected thing is this action).
- util.scand with an empty string returns the current date.
- Actually support *linked* test-cases in the integration-tests (in the same way as the unit-tests).
- Add a new `"type": "other"`-kind to the test-manifest, to support *linked* test-cases in the unit/integration-tests without requiring the PDF document in question to also be a reference-test.
Based on this compatibility information, given that IE 11 is now *explicitly* unsupported, we should no longer need to bundle a `URL` polyfill in any builds: https://developer.mozilla.org/en-US/docs/Web/API/URL/URL#browser_compatibility
Note that the caveat listed for older Safari-versions doesn't apply to any code in the PDF.js library, since we never call `new URL(url, undefined)` in the code-base.
Note also that Node.js has a web-compatible `URL` implementation, which according to the "History" section at https://nodejs.org/api/url.html#url_the_whatwg_url_api has been available since Node.js `10.0.0` (according to https://nodejs.org/en/about/releases/ that branch is one month away from being EOL-ed).
Given that *all* data has been loaded on the main-thread, and then transferred to the worker-thread, ever since PR 8617 (almost four years ago) it should no longer be necessary to keep this special-case around.
Given that the `webViewerOpenFileViaURL` helper function is being defined in *all* builds anyway, the current pre-processor usage doesn't really improve readability in my opinion.
Currently `destRef === null`, which will only happen in documents with corrupt destinations, will (unsurprisingly) throw when trying to lookup the pageNumber. To avoid this, we can simply use the same format as in 1a2cdaffc5/web/pdf_link_service.js (L128)
The rotation handling that's currently living in `PDFViewerApplication` is *very* old, and pre-dates the introduction of the viewer components by years.
As can be seen in the `BaseViewer.pagesRotation` setter, we're not actually normalizing the rotation as intended and instead rely on the caller to handle that correctly. This is first of all inconsistent, given how other setters are implemented, and secondly it could also lead to the rotation being set to a value outside of the `[0, 360)`-range.
Finally, for improved consistency the rotation handling in `PageViewport` is updated similarly. Please note that this case, it's *not* changing the pre-existing logic.
This improves and simplifies #13102 in order to make printing of test-cases
like the one in bug 1698414 (where the real page is bigger than the target
page) much better, see incoming screenshots.
The reason why we need to stop setting .style.width / .style.height is to get
the right auto-sizing behavior in both axes. This shouldn't change behavior as
long as the print resolution is >= the CSS resolution, which seems like a
reasonable assumption.
If you try to print with a lower resolution than CSS, then instead of an
stretched canvas, you'd get a centered CSS-quality canvas, which seems
sensible. This could maybe be fixed with some CSS hackery (some combination of
min / max and viewport units perhaps?), but I think it's more trouble than it's
worth.
- implement few positioning properties: position, width, height, anchor;
- implement font element;
- implement fill element (used by font) and its children (linear, radial, ...);
- font property is inherited from ancestor container (see https://www.pdfa.org/wp-content/uploads/2020/07/XFA-3_3.pdf#page=43) so let CSS handles that stuff;
- in order to reduce the number of properties to set, only set non default properties and put the default in CSS;
- set a background to some containers to be able to see them (will be removed in a future commit).
The intention, in PR 12493, was that the page we're adding to the browser history should behave as if it were a "regular" internal destination (to properly convey user intent).
Unfortunately, since I didn't consider all the edge-cases correctly, it ended up behaving like a URL-hash instead which obviously wasn't intended. Note that currently this isn't a problem, however it can become an issue (in some cases) with upcoming re-factoring around `PDFHistory` and OpenAction support[1].
---
[1] I've started working on fixing the following TODO, which will require a couple of smaller tweaks here and there: 9d0ce6e79f/web/app.js (L1680-L1681)
In the `getAll`-method, we can have just one *explicit* loop rather than two indirect ones via the old `Object.assign`-call.
Also, changes the `get`-method to be slightly more compact (while keeping the logic intact).
Looking at this now, I cannot understand why we'd need to initialize `this.prefs` with all of the values from `this.defaults`.
Not only does this *indirectly* require one extra loop, via the `Object.assign`-call, but it also means that in GENERIC-builds changes to default-preference values might not be picked-up unless the the existing user-prefs are cleared (if the user had *manually* set prefs previously).
I'm adding these temporary l10n strings, so that I don't have to stare at a bunch of warning messages in the web-console when using the development viewer.
Given that the `enableXfa` parameter must to be passed to the API/Worker, and thus included in the `getDocument` call, it's not necessary to include it when initializing the `PDFViewer`-instance used in the default viewer. (Also, in `AppOptions`, the parameter is clearly marked with `OptionKind.API`.)
Furthermore, we probably don't want to display the fallback bar (in Firefox) for XFA documents when `enableXfa = true` is set.
While it's still not entirely clear if this would've prevented the issue as reported, given that the particular use-case reported apparently no longer applies, this small change really cannot hurt in general *and* it won't effect "regular" viewer builds in any way.
Given how the compatibility-values are being handled, it's not actually possible to override a *truthy* default-value with a *falsy* compatibility-value.
This is a simple oversight on my part, and with modern ECMAScript features this is very easy to support.
With the changes made in the previous patch, we can now list "disableTelemetry" in the `AppOptions` only for the `CHROME`-builds and thus remove the special-casing in the `checkChromePreferencesFile` helper function.
Currently there's a lot of duplication in the `buildLib` and `buildDefaultPreferences` functions, which seem quite unfortunate. Hence this patch extracts the common functionality in a new `buildLibHelper` function instead.
Originally the default preferences where simply placed in a JSON-file, checked into the repository, which over time became impractical, annoying, and error-prone to maintain; please see PR 10548.
While that improved the overall situation a fair bit, it however inherited one quite unfortunate property of the old JSON-based solution[1]: It's still not possible for *different* build targets to specify their *own* default preference values.
With some preferences, such as e.g. `enableScripting`, it's not inconceivable that you'd want to (at least) support build-specific default preference values. Currently that's not really possible, which is why this PR re-factors the default preferences generation to support this.
---
[1] This fact isn't really clear from the `AppOptions` implementation, unless you're familiar with the `gulpfile.js` code, which could lead to some confusion for those new to this part of the code-base.
A number of the currently supported *scripting* events only make sense in the "normal" viewer mode, and not when PresentationMode is active. For example:
- Changing the zoom-level will outright break rendering in PresentationMode, since it relies on "page-fit" being used.
- Focusing a particular (AcroForm) element won't work, and could break keyboard navigation, since forms should not be editable in PresentationMode (see issue 12232).
While this will perhaps not be perfect for *every* PDF document with mixed page orientation, based on the large number of bugs/issues seen over the years I'm however pretty convinced that it'll be an overall improvement in a majority of cases.
In order to improve things further, we'd probably need Firefox to support e.g. `@page` such that the viewer can provide better information to the print engine.
Currently, with `enablePrintAutoRotate = true` set, we're forced to loop through all the pages *twice* when checking for any landscape pages.
This seems completely unnecessary now, and using only *one* loop should be marginally more efficient in general.
Currently landscape pages are rotated *clockwise*, which for most documents feel wrong since holding the printed pages at their *left* edge causes the landscape pages to be viewed "upside down".
In general, since most documents are LTR ones, it feels more appropriate to instead rotate landscape pages *counterclockwise* for printing.
Similar to the existing `annotationsPromise` and `_jsActionsPromise` properties, the new `_xfaPromise` should obviously also be reset, since otherwise you might end up holding onto a lot of data for pages that are no longer active.
(That caching wasn't present in the original version of PR 13069, which is why I didn't spot it until now.)
- add an option to enable XFA rendering if any;
- for now, let the canvas layer: it could be useful to implement XFAF forms (embedded pdf in xml stream for the background and xfa form for the foreground);
- ui elements in template DOM are pretty close to their html counterpart so we generate a fake html DOM from template one:
- it makes easier to translate template properties to html ones;
- it makes faster the creation of the html element in the main thread.
Given that https://bugzilla.mozilla.org/show_bug.cgi?id=1699219 has enabled scripting for all Firefox-channels, it seems reasonable to simply set `enableScripting = true` unconditionally in the viewer preferences/options.
For now, this patch leaves the standalone viewer-components alone (such as e.g. `BaseViewer`), and if those are used scripting will thus have to be manually enabled (see e.g. the "simpleviewer"/"singlepageviewer" examples).
While there is nothing *outright* wrong with the existing implementation, it can however lead to increased memory usage in one particular case (that I completely overlooked when implementing this):
For "data:"-URLs, which by definition contains the entire PDF document and can thus be arbitrarily large, we obviously want to avoid sending, storing, and/or logging the "raw" docBaseUrl in that case.
To address this, this patch makes the following changes:
- Ignore any non-string in the `docBaseUrl` option passed to `getDocument`, since those are unsupported anyway, already on the main-thread.
- Ignore "data:"-URLs in the `docBaseUrl` option passed to `getDocument`, to avoid having to send what could potentially be a *very* long string to the worker-thread.
- Parse the `docBaseUrl` option *directly* in the `BasePdfManager`-constructors, on the worker-thread, to avoid having to store the "raw" docBaseUrl in the first place.
It seems reasonable to place this alongside the *similar* `getFilenameFromUrl` helper function. This way, with the changes in the next patch, we also avoid having to expose the `isDataScheme` function in the API itself and we instead expose `getPdfFilenameFromUrl` in the API (which feels overall more appropriate).
The issue that this patch fixes is extremely unlikely, but still theoretically possible, and I really should've caught this earlier.
Note how `BaseViewer.pagesPromise` will only be defined when a document is active, see below, and that if a printing event (triggered from scripting) arrives while the document is been closed there's a small chance that the promise isn't defined.
eb92ed12f2/web/base_viewer.js (L426-L428)
This extends PR 13033 slightly, with a heuristic to support corrupt PDF documents where the `LineAnnotation`s have an empty /Rect-entry. Please note that while I have no idea if this is "correct", this patch at least makes us output the same /BBox as re-saving in Adobe Reader does.
This builds on top of #13100, but this changes printing behavior intentionally
so I thought it was worth discussing separately, to improve the rendering on
test-cases like the one in https://bugzil.la/1697778.
This matches what e.g. Evince does when you print the PDF in there on an A4
printer.
We use margins to center horizontally, and flex to center vertically. The
reasoning for this is that it should have better browser support (though maybe
pdf.js no longer supports browsers without flex support?) and it's just as
simple.
@supports() is not supposed to report support for page descriptors, this is
depending on a Chromium bug, which doesn't treat as invalid:
```
<div style="size: 1pt 1pt">
```
Even though it should. That is
https://bugs.chromium.org/p/chromium/issues/detail?id=1079214
There's no need to use @supports for this. If the descriptor is not accepted it
will just be ignored.
That way, when Firefox implements @page { size }, which is in progress, it will
get the right behavior.
First, there's just no need to do something like this, this is simpler and
closer to what the screen renderer does.
Second, this causes overflow, which Firefox tries to compensate for when
fitting to page width, and fails at it. That is tracked in:
https://bugzilla.mozilla.org/show_bug.cgi?id=1698136
But this bug works around it by not causing overflow.
For modern browsers, we could avoid the duplication setting the style attribute
by using something like width: min/max-content, but this is not a big deal I
think, let me know if you'd prefer that.
Also I had to add a max-height for Chromium not to create extra pages. This
is harmless in Firefox and workarounds the Chromium bug, so so be it.
This is mostly done using `gulp lint --fix` with a few manual changes in
the following diff:
```diff
diff --git a/src/core/pattern.js b/src/core/pattern.js
index 365491ed3..eedd8b686 100644
--- a/src/core/pattern.js
+++ b/src/core/pattern.js
@@ -105,7 +105,7 @@ const Pattern = (function PatternClosure() {
return Pattern;
})();
-var Shadings = {};
+const Shadings = {};
// A small number to offset the first/last color stops so we can insert ones to
// support extend. Number.MIN_VALUE is too small and breaks the extend.
@@ -597,16 +597,15 @@ Shadings.Mesh = (function MeshClosure() {
if (!(0 <= f && f <= 3)) {
throw new FormatError("Unknown type6 flag");
}
- var i, ii;
const pi = coords.length;
- for (i = 0, ii = f !== 0 ? 8 : 12; i < ii; i++) {
+ for (let i = 0, ii = f !== 0 ? 8 : 12; i < ii; i++) {
coords.push(reader.readCoordinate());
}
const ci = colors.length;
- for (i = 0, ii = f !== 0 ? 2 : 4; i < ii; i++) {
+ for (let i = 0, ii = f !== 0 ? 2 : 4; i < ii; i++) {
colors.push(reader.readComponents());
}
- var tmp1, tmp2, tmp3, tmp4;
+ let tmp1, tmp2, tmp3, tmp4;
switch (f) {
// prettier-ignore
case 0:
@@ -729,16 +728,15 @@ Shadings.Mesh = (function MeshClosure() {
if (!(0 <= f && f <= 3)) {
throw new FormatError("Unknown type7 flag");
}
- var i, ii;
const pi = coords.length;
- for (i = 0, ii = f !== 0 ? 12 : 16; i < ii; i++) {
+ for (let i = 0, ii = f !== 0 ? 12 : 16; i < ii; i++) {
coords.push(reader.readCoordinate());
}
const ci = colors.length;
- for (i = 0, ii = f !== 0 ? 2 : 4; i < ii; i++) {
+ for (let i = 0, ii = f !== 0 ? 2 : 4; i < ii; i++) {
colors.push(reader.readComponents());
}
- var tmp1, tmp2, tmp3, tmp4;
+ let tmp1, tmp2, tmp3, tmp4;
switch (f) {
// prettier-ignore
case 0:
@@ -897,7 +895,7 @@ Shadings.Mesh = (function MeshClosure() {
decodeType4Shading(this, reader);
break;
case ShadingType.LATTICE_FORM_MESH:
- var verticesPerRow = dict.get("VerticesPerRow") | 0;
+ const verticesPerRow = dict.get("VerticesPerRow") | 0;
if (verticesPerRow < 2) {
throw new FormatError("Invalid VerticesPerRow");
}
```
As part of testing this, I've diffed the output of `gulp mozcentral` with/without this patch and the *only* difference is the incremented `version`/`build` numbers.
A significant portion of the code-base has now been converted to use `let`/`const`, rather than `var`, hence it should be possible to simply enable the ESLint `no-var` rule globally.
This way we can ensure that new code won't accidentally use `var`, and it also removes the need to manually enable the rule in various folders.
Obviously it makes sense to continue the efforts to replace `var`, but that should probably happen on a file and/or folder basis.
Please note that this patch excludes the following code:
- The `extensions/` folder, since that seemed easiest for now (and I don't know exactly what the support situation is for the Chromium-extension).
- The entire `external/` folder is ignored, since most of it's currently excluded from linting.
For the code that isn't imported from elsewhere (and should be ignored), we should probably (at some point) bring the code up to the same linting/formatting standard as the rest of the code-base.
- Various files in the `test/` folder are ignored, as necessary, since the way that a lot of this code is loaded will require some care (or perhaps larger re-factoring) when removing `var` usage.
While the JSDocs have never advertised `getDocument` as supporting Node.js `Buffer`s, that apparently doesn't stop users from passing such data structures to `getDocument`.
In theory the existing `instanceof Uint8Array` check ought to have caught Node.js `Buffer`s, however for reasons that I don't even pretend to understand that check actually passes. Hence this patch which, *only* in Node.js environments, will special-case `Buffer`s to hopefully provide a slightly better out-of-the-box behaviour in Node.js environments[1].
---
[1] Although I'm not sure that we necessarily want to advertise this in the JSDocs, given the specialized use-case.
Replace the `objectFromEntries` helper function with an `objectFromMap` one instead, and simplify the data lookup in the AnnotationStorage.getValue method
(This is something that I happened to notice while working on the previous patch.)
The code in the `examples/mobile-viewer/viewer.js` file is essentially copied from an older version of the default viewer, hence we can slightly simplify the `animationStarted` handling here.
Currently errors occurring within the `src/display/{text_layer, annotation_layer}.js` files are not being handled properly by the test-suite, and the tests simply time out rather than failing as intended.
This makes it *very* easy to accidentally overlook a certain type of errors, see e.g. https://github.com/mozilla/pdf.js/pull/13055#discussion_r589005041, which this patch will thus prevent.
Note that the majority of these changes were done automatically, by using `gulp lint --fix`, and the manual changes were limited to the following diff:
```diff
diff --git a/src/core/cff_parser.js b/src/core/cff_parser.js
index d684c200e..2e2b811e4 100644
--- a/src/core/cff_parser.js
+++ b/src/core/cff_parser.js
@@ -555,7 +555,7 @@ const CFFParser = (function CFFParserClosure() {
stackSize %= 2;
validationCommand = CharstringValidationData[value];
} else if (value === 10 || value === 29) {
- var subrsIndex;
+ let subrsIndex;
if (value === 10) {
subrsIndex = localSubrIndex;
} else {
@@ -886,15 +886,15 @@ const CFFParser = (function CFFParserClosure() {
format = bytes[pos++];
switch (format & 0x7f) {
case 0:
- var glyphsCount = bytes[pos++];
+ const glyphsCount = bytes[pos++];
for (i = 1; i <= glyphsCount; i++) {
encoding[bytes[pos++]] = i;
}
break;
case 1:
- var rangesCount = bytes[pos++];
- var gid = 1;
+ const rangesCount = bytes[pos++];
+ let gid = 1;
for (i = 0; i < rangesCount; i++) {
const start = bytes[pos++];
const left = bytes[pos++];
@@ -938,7 +938,7 @@ const CFFParser = (function CFFParserClosure() {
}
break;
case 3:
- var rangesCount = (bytes[pos++] << 8) | bytes[pos++];
+ const rangesCount = (bytes[pos++] << 8) | bytes[pos++];
for (i = 0; i < rangesCount; ++i) {
let first = (bytes[pos++] << 8) | bytes[pos++];
if (i === 0 && first !== 0) {
@@ -1173,7 +1173,7 @@ class CFFDict {
}
}
-var CFFTopDict = (function CFFTopDictClosure() {
+const CFFTopDict = (function CFFTopDictClosure() {
const layout = [
[[12, 30], "ROS", ["sid", "sid", "num"], null],
[[12, 20], "SyntheticBase", "num", null],
@@ -1229,7 +1229,7 @@ var CFFTopDict = (function CFFTopDictClosure() {
return CFFTopDict;
})();
-var CFFPrivateDict = (function CFFPrivateDictClosure() {
+const CFFPrivateDict = (function CFFPrivateDictClosure() {
const layout = [
[6, "BlueValues", "delta", null],
[7, "OtherBlues", "delta", null],
@@ -1265,11 +1265,12 @@ var CFFPrivateDict = (function CFFPrivateDictClosure() {
return CFFPrivateDict;
})();
-var CFFCharsetPredefinedTypes = {
+const CFFCharsetPredefinedTypes = {
ISO_ADOBE: 0,
EXPERT: 1,
EXPERT_SUBSET: 2,
};
+
class CFFCharset {
constructor(predefined, format, charset, raw) {
this.predefined = predefined;
@@ -1695,7 +1696,7 @@ class CFFCompiler {
// For offsets we just insert a 32bit integer so we don't have to
// deal with figuring out the length of the offset when it gets
// replaced later on by the compiler.
- var name = dict.keyToNameMap[key];
+ const name = dict.keyToNameMap[key];
// Some offsets have the offset and the length, so just record the
// position of the first one.
if (!offsetTracker.isTracking(name)) {
```
Rather than first checking if data exists before fetching it from storage, we can simply do the lookup directly and then check its value.
Note that this follows the same pattern as utilized in the `AnnotationStorage.setValue` method.
Given that it's only used with `Map`s, and that it's currently implemented in such a way that we (indirectly) must iterate through the data *twice*, some simplification cannot hurt here.
Note that the only reason that we're not using `Object.fromEntries(...)` directly, at each call-site, is that that one won't guarantee that a `null` prototype is being used.
*Please note:* Given the pre-existing issues raised in PR 13056, which seem to block immediate progress there, this patch extracts some *overall* improvements of the scripting/sandbox destruction in `PDFScriptingManager`.
As can be seen in `BaseViewer.setDocument`, it's currently necessary to *manually* delay the `PDFScriptingManager`-destruction in order for things to work correctly. This is, in hindsight, obviously an *extremely poor* design choice on my part; sorry about the churn here!
In order to improve things overall, the `PDFScriptingManager._destroyScripting`-method is re-factored to wait for the relevant events to be dispatched *before* sandbox-destruction occurs.
To avoid the scripting/sandbox-destruction hanging indefinitely, we utilize a timeout to force-destroy the sandbox after a short time (currently set to 1 second).
These tests, and their [accompanying Wiki page](https://github.com/mozilla/pdf.js/wiki/Required-Browser-Features), haven't received any real updates for *many years* and are sufficiently out of date to be effectively useless now.
Providing *irrelevant* compatibility information seems overall worse than not providing any information, and as suggested in the issue it'd probably be better to use https://github.com/mozilla/pdf.js#online-demo for checking if a particular platform/browser is supported.
Thanks to version control, it's easy to restore these files should the need ever arise. However, re-introducing these tests would essentially require updating every single test-case *and* a commitment to keeping them up to date with future code changes.
By moving this code from the `BaseViewer` and into `PDFScriptingManager`, all of the scripting initialization/handling code is now limited to just one file/class which help overall readability (in my opinion). Also, this patch is a *net reduction* in number of lines of code which can never hurt.
As part of these changes, the intermediary "pageopen"/"pageclose" events are now removed in favor of using the "regular" viewer events directly in `PDFScriptingManager`. Hence this removes some (strictly unnecessary) indirection in the current code, when handling PageOpen/PageClose events, which leads to overall fewer function calls in this part of the code.
The *main* purpose of this patch is to allow scripting to be used together with the viewer components, note the updated "simpleviewer"/"singlepageviewer" examples, in addition to the full default viewer.
Given how the scripting functionality is currently implemented in the default viewer, trying to re-use this with the standalone viewer components would be *very* hard and ideally you'd want it to work out-of-the-box.
For an initial implementation, in the default viewer, of the scripting functionality it probably made sense to simply dump all of the code in the `app.js` file, however that cannot be used with the viewer components.
To address this, the functionality is moved into a new `PDFScriptingManager` class which can thus be handled in the same way as all other viewer components (and e.g. be passed to the `BaseViewer`-implementations).
Obviously the scripting functionality needs quite a lot of data, during its initialization, and for the default viewer we want to maintain the current way of doing the lookups since that helps avoid a number of redundant API-calls.
To that end, the `PDFScriptingManager` implementation accepts (optional) factories/functions such that we can maintain the current behaviour for the default viewer. For the viewer components specifically, fallback code-paths are provided to ensure that scripting will "just work"[1].
Besides moving the viewer handling of the scripting code to its own file/class, this patch also takes the opportunity to re-factor the functionality into a number of helper methods to improve overall readability[2].
Note that it's definitely possible that the `PDFScriptingManager` class could be improved even further (e.g. for general re-use), since it's still heavily tailored to the default viewer use-case, however I believe that this patch is still a good step forward overall.
---
[1] Obviously *all* the relevant document properties might not be available in the viewer components use-case (e.g. the various URLs), but most things should work just fine.
[2] The old `PDFViewerApplication._initializeJavaScript` method, where everything was simply inlined, have over time (in my opinion) become quite large and somewhat difficult to *easily* reason about.
These changes will be necessary for the next patch, since we don't want to accidentally pull in the entire default viewer in the standalone viewer components.
At this point in time the standalone `acroforms` example is no longer needed, since both the `simpleviewer` and `singlepageviewer` examples already support AcroForms by default.
Moreover, the `acroforms` example no longer illustrate best practices, given the direct usage of `PDFPageView` for a multipage document.
Given that scripting is now enabled in Firefox Nightly (but only there), it seems weird to not have scripting enabled by default in `gulp server` mode.
Rather than having to spell out the English fallback strings at *every* single `IL10n.get` call-site throughout the viewer, we can simplify things by collecting them in *one* central spot.
This provides a much better overview of the fallback l10n strings used, which makes future changes easier and ensures that fallback strings occuring in multiple places cannot accidentally get out of sync.
Furthermore, by making the `fallback` parameter of the `IL10n.get` method *optional*[1] many of the call-sites (and their surrounding code) become a lot less verbose.
---
[1] It's obviously still possible to pass in a fallback string, it's just not required.
Now that we have scripting support, warning about e.g. JavaScript actions doesn't seem necessary anymore. Especially considering that scripting-related actions are/will not be parsed by the `Catalog.parseDestDictionary` method anyway, since it's intended for handling "simple" actions.
All of this code predates the existence of native JS classes, however we can now clean this up a bit. This patch thus let us remove some variable "shadowing" from the code.
This helper function is first of all only called *twice*, and secondly it also leads to unnecessary intermediate allocations given how the `TypedArray`s are handled.
Hence we can simply inline this small function, and thus directly allocate the combined `TypedArray` instead.
The `compareByteArrays` is first of all duplicated in multiple closures in the `src/core/crypto.js` file. Secondly, despite its name, it's also functionally equivalent to the now existing `isArrayEqual` helper function.
The `isArrayEqual` helper function is changed to use a standard `for`-loop, rather than `Array.prototype.every`, since that ought to be slightly more efficient given that we're now using it with (potentially) larger data.
All of this code predates the existence of native JS classes, however we can now clean this up a bit. This patch thus let us remove some variable "shadowing" from the code.
Note that this particular helper function is, with the exception of the `GENERIC` default viewer and the (unsupported) SVG-backend, mostly unused at this point in time. Hence we should be able to clean-up this helper function slightly.
Also, fixes a small inconsistency in the `SVGGraphics` initialization in the viewer, by passing in the `disableCreateObjectURL` compatibility-option. Given that the SVG-backend isn't officially supported/recommended this shouldn't have been an issue, but given that I spotted this it can't hurt to fix it.
For any viewer component not listed in `web/pdf_viewer.component.js`, it shouldn't be necessary to provide a default value for the `l10n`-parameters.
Note also that these *specific* components are heavily tailored towards the default viewer use-case, rather than for general usage.
Given that `PDFFindBar` is written *specifically* for the default viewer, rather than general usage (as opposed to the `PDFFindController`), we should be able to simply assume that the `findResultsCount` DOM-element is always present. Even more so, when we're purposely not doing any similar checks for other DOM-elements in this code.
Also, remove unnecessary `null` defaults for the various DOM-element options in the constructor, since the code simply assumes that all of the relevant DOM-elements are in fact available.
Note how the `PDFAttachmentViewer` handles PDF file attachments specially, by opening them in a new window/tab, rather than forcing them to be downloaded. This is done to improve the overall UX, since browsers in general are able to handle PDF files internally.
However, for file *annotations* we're currently not attempting to do the same thing and are instead just downloading them directly. In order to unify the behaviour, without having to duplicate a lot of code, the opening of PDF file attachments is thus moved into a new `DownloadManager.openOrDownloadData` method.
- strokeColor corresponds to borderColor;
- support fillColor and textColor;
- support colors on the different annotations;
- fix typo in aforms (+test).
This is similar to the other methods, and the only reason for this not having been done originally is that the `cancel` functionality is a later addition.
Rather than converting the `AnnotationStorage`-data to an Object, before sending it to the worker-thread, we should be able to simply send the internal `Map` directly.
The "structured clone algorithm" doesn't have a problem with `Map`s, however the `LoopbackPort` used when workers are *disabled* (e.g. in Node.js environments) didn't use to support them. With PR 12997 having lifted that restriction, we should now be able to simply send the `AnnotationStorage`-data as-is rather than having to iterate through it to first create an Object.
*Please note:* The changes in `src/core/annotation.js` could have been a lot more compact if we were able to use optional chaining in the `src/core` folder. Unfortunately that's still not possible, since SystemJS is being used in the development viewer (i.g. `gulp server`) and fixing that is *still* blocked by [bug 1247687](https://bugzilla.mozilla.org/show_bug.cgi?id=1247687).
*Please note:* The `defer` parameter has been enabled by default ever since PR 9777 (in 2018), which first shipped in PDF.js release `2.0.943`.
With workers *disabled*, e.g. in Node.js environments, this has been used ever since without any problems reported[1].
The impetus for this change was that I happened to notice that *if* the `LoopbackPort` was used with synchronous event dispatching, we'd simply send that data as-is to the listeners. This created an inconsistency in the data returned from the `pdf.worker.js` file, since `postMessage` used with *actual* workers (or the `LoopbackPort` with `defer = true`) will ignore/throw when encountering unclonable data.
Originally my intention was simply to just call `cloneValue` regardless of the event dispatching used in `LoopbackPort`, however looking at the use-cases (or lack thereof) of the `LoopbackPort` it seemed reasonable to simply remove the `defer` parameter instead.
This patch is tagged "[api-minor]" since the `LoopbackPort` is still exposed in the API, although I really hope that no third-party is using this (since disabling workers leads to bad performance).
Finally, this patch changes a `forEach` loop to `for...of` and makes uses of optional changing in existing code.
---
[1] As evident by the `npm test` command run by Github Actions, and previously by Travis.
With the previous patch this function is now *only* accessed on the worker-thread, hence it's no longer necessary to include it in the *built* `pdf.js` file.
With the previous patch this functionality is now *only* accessed on the worker-thread, hence it's no longer necessary to include it in the *built* `pdf.js` file.
The only reason, as far as I can tell, for parsing the Metadata on the main-thread is how it was originally implemented. When Metadata support was first implemented, it utilized the [`DOMParser`](https://developer.mozilla.org/en-US/docs/Web/API/DOMParser) which isn't available in workers.
Today, with the custom XML-parser being used, that's no longer an issue and it seems reasonable to move the Metadata parsing to the worker-thread[1], since that's where all parsing should happen (for performance reasons).
Based on these changes, we'll be able to reduce the now unnecessary duplication of the XML-parser (and related code) in both of the *built* `pdf.js`/`pdf.worker.js` files.
Finally, this patch changes the `_repair` method to use "Array + join" rather than string concatenation.
---
[1] This needed the previous patch, to enable sending of `Map`s between threads with workers disabled.
I happened to look at this code, and I can't for the life of me figure out why I didn't just implement it like this patch in the first place (since the current format feels overly verbose).
*This is somewhat similar to PR 12931.*
For PDF documents where fonts are completely missing in the /Resources dictionaries, there's basically no "correct" way of rendering the document.
Hence it's very unlikely that another PDF viewer will do a better job than PDF.js in these cases, and consequently it seems highly questionable if the fallback bar really helps here.
Given that these event listeners should essentially never be needed, but are included simply to avoid breakage in edge-cases, it can't hurt to make this code slightly less verbose.
- Mark `BaseViewer.initializeScriptingEvents` as an `async` method, since that's actually how it's being used in the default viewer (see `PDFViewerApplication-_initializeJavaScript`).
- Change `BaseViewer._pageWidthScaleFactor` to access the *internal* scroll/spread-modes directly, rather than using the getters, since that's consistent with the rest of the code (and not just for these properties).
For reasons that I now can't for the life of me understand, I included handling of the `PresentationModeState.CHANGING`-case despite it not actually doing anything.
Given that these HTML elements are not being used at all in `MOZCENTRAL`-builds, note the preprocessor check in `PDFViewerApplication._otherError`, we obviously don't need the HTML code either.
The following checks are all unneeded, and could easily cause confusion when reading the code. (All of them are my fault as well, since I've sometimes added those checks without really thinking about the surrounding code.)
- In `PartialEvaluator.hasBlendModes` there cannot be any `MissingDataException`s thrown, given that the `Page.getOperatorList` method waits for all the necessary /Resources to load first. Furthermore, note also that if an error is thrown from `PartialEvaluator.hasBlendModes` then it'd completely break rendering of that page, since any errors thrown from `Page.getOperatorList` are simply sent to the main-thread.
- In `PartialEvaluator.handleColorN` there cannot be any `MissingDataException`s thrown, given that again the `Page.getOperatorList` method waits for all the necessary /Resources to load before operatorList parsing starts.
- In `XRef.readXRef` there cannot be any `MissingDataException`s thrown, given that we're *explicitly* requesting (and waiting for) the entire document in `pdfManagerReady` (in `src/core/worker.js`) before re-parsing of a corrupt document starts.
* don't set a value in annotationStorage by default:
- having an undefined when the annotation is rendered for saving/printing means nothing has changed so use normal appearance
- aims to fix https://bugzilla.mozilla.org/show_bug.cgi?id=1681687
* change the way to compute font size when this one is null in DA:
- make fontSize proportional to line height
- in multiline case, take into account the number of lines for text entered to adapt the font size
- use ascent of the fallback font instead of the one from pdf to position spans
- use TextMetrics.fontBoundingBoxAscent if available or
- use a basic heuristic to guess ascent in drawing char on a canvas
- compute ascent as a ratio of font height
Some of the localization strings (e.g. "loading_error") are repeated multiple times throughout the `web/app.js` file, which means that we need to duplicate the fallback strings as well. Furthermore, the signature of the `IL10n.get` method makes the call-sites quite verbose.
By adding a new helper method, in `PDFViewerApplication`, we're able to gather the localization fallback strings in one central spot in `web/app.js` and also make the lookup of the error/warning messages more compact.
This code is *very* old and it even predates the existence of arrow functions. Hence we can now reduce the overall verbosity by not having to explicitly spell out `PDFViewerApplication` everywhere.
This feature was Firefox-specific, and it's now been removed from the HTML specification and it's disabled by default starting with Firefox 85. Hence it seems completely unnecessary to keep this code in the default viewer.
Please refer to https://groups.google.com/g/mozilla.dev.platform/c/tc11BCenm2c and the resources that it links to.
Given that `getComputedStyle` only works on visible elements, the result of PR 12354 is that if the viewer is placed in a *hidden* `iframe` the viewer will now be broken. This obviously wasn't the intention of that PR, hence I believe that we should limit the `position: absolute;` check slightly to avoid this.
The *third* page of the referenced PDF document currently fails to render completely, since one of its font files fail to load.
Since that error isn't handled, a large part of the text is thus missing which looks quite bad. By "replacing" the font data with an *empty* stream, we'll thus be able to fallback to rendering the text with a standard font (instead of using `ErrorFont`). While there's obviously no guarantee that things will look perfect, actually rendering the text at all should be an improvement in general.
Also, print a warning in `PartialEvaluator.loadFont` when the `PartialEvaluator.translateFont` method rejects, since that'd have helped debug/fix the issue faster.
*As far as I can tell, this has been broken ever since PR 3289 (back in 2013) without anyone noticing.*
For any non-`MissingDataException` errors encountered in `ObjectLoader._walk`, we're simply throwing immediately which thus has the potential to *completely* break rendering of an entire page.
In practice this is obviously only an issue for PDF documents which are in one way or another corrupt, since that's the only way that `XRef.fetch` will throw non-`MissingDataException` errors. To make matters worse these errors are *intermittent*, since they can only occur if the document is still loading when the `ObjectLoader`-code runs (note the early return in `ObjectLoader.load`).
Please note that we cannot simply catch the error and let "normal" parsing continue in `ObjectLoader._walk`, since that could lead to errors elsewhere given that resources "below" the current one (in the graph) might not be checked as intended then.
All-in-all, the only way to make absolutely sure that we won't cause *unexpected* `MissingDataException`s somewhere else in the code-base is to fallback to fetching the *entire* document in this edge-case.
aria-controls state
In testing, screen readers such as JAWS have trouble understanding the expanded state of the buttons that expand hidden menus due to lacking aria-expanded attribute. Also, given that the buttons do not contain the controlled/shown element, they should also define the aria-controls attribute with associated element id per https://www.w3.org/TR/wai-aria-1.1/#aria-expanded
This fixes adds these requirements for the sidebar, find, and secondary toolbar buttons.
Currently it's not *immediately* clear from the code itself, unless you look at the definition of `this._pageLabels`, that the default value is `null`.[1]
We can improve this, and also reduce the amount of code, by using modern ECMAScript features such as optional chaining and nullish coalescing.
---
[1] Keep in mind that an *empty* string is actually a valid page label, according to the PDF specification.
- in order to evaluate SOM expressions nodes and their attributes must be checked in the same order as in the xml;
- add an object XFAObjectArray with a parameter max to handle multiple children with the same name.
Given that we don't focus the viewer *itself* (among other things) when the viewer is embedded, I suppose that it makes some sense to not focus the `PasswordPrompt` input-field either on load.
In order to improve the overall UX here, if an *incorrect* password was provided we'll still focus the input-field.
Fixes 12951 (assuming we care to do so, of course).
- the parser is base on a class extending XMLParserBase
- it handle xml namespaces:
* each namespace is assocated with a builder
* builder builds nodes belonging to the namespace
* when a node is inserted in the parent namespace compatibility is checked (if required)
- to avoid name collision between xml names and object properties, use Symbol.
Given that IE 11/Edge is now unsupported in PDF.js, and that Microsoft Edge is now a Chromium-browser, we can avoid (some) unnecessary bloat in the built CSS files.
Given that `FontFaceObject` is not exposed in the public API, but only accessed internally, there's no need to assume that a `FontFaceObject`-instance is ever initialized without `onUnsupportedFeature` being provided. This is also consistent with the `BaseFontLoader` implementation.
With PR 10539, we'll now always attempt to fallback to the PDF.js built-in font renderer for fonts that fail to load (i.e. are rejected by the sanitizer). Generally speaking, these errors are the result of insufficient validation in the PDF.js font code, however in almost all cases we've seen thus far our built-in font renderer manages just fine.
However, we still trigger the `onUnsupportedFeature` reporting, which in Firefox causes the fallback bar to be displayed. Given that, in a majority of cases[1], things look fine it seems unfortunate to bother the user with the fallback bar here.
Note that even though we no longer show the fallback bar in this case, we still report telemetry as before.
---
[1] The only *known* case where things aren't fine with the built-in font renderer is issue 10232, however that document is sufficiently broken that there's a couple of other things that will trigger the fallback bar.
Given that we'll only cache `/XObject`s of the `Image`-type globally, we can utilize that in `PartialEvaluator.getTextContent` as well. This way, in cases such as e.g. issue 12098, we can avoid having to fetch/parse `/XObject`s that we already know to be `Image`s. This is helpful, since `Stream`s are not cached on the `XRef` instance (given their potential size) and the lookup can thus be somewhat expensive in general.
Also, skip a redundant `RefSetCache.has` check in the `GlobalImageCache.getData` method.
The widths property should be an object to match what metrics returns.
In ZapfDingbats.pdf I was getting a data clone error with pdfBug enabled.
In buildCharCodeToWidth() there was an encoding with the name "at" which
is also the name of a method on an array. buildCharCodeToWidth assumes an
object is passed in, so when it checked for the "at" property, it found the
method and copied it over.
This only seemed to affect Firefox.
When implementing the `GlobalImageCache` functionality I was mostly worried about the effect of *very large* images, hence the maximum number of cached images were purposely kept quite low[1].
However, there's one fairly obvious problem with that approach: In documents with hundreds, or even thousands, of *small* images the `GlobalImageCache` as implemented becomes essentially pointless.
Hence this patch, where the `GlobalImageCache`-implementation is changed in the following ways:
- We're still guaranteed to be able to cache a *minimum* number of images, set to `10` (similar as before).
- If the *total* size of all the cached image data is below a threshold[2], we're allowed to cache additional images.
This patch thus *improve*, but doesn't completely fix, issue 12098. Note that that document is created by a *very poor* PDF generator, since every single page contains the *entire* document (with all of its /Resources) and to create the individual pages clipping is used.[3]
---
[1] Currently set to `10` images; imagine what would happen to overall memory usage if we encountered e.g. 50 images each 10 MB in size.
[2] This value was chosen, somewhat randomly, to be `40` megabytes; basically five times the [maximum individual image size per page](6249ef517d/src/display/api.js (L2483-L2484)).
[3] This surely has to be some kind of record w.r.t. how badly PDF generators can mess things up...
Given how our existing code-base is written, we simply can't (and probably don't want to) enable *all* of this ESLint rule. The reason is that we have a lot of code, especially related to functions, which is fine at runtime but might to a static analyzer *look* like a problem.
However, even in its most *basic* version this ESLint rule can help catch some outright bugs; see PR 12891.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/no-use-before-define
Note how, in the `if (this.stateManager.stateStack.length !== 0) {` branch, we're attempting to access the not yet defined variable[1] `args`. If this code-path is ever hit, an Error will be thrown and parsing will thus be aborted immediately (likely leading to e.g. rendering bugs).
Note that I found this purely by accident, since I happened to glance at the LGTM report. However, I've since found that the error is also present during the unit-test[2] and with this patch we're actually testing the *intended* thing here.
As part of fixing this, and to avoid re-introducing a similar bug in the future, we'll now instead always reset `args.length` *before* attempting to read the next operator.
Also, we can use the existing `EvaluatorPreprocessor.savedStatesDepth` getter to simplify the save/restore detection a tiny bit.
---
[1] The ESLint rule `no-use-before-define` would have helped catch this problem, but unfortunately we cannot enable that without quite a bit of refactoring all over the code-base.
[2] The unit-test was updated such that it would fail in the `master`-branch.
- For wrapped scrolling, we unfortunately need to do a fair bit of parsing of the *current* page layout. Compared to e.g. the spread-modes, where we can easily tell how the pages are laid out, with wrapped scrolling we cannot tell without actually checking. In particular documents with varying page sizes require some care, since we need to check all pages on the "row" of the current page are visible and that there aren't any "holes" present. Otherwise, in the general case, there's a risk that we'd skip over pages if we'd simply always advance to the previous/next "row" in wrapped scrolling.
- For horizontal scrolling, this patch simply maintains the current behaviour of advancing *one* page at a time. The reason for this is to prevent inconsistent behaviour for the next and previous cases, since those cannot be handled identically. For the next-case, it'd obviously be simple to advance to the first not completely visible page. However for the previous-case, we'd only be able to go back *one* page since it's not possible to (easily) determine the page layout of non-visible pages (documents with varying page sizes being a particular issue).
- For vertical scrolling, this patch maintains the current behaviour by default. When spread-modes are being used, we'll now attempt to advance to the next *spread*, rather than just the next page, whenever possible. To prevent skipping over a page, this two-page advance will only apply when both pages of the current spread are visible (to avoid breaking documents with varying page sizes) and when the second page in the current spread is fully visible *horizontally* (to handle larger zoom values).
In order to reduce the performance impact of these changes, note that the previous/next-functionality will only call `getVisibleElements` for the scroll/spread-modes where that's necessary and that "normal" vertical scrolling is thus unaffected by these changes.
To support these changes, the `getVisibleElements` helper function will now also include the `widthPercent` in addition to the existing `percent` property.
The `PDFViewer._updateHelper` method is changed slightly w.r.t. updating the `currentPageNumber` for the non-vertical/spread modes, i.e. won't affect "normal" vertical scrolling, since that helped simplify the overall calculation of the page advance.
Finally, these new `BaseViewer` methods also allow (some) simplification of previous/next-page functionality in various viewer components.
*Please note:* There's one thing that this patch does not attempt to change, namely disabling of the previous/next toolbarButtons respectively the firstPage/lastPage secondaryToolbarButtons. The reason for this is that doing so would add quite a bit of complexity in general, and if for some reason `BaseViewer._getPageAdvance` would get things wrong we could end up incorrectly disabling the buttons. Hence it seemed overall safer to *not* touch this, and accept that the buttons won't be `disabled` despite in some edge-cases no further scrolling being possible.
Given how recently logical assignment operators support were added to browsers, we're intentionally translating them even in `SKIP_BABEL = true` builds for the time being.
However, in either `MOZCENTRAL` or `TESTING` builds it should be OK to simply leave logical assignment operators intact.
With the changes in PR 12831, it's no longer necessary to keep track of the `fontName`-string separately since it's available through the `_defaultAppearanceData`-property as well.
As can be seen in 2cba290361/src/core/evaluator.js (L986) the `gStateObj` (which is actually an Array despite its name), is wrapped in Array when it's inserted into the OperatorList. Hence we obviously need to take this into account when accessing it in `TranslatedFont._removeType3ColorOperators`; this mistake happened because we don't have any test-cases for this particular code-path as far as I know.
By changing this a `shadow`ed getter, we can simply access it directly and not worry about it being initialized. I have no idea why I didn't just implement it this way in the first place.
* Add a parser to get font data from the default appearance
- pdfium & poppler use a special parser too to get these info.
* Update src/core/default_appearance.js
Co-authored-by: Jonas Jenwald <jonas.jenwald@gmail.com>
Co-authored-by: Jonas Jenwald <jonas.jenwald@gmail.com>
This adds "optional chaining" and "nullish coalescing" to the list of features needed when using the *modern*, i.e. non-translated/non-polyfilled, build of the PDF.js library.
- Don't explicitly mention "ES5" in https://mozilla.github.io/pdf.js/getting_started/#download, since that's no longer accurate. As-is, this will more than likely give users the wrong impression regarding our *actual* browser support.
- Reword the "Getting the Code"-section of the README, since an ES6-compatible browser is no longer sufficient for development purposes given our usage of modern ECMAScript features.
- Update the "Building PDF.js"-section of the README, to explicitly mention both the `gulp generic` and `gulp generic-es5` commands.
Hopefully this will help reduce *some* user confusion, but we should perhaps also consider changing the "-es5"-suffix in our build-scripts and build-folders. (That would however require coming up with a "better" suffix, which might not be trivial.)
The whole purpose of showing a notification on the `sidebarToggle` button, when the sidebar is closed, was to give users *some* kind of indication that the PDF document contains outline/attachments/layers without having to manually open the sidebar to check.
However, in the implementation in PR 7959, I also added notifications for each view-buttons in the sidebar. Looking back at this, I've always questioned the value of the last part, since the view-buttons already have a `disabled`-state which shows if they're available or not. Hence we're actually, in a sense, duplicating notifications for the outline/attachments/layers-buttons without adding (in my opinion) all that much overall value.
All-in-all, I'm thus proposing that we only display the notification on the `sidebarToggle`-button itself, since that should really be sufficient here, which also allows us to simplify the relevant code a fair bit.
- aims to fix issue #12868: apply zoom factor to linewidth after setting it to 1.
- only apply 1px-width when required
- the sign of getSinglePixelWidth is used to know if 1px-width is required
- Remove a *duplicated* reference test, see "issue12810", from the manifest.
- Improve the spelling in a couple of comments in `src/core/canvas.js`, most notable of the word "parallelogram".
- Update a comment, also in `src/core/canvas.js`, to actually agree with the value used to reduce confusion when reading the code.
While PR 12725 fixed bug 1671312 as reported, i.e. the "In the upper right corner "Purposes' has bad kerning."-part, it however broke other parts of the text rendering.
Note in particular the tables, e.g. on page 2 and beyond, where the glyphs are now rendered too close together. The reason for this is that the fonts in question are non-embedded ArialNarrow, which we just replace with Helvetica which obviously is not narrow. Given that the font replacement isn't a perfect fit for non-embedded ArialNarrow, we still need to re-measure the glyph widths in this case.
Note first of all how the `PDFDocumentProxy.getJSActions` method in the API caches the result, which makes repeated lookups cheap enough to not really be an issue.
Secondly, with the previous patch, we're now only dispatching "pageopen"/"pageclose"-events when there's actually a sandbox that listens for them.
All-in-all, with these changes we can thus simplify the default-viewer "pageopen"-event handler a fair bit.
This patch is a rebased *and* refactored version of PR 9448, such that it applies cleanly given that `PDFFindController` has changed since that PR was opened; obviously keeping the original author information intact.
This patch will thus ensure that e.g. fractions, and other things that we normalize before searching, will still be highlighted correctly in the textLayer.
Furthermore, this patch also adds basic unit-tests for this functionality.
*Note:* The `[api-minor]` tag is added, since third-party implementations of the `PDFFindController` must now always use the `pageMatchesLength` property to get accurate length information (see the `web/text_layer_builder.js` changes).
Co-authored-by: Ross Johnson <ross@mazira.com>
Co-authored-by: Jonas Jenwald <jonas.jenwald@gmail.com>
The "pageopen"/"pageclose"-events are only necessary if, and only if, there's actually a sandbox to dispatch the events in. Hence we shouldn't dispatch those events unconditionally, as soon as `enableScripting` is set, but rather initialize that functionality only when needed.
Furthermore, in `web/app.js`, there's currently a bug since we're attempting to *manually* simulate a "pageopen"-event for a page that may not actually have been rendered at the time. With the modified `BaseViewer.initializeScriptingEvents` method, we'll now dispatch a correct "pageopen"-event here.
Not only was long text in popups no longer wrapped correctly, the
alignment was also center instead of left (or right, depending on the
locale used) for both text in popups and the other parts within the
annotation's section, such as the icon.
* add a comment to explain how minimal linewidth is computed.
* when context.linewidth < 1 after transform, firefox and chrome
don't render in the same way (issue #12810).
* set lineWidth to 1 after transform and before stroking
- aims fix issue #12295
- a pixel can be transformed into a rectangle with both heights < 1.
A single rescale leads to a rectangle with dim equals to 1 and
the other to something greater than 1.
* change the way to render rectangle with null dimensions:
- right now we rely on the lineWidth set before "re" but
it can be set after "re" and before "S" and in this case the rendering
will be wrong.
- render such rectangles as a single line.
Note that these changes were done automatically, using `gulp lint --fix`.
With this rule, we'll thus enforce a *consistent* formatting of zero-lengths in our CSS files.
Please find additional details about the Stylelint rule at https://stylelint.io/user-guide/rules/length-zero-no-unit
With the updated default viewer UI, some `dir`-dependent CSS rules are now redundant since *identical* rules are being specified for both LTR and RTL mode; after PR 12807 landed I've found even more of these cases.
Note in particular that the findbar-button rules can be simplified quite a bit, since there's a fair amount of unnecessary duplication in the CSS.
There's built-in ESLint rule, see `sort-imports`, to ensure that all `import`-statements are sorted alphabetically, since that often helps with readability.
Unfortunately there's no corresponding rule to sort `export`-statements alphabetically, however there's an ESLint plugin which does this; please see https://www.npmjs.com/package/eslint-plugin-sort-exports
The only downside here is that it's not automatically fixable, but the re-ordering is a one-time "cost" and the plugin will help maintain a *consistent* ordering of `export`-statements in the future.
*Note:* To reduce the possibility of introducing any errors here, the re-ordering was done by simply selecting the relevant lines and then using the built-in sort-functionality of my editor.
Given that the PDF document in the issue contains the same very large JPEG image *three* times, this patch includes a test-case where only the first page has been extracted from it.
Add support for finding/highlighting the outlineItem, corresponding to the currently visible page, in the sidebar (issue 7557, bug 1253820, bug 1499050)
This implementation is inspired by the behaviour in (recent versions of) Adobe Reader, since it leads to reasonably simple and straightforward code as far as I'm concerned.
*Specifically:* We'll only consider *one* destination per page when finding/highlighting the current outline item, which is similar to e.g. Adobe Reader, and we choose the *first* outline item at the *lowest* level of the outline tree.
Given that this functionality requires not only parsing of the `outline`, but looking up *all* of the destinations in the document, this feature can when initialized have a non-trivial performance overhead for larger PDF documents.
In an attempt to reduce the performance impact, the following steps are taken here:
- The "find current outline item"-functionality will only be enabled once *one* page has rendered and *all* the pages have been loaded[1], to prevent it interfering with data regular fetching/parsing early on during document loading and viewer initialization.
- With the exception of a couple of small and simple `eventBus`-listeners, in `PDFOutlineViewer`, this new functionality is initialized *lazily* the first time that the user clicks on the `currentOutlineItem`-button.
- The entire "find current outline item"-functionality is disabled when `disableAutoFetch = true` is set, since it can easily lead to the setting becoming essentially pointless[2] by triggering *a lot* of data fetching from a relatively minor viewer-feature.
- Fetch the destinations *individually*, since that's generally more efficient than using `PDFDocumentProxy.getDestinations` to fetch them all at once. Despite making the overall parsing code *more* asynchronous, and leading to a lot more main/worker-thread message passing, in practice this seems faster for larger documents.
Finally, we'll now always highlight an outline item that the user manually clicked on, since only highlighting when the new "find current outline item"-functionality is used seemed inconsistent.
---
[1] Keep in mind that the `outline` itself already isn't fetched/parsed until at least *one* page has been rendered in the viewer.
[2] And also quite slow, since it can take a fair amount of time to fetch all of the necessary `destinations` data when `disableAutoFetch = true` is set.
Given that the API will now, after PR 12039, automatically pick the correct factories to use depending on the environment (browser vs. Node.js), we can utilize that in the unit-tests as well. This way we don't have to manually repeat the same initialization code in *multiple* unit-tests.
*Note:* The *official* PDF.js API is defined in `src/pdf.js`, hence the new exports in `src/display/api.js` will not affect that.
Also, updates the unit-test `FileReaderFactory` helpers similarily.
*Drive-by change:* Fix the `CMapReaderFactory` usage in the annotation unit-tests, since the cache should only contain raw data and not a Promise. While this obviously works as-is, having unit-tests that "abuse" the intended data format can easily lead to unnecessary failures if changes are made to the relevant `src/core/` code.
Given that we're already using the `external/dist/` folder for things we simply want to copy to pdfjs-dist during building, this patch *should* hopefully work since it's based on the suggestion in issue 12827.
As long as this only requires a *single/small* file, to fix the TypeScript definitions in es5-builds, this solution seem acceptable as far as I'm concerned. (Although, please note that I don't know enough about TypeScript to actually test the patch.)
With the code dispatching a "pageopen" event on the existing (general) `BaseViewer` event "pagesinit", in practice this means that the `Set` is always being created. Hence we can simplify the method overall, by always initializing the `this._pageOpenPendingSet` property.
Given that "pageopen" events are not guaranteed to occur, if the page becomes inactive *before* it finishes rendering, we should probably also avoid dispatching a "pageclose" event in that case to avoid confusing/inconsistent state in any event handlers.
Currently any errors thrown in `preEvaluateFont`, which is a *synchronous* method, will not be handled at all in the `loadFont` method and we were thus failing to return an `ErrorFont`-instance as intended here.
Also, add an *explicit* check in `PartialEvaluator.preEvaluateFont` to ensure that Type0-fonts always have a *valid* dictionary.
Ensure that `PDFViewerApplication._contentLength` is always updated with the *correct* length, as returned by `PDFDocumentProxy.getDownloadInfo`, and only let the `PDFViewerApplication._initializeMetadata` method overwrite if it's not already been set.
Finally, in `PDFViewerApplication._initializeJavaScript`, the fallback `_contentLength` handling is now moved to just after the fallback `documentInfo` handling, such that all the fallback code is in one place within the method.
There's been a number of (somewhat) recent issues where people are having trouble using pdfjs-dist together with `webpack`, since that library purposely doesn't declare any dependencies; refer to PR 11474 for additional context.
In an *attempt*, although I don't know how much this will actually help in practice (given my limited `webpack` experience), let's try to list `worker-loader` as a *peer*-dependency to see if that helps. This should, unless I'm completely misunderstanding https://docs.npmjs.com/cli/v6/configuring-npm/package-json#peerdependencies, prevent `worker-loader` from being installed by default for *all* pdfjs-dist users while still indicating the dependency for those who need it.
Once the next PDF.js release is made, the `webpack` example will no longer work since the non-translated builds now use ECMAScript features not supported by older `webpack`-versions.
With the updated default viewer UI, a couple of `dir`-dependent CSS rules have now become redundant since *identical* rules are being specified for both LTR and RTL mode.
Furthermore, there's also some unnecessary re-defining of the `toolbarButton`/`secondaryToolbarButton`-icon related CSS rules.
Finally, for the toggle-buttons there's a particular styling applied to the `:hover:active` state, however the color wasn't defined with CSS variables.
With the updated default viewer UI, a couple of the toolbarButton icons are now *vertically* symmetrical; hence we can remove some now unneeded `transform: scaleX(-1);` rules from the viewer CSS.
Note how the `onerror` functionality is not being used in the GENERIC `DownloadManager`, since we have no way of knowing if downloading succeeded.
Hence this functionality is only *possibly* useful in MOZCENTRAL builds, however as outlined in the existing comments it's unlikely to be helpful in practice. Generally speaking, if downloading failed once in [`PdfStreamConverter.jsm`](https://searchfox.org/mozilla-central/rev/809ac3660845fef6faf18ec210232fdadc0f1ad9/toolkit/components/pdfjs/content/PdfStreamConverter.jsm#294-406) it seems very likely that it would fail again; all-in-all I'm thus suggesting that we just remove the `onerror` functionality altogether here.
Currently this code is duplicated no less than three times in the `web/app.js` file, and by introducing a helper method we can avoid unnecessary repetition.
There's a fair number of cases where `FirefoxCom.request`-calls are manually wrapped in a Promise to make it asynchronous. We can reduce the amount of boilerplate code in these cases by introducing a new `FirefoxCom.requestAsync` method instead.
Furthermore, a couple of `FirefoxCom.request`-calls in the `DownloadManager` are also changed to be asynchronous rather than using callback-functions.
With this patch, we're thus able to replace a lot of *direct* usages of `FirefoxCom.request` with the new `FirefoxCom.requestAsync` method instead.
*Please note:* It's highly recommended to ignore whitespace-only changes when looking at this patch.
Besides modernizing this code, by converting it to a standard class, the existing JSDoc comments are updated to actually agree better with the way that this functionality is used now. (The next patch will reduce usage of `FirefoxCom.request` significantly, hence the JSDocs for the optional `callback` is removed to not unnecessarily advertise that functionality.)
Finally, the unnecessary/unused `return` statement at the end of `FirefoxCom.request` is also removed.
This is the "modern" way of removing a node from the DOM, which has the benefit of being a lot shorter and more concise.
Also, this patch removes the `return` statement from the "pdf.js.response" event listener, since it's always `undefined`, given that none of the `callback`-functions used here ever return anything (and don't need to either). Generally speaking, returning a value from an event listener isn't normally necessary either.
This method currently accepts a callback-function, which does feel a bit old fashioned now. At the time that this code was introduced, native Promises didn't exist yet and there's a custom Promise-implementation used instead.
However, today with Promises and async/await being used *a lot* it seems reasonable to change `DefaultExternalServices.fallback` to an `async` method instead such that the callback-function can be removed.
Note how the end of the `{PDFOutlineViewer, PDFAttachmentViewer, PDFLayerViewer}.render` methods share *almost* identical code, hence we can reduce some duplication by introducing the new `BaseTreeViewer` helper method here.
Furthermore, setting `this._lastToggleIsShow` can be made ever so slightly more efficient, since we don't care about the number of ".treeItemsHidden"-classes but only want to know if at least one exists.
This follows the same principle as the `once` option that exists in the native `addEventListener` method, and will thus automatically remove an `EventBus` listener when it's invoked; see https://developer.mozilla.org/en-US/docs/Web/API/EventTarget/addEventListener#Parameters
Finally, this patch also tweaks some the existing `EventBus`-code to use modern features such as optional chaining and logical assignment operators.
Given that we already have a `PresentationModeState`-enumeration, we should use that with the "presentationmodechanged" event rather than including separate properties. Note that this new behaviour, of including an enumeration-value in the event, is consistent with lots of other existing viewer-events.
To hopefully avoid issues in custom implementations of the default viewer, any attempt to access the removed properties will now throw.
Similar to e.g. the "locale" option, this in *only* done for those build-targets where the "sandboxBundleSrc" is actually defined.
With these changes we can remove an `AppOptions` dependency from the `web/generic_scripting.js` file, thus limiting *direct* `AppOptions` usage in the default viewer files.
Given that the `dispatchEventInSandbox` method (on the scripting-classes) is asynchronous, there's a very real risk that the events won't be dispatched/handled until *after* their associated functionality has actually run (with the "Will..." events being particularily susceptible to this issue).
To reduce the likelihood of that happening, we can simply `await` the `dispatchEventInSandbox` calls as necessary. A couple of methods are now marked as `async` to support these changes, however that shouldn't be a problem as far as I can tell.
*Please note:* Given that the browser "beforeprint"/"afterprint" events are *synchronous*, we unfortunately cannot await the `WillPrint`/`DidPrint` event dispatching. To fix this properly the web-platform would need support for asynchronous printing, and we'll thus have to hope that things work correctly anyway.
Note that currently the `DidSave` event is not *guaranteed* to actually be dispatched if there's any errors during saving, which is easily fixed by simply moving it to occur in the `finally`-handler in `PDFViewerApplication.save` method.
For the `WillPrint`/`DidPrint` events, things are unfortunately more complicated. Currently these events will *only* be dispatched iff the printing request comes from within the viewer itself (e.g. by the user clicking on the "Print" toolbar button), however printing can be triggered in a few additional ways:
- In the GENERIC viewer:
- By the <kbd>Ctrl</kbd>+<kbd>P</kbd> keyboard shortcut.
- In the MOZCENTRAL viewer, i.e. the Firefox built-in viewer:
- By the <kbd>Ctrl</kbd>+<kbd>P</kbd> keyboard shortcut.
- By the "Print" item, as found in either the Firefox "Hamburger menu" or in the browser-window menu.
In either of the cases described above, no `WillPrint`/`DidPrint` events will be dispatched. In order to *guarantee* that things work in the general case, we thus have to move the `dispatchEventInSandbox` calls to the "beforeprint"/"afterprint" event handlers instead.
Similar to other markers that we currently skip, by ignoring unsupported Coding style default (COD) options we'll at least render *something* here (although some JPEG 2000 images may look slightly wrong).
Note that if the unsupported COD options lead to additional errors, during parsing, we'll still abort parsing of the JPEG 2000 image.
Rather than calling `getJavaScript` in the API and then ignoring the result, when "enableScripting" is set, it should be more efficient/faster to simply skip it altogether instead.
Finally, the `setTimeout` call at the end of `PDFViewerApplication._initializeAutoPrint` is removed, since it doesn't seem necessary any more as far as I can tell.[1]
Note that when this functionality was originally added, back in PR 2839, it seems that `pagesPromise` simply waited for the `getPage` calls of *all* pages to resolve. Today, on the other hand, the viewer fetches *and* renders the first page *before* doing the remaining `getPage` calls, and only afterwards is `pagesPromise` resolved. Hence it's not really clear why we now need to delay printing even further with a `setTimeout` call.
---
[1] The patch was tested with the following documents: https://github.com/mozilla/pdf.js/blob/master/test/pdfs/bug1001080.pdf and https://github.com/mozilla/pdf.js/blob/master/test/pdfs/issue6106.pdf
The `gulp components` task is only necessary when running the reference-tests, since they use the `SimpleLinkService` during the `annotationLayer` sub-tests.
However, unit-tests don't actually use any part of the `gulp components` build, and we can thus reduce the overall runtime of the standalone unit-tests by not building unnecessary files.
This will, in a very simple way using the existing events, thus allow the viewer to remove the "beforeunload" `window` event listener when the document is closed.
Generally speaking we want to avoid having *global* event listeners for the PDF document instance, which is why the `EventBus` exists, and instead reserve global events for the viewer itself. However, the `AnnotationStorage` "beforeunload" event unfortunately needs to be document-specific and we should thus ensure that it's correctly removed when the document is destroyed.
These callbacks should not be necessary *before* the document has been initialized. Furthermore, move the functionality to a new helper-method since `PDFViewerApplication.load` is already quite large.
Given that this relies on accessing properties on the `PDFDocumentProxy`-instance, it seems more appropriate for this code to live in `PDFViewerApplication`.
It seems that the timeout is way too short in practice, since this new integration-test failed *intermittently* already in PR 12702 (which is where the test was added).
The ideal solution here would be to simply await an event, dispatched by the viewer, however that unfortunately doesn't appear to be supported by Puppeteer.
Instead, the solution implemented here is to add a new method in `PDFViewerApplication` which Puppeteer can query to check if the scripting/sandbox has been fully initialized.
There's really no point, as far as I can tell, to attempt to dispatch an event in a non-existent sandbox. Generally speaking, even trying to do this *could* possibly even lead to errors in some cases.
Furthermore, utilize optional chaining to simplify some `dispatchEventInSandbox` calls throughout the viewer.
Finally, replace superfluous `return` statements with `break` in the switch-statement in the `updateFromSandbox` event-handler.
There's no really compelling reason, as far as I can tell, to introduce the `ENABLE_SCRIPTING` build-target, instead of simply re-using the existing `TESTING` build-target for the new `gulp integrationtest` task.
In general there should be no problem with just always enable scripting in TESTING-builds, and if I were to *guess* the reason that this didn't seem to work was most likely because the Preferences ended up over-writing the `AppOptions`.
As it turns out the GENERIC-viewer has already has built-in support for disabling of Preferences, via the `AppOptions`, and this can be utilized in TESTING-builds as well to ensure that whatever `AppOptions` are set they're always respected.
For DOM events all event names are lower-case, and the newly added PDF.js scripting-events thus "stick out" quite a bit. Even more so, considering that our internal `eventBus`-events follow the same naming convention.
Hence this patch, which changes the "updateFromSandbox"/"dispatchEventInSandbox" events to be lower-case instead.
Furthermore, using DOM events for communication *within* the PDF.js code itself (i.e. between code in `web/app.js` and `src/display/annotation_layer.js/`) feels *really* out of place.
That's exactly the reason that we have the `EventBus` abstraction, since it allowed us to remove prior use of DOM events, and this patch thus re-factors the code to make use of the `EventBus` instead for scripting-related events.
Obviously for events targeting a *specific element* using DOM events is still fine, but the "updatefromsandbox"/"dispatcheventinsandbox" ones should be using the `EventBus` internally.
*Drive-by change:* Use the `BaseViewer.currentScaleValue` setter unconditionally in `PDFViewerApplication._initializeJavaScript`, since it accepts either a string or a number.
- Update the `LinkAnnotationElement._bindJSAction` call-site to actually agree with the JSDocs, by passing in the `data`.
- Prevent the links created by `LinkAnnotationElement._bindJSAction` from being displayed with empty hashes; compare with e.g. `LinkAnnotationElement. _bindNamedAction`.
- The overall indentation-level in `WidgetAnnotationElement._setEventListener` can be reduced slightly by using early returns, which improves the overall readability of this method a bit. (We're also able to avoid unnecessary `in` usage here.)
- The code can also be made *slightly* more efficient overall, by moving the `this.data.actions` check into `WidgetAnnotationElement._setEventListeners` instead. This way we can avoid useless `this._setEventListener`-calls when there are no actions present.
- Actually remove the `isDown` property when destroying the scripting-instance.
- Mark all `mouseState` usage as "private" in the various classes.
- Ensure that the `AnnotationLayer` actually treats the parameter as properly *optional*, the same way that the viewer components do.
- For now remove the `mouseState` parameter from the `PDFPageView` class, and keep it only on the `BaseViewer`, since it's questionable if all of the scripting-functionality will work all that well without e.g. a full `BaseViewer`.
- Append the `mouseState` to the JSDoc for the `AnnotationElement` class, and just move its definition into the base-`AnnotationElement` class.
* the goal is to execute actions like Open or OpenAction
* can be tested with issue6106.pdf (auto-print)
* once #12701 is merged, we can add page actions
Similar to other markers that we currently skip, by ignoring the Coding style component (COC) marker we'll at least prevent outright errors (although some JPEG 2000 images may look slightly wrong).
This new event essentially mirrors the existing "pagesinit" event, and will allow e.g. a custom implementation of the viewer to be notified before the current PDF document is removed from the viewer.
By using this new event, we're thus able to dispatch a "pageclose" event for JavaScript actions when closing the existing document.
Having looked at the Acrobat JavaScript specification, see https://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/AcrobatDC_js_api_reference.pdf#G5.1963437, I suppose that introducing these two new events is probably the easiest solution overall.
However there's a number of things that, as far as I'm concerned, will help the overall implementation:
- Only dispatch these new events when `enableScripting = true` is set.
- Handle them *separately* from the existing "pagechanging" event dispatching, to avoid too much clutter.
- Don't dispatch either of the events if the page didn't actually change.
- When waiting for pages to render, don't dispatch "pageopen" if the page is no longer active when rendering finishes.
- Ensure that we only use *one* "pagerendered" event listener.
- Ensure that "pageopen" is actually dispatched when the document loads.
I suppose that we *could* avoid adding the "pageclose" event, and use the existing "pagechanging" event instead, however having a separate event might allow more flexibility in the future. (E.g. I don't know if we'll possibly want to dispatch "pageclose" on document close, as mentioned briefly in the specification.)
* move set/clear|Timeout/Interval and crackURL code in pdf.js
* remove the "backdoor" in the proxy (used to dispatch event) and so return the dispatch function in the initializer
* remove listeners if an error occured during sandbox initialization
* add support for alert and prompt in the sandbox
* add a function to eval in the global scope
It appears that the PDF document in [bug 1292316](https://bugzilla.mozilla.org/show_bug.cgi?id=1292316) now renders "correctly"[1] when compared to e.g. Adobe Reader and PDFium. Most likely this bug was fixed by a *somewhat* recent patch, or patches, to the `XRef.indexObjects` method.
Before just closing [bug 1292316](https://bugzilla.mozilla.org/show_bug.cgi?id=1292316) as WFM, I figured that it probably can't hurt to add it as a new test-case to avoid accidentally regressing this document in the future.
---
[1] Given that the XRef table is corrupt, and that we're forced to recover, there's generally speaking probably some question as to what actually constitutes "correct" in this case.
Currently only the `gulp unittest` task actually set the time-zone, which means that locally I'm now getting failures with e.g. `gulp test`.
*Please note:* I firmly believe that the unit-tests in question should be re-written, since even with this patch applied there's failures when running http://localhost:8888/test/unit/unit_test.html directly in a browser.
There doesn't seem to be anything definitive about this in
the spec, but from experimenting, it seems acrobat lets
PDFs override the widths of the standard fonts.
Given that the GENERIC default viewer supports opening more than one document, and that a unique scripting-instance is now used for each document, the changes made in this patch seem appropriate.
While it's not entirely clear to me that it's ultimately desirable to use the `pdf.sandbox.js` in the Chromium-extension, given that the MOZCENTRAL-build uses `pdf.scripting.js` directly in a *custom* sandbox, the current state isn't that great since setting `enableScripting = true` with the Chromium-extension will currently fail completely.
Hence this patch, which should at least unbreak things for now.
Since the `close` method has become quite large, this small re-factoring shouldn't hurt (and may also be useful with future changes to the `_initializeJavaScript` method).
I completely missed this previously, but we obviously should remove the scriptElement as well to *really* clean-up everything properly.
Given that there's multiple existing usages of `loadScript` in the code-base, the safest/quickest solution seemed to be to have call-sites opt-in to remove the scriptElement using a new parameter.
This patch *attempts* to actually implement what's described for the `Count`-entry in the PDF specification, see https://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#G11.2095911, which I mostly ignored back in PR 10890 since it seemed unnecessarily complicated[1].
Besides issue 12704, I've also tested a couple of other documents (e.g. the PDF specification) and these changes don't *seem* to break anything else; additional testing would be helpful though!
---
[1] At the time, all PDF documents that I tested worked even with a very simple approach and I thus hoped that it'd would suffice.
Similar to the previous patch, the GENERIC default viewer is capable of opening more than *one* PDF document and we should ensure that we handle that case correctly.
I was actually quite surprised to find that, despite the various `scripting`-getters implementing `destroySandbox` methods, there were no attempts at actually cleaning-up either the "sandbox" or removing the globally registered event listeners.
This patch also changes the method to skip *all* data fetching when "enableScripting" isn't active. Finally, simplifies some event-data accesses in the "updateFromSandbox" listener.
Another possible option here could be to use the `contentLength`, when it exists, and then using e.g. a custom event to always update the "filesize" in the sandbox "after the fact" with the result of the `getDownloadInfo`-call.
We can easily avoid unnecessary API-calls here, since most of the time the `metadata` will already be available here. In the *rare* case that it's not available, we can simply wait for the existing `getMetadata`-call to resolve.
This will be useful in the following patch, and note that there's also an old issue (see 5765) which asked for such an event. However, given that the use-case wasn't *clearly* specified, and that we didn't have an internal use for it at the time it wasn't implemented.
Also, ensure that all of the metadata-related properties are actually reset when the document is closed.
This simplifies not just this code, but the unit-tests as well, and should be sufficient as far as I can tell.
Note also that currently, in the *built* `pdf.sandbox.js` file, there's even a line reading `testMode = testMode && false;` because of an accidentally flipped pre-processor statement.
Finally, in the `scripting_spec.js` unit-test, defines `sandboxBundleSrc` at the top of the file to make it easier to find and/or change it when necessary.
Compared to the, previously removed, `sandbox`/`watch-sandbox` gulp-tasks, these ones should work even when run against an non-existent/empty `build`-folder.
Also, to ensure that the development viewer actually works out-of-the-box, `gulp server` will now also include `gulp watch-dev-sandbox` to remove the need to *manually* invoke the build-tasks.
Finally, this patch also removes the `web/devcom.js` file since it shouldn't actually be needed, assuming that the "sandbox"-loading code in the `web/genericcom.js` file is actually *correctly* implemented.
The way that the `pdf.sandbox.js` building was implemented feels all kinds of inconsistent/wrong, and it "sticks out" quite a bit when compared to the rest of the `gulpfile.js`. This patch thus attempts to improve the current situation slightly, to hopefully make future maintenance easier.
One thing that strikes you, pretty immediately, when looking at PR 12604 is that the two new `gulp`-tasks added (i.e. `sandbox` and `watch-sandbox`) don't even work!?
The reason for this is that they implicitly dependent upon the result of the `buildnumber`-task, which isn't listed as a dependency. (Try running `gulp clean` *first*, and invoking any of the new `gulp`-tasks will inevitably fail.)
Furthermore, there's another (potentially big) problem with the implementation of e.g. the `gulp sandbox` task, since it doesn't actually wait for all building to complete before the task is considered as "done". This has the potential to cause all sorts of subtle bugs elsewhere, and the fact that things even "work" as-is can probably be attributed mostly to luck.
Unfortunately there's no *perfect* way to improve things here, since the `pdf.sandbox.js` file depends on including the `pdf.scripting.js` file as a string, however I firmly believe that improvements are still possible here.
To that end, this patch updates all relevant build-targets to create a *temporary* `pdf.scripting.js` file as part of the setup in the `gulp`-tasks, and then reads that file during the `pdf.sandbox.js` building.
This at least allows us to bring all of this "sandbox"-build code much more in-line with the existing build-system.
Given the somewhat "specialized" nature of the `pdf.sandbox.js` building, it ought to be possible to re-factor how some of the options are handled.
Note in particular that the `gulp-strip-comments` dependency seems somewhat unncessary, since the *main* source of comments are just the default license header. Hence I seems much more reasonable to simply not include that to begin with, rather than removing it after the fact (the few remaining Webpack-related should be few/small enough to not really matter much in practice).
This way we're able to further reduce the special-casing related to the `pdf.sandbox.js`-building, which will make future changes/maintenance easier by bringing this code more in-line with existing patterns in `gulpfile.js`.
(If we really want to reduce the filesize, we might want to consider always minifying the `GENERIC`-build of the `pdf.sandbox.js` file.)
Each quadrilateral needs to have its own link element, so the first
quadrilateral can use the already created element, but the next
quadrilaterals need to clone that element.
Not only does this reduce boilerplate since the documentation is the
same for all annotation classes, it also wasn't correct for the
annotation types that support quadpoints since they return an array of
section elements instead of a single one.
There's no good reason, as far as I can tell, to use search-and-replace to include the *stringified* `pdf.scripting.js` file in the built `pdf.sandbox.js` file. Instead we could, and even should, utilize the existing `PDFJSDev.eval(...)`-functionality, which is not only simpler but will also be more efficient as well (no need for a regular expression).
The current location feels somewhat strange, and also inconsistent with the existing way that bundling is done.
Finally, add the version/build numbers at the top of the *built* `pdf.sandbox.js` files, since all other built files include that information given that it's often helpful to be able to easily determine the *exact* version.
[Regression] Prevent the *built* `pdf.scripting.js`/`pdf.sandbox.js` files from accidentally including most of the main-thread code (PR 12631 follow-up)
*This is a recent regression, which I stumbled upon while working on cleaning-up the gulpfile related to `pdf.sandbox.js` building.*
By placing the `ColorConverters` functionality in the `src/display/display_utils.js` file, you end up including a *significant* chunk of the `pdf.js` file in the built `pdf.scripting.js`/`pdf.sandbox.js` files.
Given that I cannot imagine that this was actually intended, since it inflates the built files with unnecessary/unused code, this moves `ColorConverters` to a new file instead (thus breaking the dependencies).
To hopefully reduce the risk future bugs, along these lines, a big comment is also placed at the top of the new file.
Finally, the `ColorConverters` is converted to a class with static methods, since this felt slightly cleaner overall.
Rather than having two slightly different ways of setting the pending/notFound appearance on the "findInput", we can simply use "data-status" in both cases since they're obviously mutually exclusive.
This seems like a very minor issue, since in general we can't really help if domains are blocked from certain networks, however in this particular case I suppose that using the Internet Archive should work.
As mentioned in https://github.com/mozilla/pdf.js/wiki/Frequently-Asked-Questions#faq-support, PDF.js version `2.6.347` is the last release with IE 11/Edge support.
Hence we should now be able to reduce unnecessary duplication in the default viewer image resources, note the files in the `web/images/` folder with a `-dark` suffix, by using only *one* SVG-image for each icon and letting the `background-color` depend on the CSS theme instead.
For the `gulp mozcentral` build-target, the resulting `web/images/` folder is reduced from `43 997` to `28 566` bytes (~35 percent).
*Please note:* I don't really know if this implementation is necessarily the *best* solution, but it seems to work well enough in e.g. Firefox Nightly and Google Chrome Beta as far as my testing goes.
In addition to the existing /Root and /Pages validation, also check that the /Pages-entry actually is a dictionary and that it has a valid /Count-entry.
This way we can avoid picking a trailer candidate which e.g. the `Catalog.numPages` getter will just end up rejecting, thus breaking PDF document loading completely.
Given that we already include the "Content-Disposition"-header filename, when it exists, it shouldn't hurt to also include the information from the "Content-Length"-header.
For PDF documents opened via a URL, which should be a very common way for the PDF.js library to be used, this will[1] thus provide a way of getting the PDF filesize without having to wait for the `getDownloadInfo`-promise to resolve[2].
With these API improvements, we can also simplify the filesize handling in the `PDFDocumentProperties` class.
---
[1] Assuming that the server is correctly configured, of course.
[2] Since that's not *guaranteed* to happen in general, with e.g. `disableAutoFetch = true` set.
* quickjs-eval.js has been generated using https://github.com/mozilla/pdf.js.quickjs/
* lazy load of sandbox code
* Rewrite tests to use the sandbox
* Add a task `watch-sandbox` which update bundle pdf.sandbox.js on change in the sandbox code
Given that it's generally faster to call *one* function and have it loop through an object, rather than looping through an object and calling a function for every iteration, this patch will reduce the total time spent in `PDFViewerApplication._readPreferences` ever so slightly.
Also, over time we've been adding more and more preferences, rather than removing them, so using the new `AppOptions.setAll` method should be generally beneficial as well.
While the effect of these changes is quite small, it does reduces the time it takes for the preferences to be fully initialized. Given the amount of asynchronous code during viewer initialization, every bit of time that we can save should thus help.
Especially considering the recently added `viewerCssTheme` preference, which needs to be read very early to reduce the risk of the viewer UI "flashing" visibly as the theme changes, I figured that a couple of small patches reducing the time spend reading preferences cannot hurt.
Given that only two debugging hash parameters (i.e. `disableWorker` and `pdfBug`) will make this method asynchronous, we can avoid what's most of the time is an unnecessary `Promise.all` invocation.
While this does work pretty well in my quick testing, it's *very much* a hack since as far as I can tell there's no support in the CSS specification for using e.g. a CSS variable to override a `@media (prefers-color-scheme: dark) {...}` block.
The solution implemented here is thus to *edit* the viewer CSS, by either removing the entire `@media ...` block in light-mode or by ensuring that its rules become *unconditionally* applied in dark-mode.
To simplify the overall implementation, since all of this does seem like somewhat of an edge-case, the `viewerCssTheme` preference will *only* be read during viewer initialization. (Similar to many other existing preferences, a reload is thus required when changing it.)
Originally the default preferences were defined in a JSON-file checked into the repository, which was loaded using SystemJS in development mode.
Over the years a number of changes have been made to this code, most notably:
- The preferences JSON-file is now generated automatically, during building, from the `AppOptions` abstraction.
- All SystemJS usage has been removed from the development viewer.
Hence the default preferences are now available *synchronously* even in the development viewer, and it's thus no longer necessary to defer to the microtask queue (since `getDefaultPreferences` is async) just to get the default preferences.
While the effect of these changes is quite small, it *does* reduces the time it takes for the preferences to be fully initialized. Given the amount of asynchronous code during viewer initialization, every bit of time that we can save should thus help.
- Add support for logical assignment operators, i.e. `&&=`, `||=`, and `??=`, with a Babel-plugin. Given that these required incrementing the ECMAScript version in the ESLint and Acorn configurations, and that platform/browser support is still fairly limited, always transpiling them seems appropriate for now.
- Cache the `hasJSActions` promise in the API, similar to the existing `getAnnotations` caching. With this implemented, the lookup should now be cheap enough that it can be called unconditionally in the viewer.
- Slightly improve cleanup of resources when destroying the `WorkerTransport`.
- Remove the `annotationStorage`-property from the `PDFPageView` constructor, since it's not necessary and also brings it more inline with the `BaseViewer`.
- Update the `BaseViewer.createAnnotationLayerBuilder` method to actaually agree with the `IPDFAnnotationLayerFactory` interface.[1]
- Slightly tweak a couple of JSDoc comments.
---
[1] We probably ought to re-factor both the `IPDFTextLayerFactory` and `IPDFAnnotationLayerFactory` interfaces to take parameter objects instead, since especially the `IPDFAnnotationLayerFactory` one is becoming quite unwieldy. Given that that would likely be a breaking change for any custom viewer-components implementation, this probably requires careful deprecation.
*Note that I wasn't able to reproduce the issue in Firefox, but only in Chromium-browsers.*
The bug, and it's feels almost trivial once you've found it, is that we're not passing the `transform` parameter as intended to `PDFPageProxy.render` when drawing thumbnails on HiDPI displays. Instead the canvas context is, for reasons that I don't even pretent to understand, *manually* scaled in `PDFThumbnailView._getPageDrawContext`, which thus doesn't guarantee that the `baseTransform` property on the `CanvasGraphics`-instances becomes correct.
The solution is really simple though, just handle the `transform` the same way in `PDFThumbnailView.draw` as in `PDFPageView.paintOnCanvas` and things should just work.
*This is a pre-existing issue that I noticed while working on PR 12613, and fixing this also brings the thumbnail code inline with the page code.*
Given the intermittent nature of all of this, it's somewhat difficult to reproduce it consistently; however the following steps should at least provide an outline:
1. Open the sidebar, and the thumbnailView, and start scrolling around.
2. *Quickly* close the sidebar, so that all thumbnails won't have time to finish rendering.
3. Either wait for the cleanup-timeout to occur, or simply run `PDFViewerApplication.cleanup()` in the console.
What *intermittently* happens here is that `WorkerTransport.startCleanup` rejects, and consequently that cleanup doesn't complete as intended, since some of the thumbnails are left in a *pending* renderingState[1].
Fixing this is simple though, and only requires updating `PDFThumbnailViewer.cleanup` along the lines of `BaseViewer.cleanup`.
---
[1] Keep in mind that thumbnails will *only* render when the thumbnailView is visible, to reduce resource usage.
This patch will help reduce memory usage, especially for longer documents, when the user scrolls around in the thumbnailView (in the sidebar).
Note how the `PDFPageProxy.cleanup` method will, assuming it's safe to do so, release main-thread resources associated with the page. These include things such as e.g. image data (which can be arbitrarily large), and also the operatorList (which can also be quite large).
Hence when pages are evicted from the `PDFPageViewBuffer`, on the `BaseViewer`-instance, the `PDFPageView.destroy` method is invoked which will (among other things) call `PDFPageProxy.cleanup` in the API.
However, looking at the `PDFThumbnailViewer`/`PDFThumbnailView` classes you'll notice that there's no attempt to ever call `PDFPageProxy.cleanup`, which implies that in certain circumstances we'll essentially keep all resources allocated permanently on the `PDFPageProxy`-instances in the API.
In particular, this happens when the users opens the sidebar and starts scrolling around in the thumbnails. Generally speaking you obviously need to keep all thumbnail *images* around, since otherwise the thumbnailView is useless, but there's still room for improvement here.
Please note that the case where a *rendered page* is used to create the thumbnail is (obviously) completely unaffected by the issues described above, and this rather only applies to thumbnails being explicitly rendered by the `PDFThumbnailView.draw` method.
For the latter case, we can fix these issues simply by calling `PDFPageProxy.cleanup` once rendering has finished. To prevent *accidentally* pulling the rug out from under `PDFPageViewBuffer` in the viewer, which expects data to be available, this required adding a couple of new methods[1] to enable checking that it's indeed safe to call `PDFPageProxy.cleanup` from the `PDFThumbnailView.draw` method.
It's really quite fascinating that no one has noticed this issue before, since it's been around since basically "forever".
---
[1] While it should be *very* rare for `PDFThumbnailView.draw` to be called for a pageView that's also in the `PDFPageViewBuffer`, given that pages are rendered before thumbnails and that the *rendered page* is used to create the thumbnail, it can still happen since rendering is asynchronous.
Furthermore, it's also possible for `PDFThumbnailView.setImage` to be disabled, in which case checking the `PDFPageViewBuffer` for active pageViews *really* matters.
* remove 1st param of _createPopup (almost useless for a method)
* prepend popup div to avoid to have them on top of some highlights (and so "disable" partially mouse events)
* add a ref test for issue #12504
* https://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/js_api_reference.pdf#page=335
* it has all the properties/methods defined in the spec
* unimplemented methods are there but with an empty body to avoid exception when calling an undefined method
* implement zoom, zoomType, layout, pageNum, ...
* in some pdf, there are actions with "event.source.hidden = ..."
* in order to handle visibility when printing, annotationStorage is extended to store multiple properties (value, hidden, editable, ...)
* When no actions then set it to null instead of empty object
* Even if a field has no actions, it needs to listen to events from the sandbox in order to be updated if an action changes something in it.
By using optional chaining, see https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Operators/Optional_chaining, it's possible to reduce unnecessary code-repetition in many cases.
Note that these changes also reduce the size of the *built* `pdf.js` file, when `SKIP_BABEL == true` is set, and for the `MOZCENTRAL` build-target that result in a `0.1%` filesize reduction from a simple and mostly mechanical code change.
Note that a number of these cases are covered by existing unit-tests, and a few others only matter for the development/build scripts.
Furthermore, I've also tried to the best of my ability to test each case *manually* to hopefully further reduce the likelihood of this patch introducing any bugs.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/no-useless-escape
The vast majority of the time, unless a Pattern is active, the `strokeColor`-property contains a "simple" colour value represented by a String. Hence it seems somewhat ridiculous to do a `hasOwnProperty` check on a String, and it's should thus be possible to improve things a tiny bit here.
Unfortunately using a simple `instanceof` check would only work for `TilingPattern`s, but not for the `ShadingIRs` given how they are implemented; see `src/display/pattern_helper.js`. (While that file could probably do with some clean-up, given the age of some of its code, that probably shouldn't happen here.)
Finally, the `this.type = "Pattern"`-property of the various Shadings/TilingPatterns were removed, since I cannot see why it's necessary when we can simply check for a `getPattern` method instead. Note that part of this code even pre-dates the main/worker-thread split, which probably in part explains why it looks the way it does.
The `PartialEvaluator.hasBlendModes` method is necessary to determine if there's any blend modes on a page, which unfortunately requires *synchronous* parsing of the /Resources of each page before its rendering can start (see the "StartRenderPage"-message).
In practice it's not uncommon for certain /Resources-entries to be found on more than one page (referenced via the XRef-table), which thus leads to unnecessary re-fetching/re-parsing of data in `PartialEvaluator.hasBlendModes`.
To improve performance, especially in pathological cases, we can cache /Resources-entries when it's absolutely clear that they do not contain *any* blend modes at all[1]. This way, subsequent `PartialEvaluator.hasBlendModes` calls can be made significantly more efficient.
This patch was tested using the PDF file from issue 6961, i.e. https://github.com/mozilla/pdf.js/files/121712/test.pdf:
```
[
{ "id": "issue6961",
"file": "../web/pdfs/issue6961.pdf",
"md5": "a80e4357a8fda758d96c2c76f2980b03",
"rounds": 100,
"type": "eq"
}
]
```
which gave the following results when comparing this patch against the `master` branch:
```
-- Grouped By browser, page, stat --
browser | page | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ---- | ------------ | ----- | ------------ | ----------- | ---- | ------ | -------------
firefox | 0 | Overall | 100 | 1034 | 555 | -480 | -46.39 | faster
firefox | 0 | Page Request | 100 | 489 | 7 | -482 | -98.67 | faster
firefox | 0 | Rendering | 100 | 545 | 548 | 2 | 0.45 |
firefox | 1 | Overall | 100 | 912 | 428 | -484 | -53.06 | faster
firefox | 1 | Page Request | 100 | 487 | 1 | -486 | -99.77 | faster
firefox | 1 | Rendering | 100 | 425 | 427 | 2 | 0.51 |
```
---
[1] In the case where blend modes *are* found, it becomes a lot more difficult to know if it's generally safe to skip /Resources-entries. Hence we don't cache anything in that case, however note that most document/pages do not utilize blend modes anyway.
Given the number of parameters, and the fact that many of them are booleans, the call-sites are no longer particularly easy to read and understand. Furthermore, this slightly improves the formatting of the JSDoc-comment, since it needed updating as part of these changes anyway.
Finally, this removes an unnecessary `numViews === 0` check from `getVisibleElements`, since that should be *very* rare and more importantly that the `binarySearchFirstItem` function already has a fast-path for that particular case.
* it's faster to generate the color code in using a table for components
* it's very likely a way faster to parse (when setting the color in the canvas)
*Please note:* I cannot be sure if this actually helps, since we've not enabled LGTM for the PDF.js repository[1], however I hope that it should stop LGTM reporting things that we're already using ESLint to enforce.
---
[1] In case the patch does nothing, we can simply revert it and move on :-)
This patch addresses a review comment, which pointed out that we should *also* handle the pageNumber-input, from PR 12493.
Given that a user *manually* changing pages using the pageNumber-input, on the toolbar, could be regarded as a pretty strong indication of user-intent w.r.t. navigation in the document, hence I suppose that updating the browser history in this case as well probably won't hurt.
All of these methods will, in one way or another, cause e.g. scrolling or zooming to occur and consequently they don't really make sense unless there's an active PDF document. Especially since all of these methods end up calling into a `BaseViewer`-instance, which already contains similar early returns in essentially all of it's methods and setters.
This fixes only those warnings, as reported by https://lgtm.com/projects/g/mozilla/pdf.js?mode=list, that make sense (as far as I'm concerned).
Hence this patch leaves the following things unaddressed:
- The "recommendation"-category, since it only complains about unused variables. However, note that all of those cases are purposely included and that there's thus ESLint-disable comments added to explictly allow them.
- The "warning"-category, which still contains two complaints. However, as far as I can tell, they are both false positives.
Given first of all the false positives of the LGTM static analyzer, and secondly that we'd need to add (essentially duplicated) disable-comments for the unused variable cases, it's not entirely clear to me if we actually want to work towards including LGTM in the PDF.js project (e.g. running alongside Travis) or if we should just close issue 11965.
Given that the `Map`-pattern apparently has undesirable performance characteristics, change this getter back to using an Object instead and check its size before returning it.
Rather than returning an *empty* Object[1] we should be returning `null` instead, since that's consistent with existing API-functionality.
To avoid having to *manually* track if the Object is empty, this patch also introduces a small helper function to check its size.
As can be seen in `src/core/fonts.js`, this method only accepts *one* parameter, hence it's somewhat difficult to understand what the Annotation-code is actually attempting to do here.
The only possible explanation that I can imagine, is that the intention was initially to call `Font.charToGlyph` *directly* instead. However, note that that'd would not actually have been correct, since that'd ignore one level of font-caching (see `this.charsCache`). Hence the unused arguments are removed, in `src/core/annotation.js`, and the `Font.charToGlyph` method is now marked as "private" as intended.
This patch removes unnecessary escape-sequence in (mostly) strings, as a first step, since the ones in regular expressions probably requires more careful testing (just in case).
The only exception is a regular expression in `src/core/annotation.js`, since we should have both unit- and reference-tests for this code *and* given [this information on MDN](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Guide/Regular_Expressions/Character_Classes#Types):
> Inside a character set, the dot loses its special meaning and matches a literal dot.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/no-useless-escape
Given that `Object.fromEntries` doesn't seem to *guarantee* that a `null` prototype is used, we thus hack around that by using `Object.assign` with `Object.create(null)`.
This provides a work-around to avoid having to conditionally try to initialize the `openAction`-object in multiple places.
Given that `Object.fromEntries` doesn't seem to *guarantee* that a `null` prototype is used, we thus hack around that by using `Object.assign` with `Object.create(null)`.
This probably ought to have been included in PR 12534, but better late than never I suppose, since it helps to more clearly demonstrate the bug in a way that a reference-test alone just cannot do.
When writing this unit-test I also noticed that it required a certain amount of "luck" to actually trigger the bug, prior to the patch, since it seems that the bug only reproduced for certain *unfortunate* sequences of TypedArray data. (The added unit-test contains one such, purposely simple, example.)
Given that we're now accessing certain API-functionality *directly* in this file, e.g. the AnnotationStorage and Optional Content configuration, ensuring that there's not a version mismatch definitely seem like a good idea to prevent any *subtle* future bugs.
The unit-test files themselves shouldn't be loaded until Jasmine has been setup/configured, however that doesn't matter for the "normal" PDF.js library files. Hence we can simply `import` them in the standard way.
This patch first of all enables linting of the files in the `test/font/` folder, and secondly it also re-factors all test files to use native `import`/`export` statements. Finally, all tests are now loaded correctly, rather than being included as scripts through the `font_test.html` file.
Different fonts incorrectly end up with *identical* hashes, despite having different /ToUnicode data.
The issue, and it's very interesting that we've apparently not seen it before, appears to be caused by the fact that different /ToUnicode entries share the *same* underlying `ArrayBuffer`, which thus becomes problematic at the `const dataUint32 = new Uint32Array(data.buffer, 0, blockCounts);` line. The simplest solution thus seem to be to just *copy* the input, when it's an `ArrayBuffer`, rather than using it as-is. (Note that if we'd stringified the input, when calling `MurmurHash3_64.update`, the issue would also have been fixed. In this case, we're already creating an unique TypedArray.)
Given that this code is used on the worker-thread, where SystemJS is still used during development, we need to (for now) handle this folder the same way as the `src/core/` one.
Since we no longer use SystemJS to load the unit-tests, there's now nothing that prevents us from using optional chaining and nullish coalescing in the `src/display/` directory.
*Please note:* Once https://bugzilla.mozilla.org/show_bug.cgi?id=1247687 is implemented, and we've removed SystemJS completely, this entire patch can (and even should) be reverted.
This is similar to the existing `getLookupTableFactory` helper function, but is implemented as outlined in issue 6774.
The re-formatting of the tables were done automatically, by using find-and-replace with regular expressions.
For reasons that I don't even pretend to understand, using this particular structure for these *very* long lookup tables allow SystemJS to process the files correctly/quickly and the development viewer thus works as intended.
While the *built* `pdf.worker.js` file still works correctly with these changes, despite these two files being excluded by Babel[1], the development viewer does not because of issues with SystemJS[2] and/or its Babel-plugin (both of which are old).
Furthermore, note also that excluding these two files from Babel-processing isn't *generally* necessary since e.g. the `gulp mozcentral` command works anyway. The explanation is rather that it's actually the source-map generation which fails for these huge sequences when building the `pdf.worker.js` file.
However, not using standard `import`/`export` statements in all files means we also need to use SystemJS when e.e. running the unit-tests. This is very unfortunate, since SystemJS (or its old Babel-version) doesn't support modern ECMAScript features such as e.g. optional chaining and nullish coalescing.
Unfortunately it also seems that https://bugzilla.mozilla.org/show_bug.cgi?id=1247687, which tracks the implementation of worker-modules in Firefox, has stalled since there hasn't been any updates for six months now.
To hopefully address all of the above, this patch is the first in a series that attempts to further reduce our reliance on SystemJS.
---
[1] The only difference being how the dependencies are handled, in the Webpack-bundled file.
[2] Parsing takes way too long and consumes too much memory, thus rendering the development viewer essentially unusable.
This brings the new `pdf.scripting.js` bundling more in-line with the pre-existing handling for the `pdf.js`/`pdf.worker.js` files:
- Add a new `src/pdf.scripting.js` file as the entry-point for the build scripts.
- Add the version/build numbers at the top of the *built* `pdf.scripting.js` files, since all other built files include that information given that it's often helpful to be able to easily determine the *exact* version.
- Tweak the `createScriptingBundle` in the gulp-file, since it looks like a little bit too much copy-and-paste in the variable names.
This mainly involves the `crypto_spec.js` file which declared most
variables before their usage, which is not really consistent with the
rest of the codebase. This also required reformatting some long arrays
in that file because otherwise we would exceed the 80 character line
limit. Overall, this makes the code more readable.
- Handle the arguments correctly in `PartialEvaluator.handleColorN`.
For TilingPatterns with a base-ColorSpace, we're currently using the `args` when computing the color. However, as can be seen we're passing the Array as-is to the `ColorSpace.getRgb` method, which means that the `Name` is included as well.[1]
Thankfully this hasn't, as far as I know, caused any actual bugs, but that may be more luck than anything else given how the `ColorSpace` code is implemented. This can be easily fixed though, simply by popping the `Name`-object off of the `args` Array.
- Cache TilingPatterns using the `Name`-string, rather than the object directly.
This is not only consistent with other caches in `PartialEvaluator`, but importantly it also ensures that the cache lookup always works correctly. Note that since `Name`-objects, similar to other primitives, uses a cache themselves a *manually* triggered `cleanup`-call could thus (theoretically) cause the `LocalTilingPatternCache` to not find an existing entry. While the likelihood of this happening is *extremely* small, it's still something that we should fix.
---
[1] The `args` Array can e.g. look like this: `[0.043, 0.09, 0.188, 0.004, /P1]`, which means that we're passing in the `Name`-object to the `ColorSpace` method.
Ensure that these tooltip-only Annotations are handled as "internalLink"s, to ensure that they behave as expected in PresentationMode (e.g. they should still use a `pointer`-cursor).
Ensure that `PDFLinkService.getDestinationHash` won't create links with empty hashes, since those don't really make a lot of sense in general (this improves things for tooltip-only Annotations).
This PDF file can be used for testing: http://mirrors.ctan.org/macros/latex/contrib/pdfcomment/doc/pdfcomment.pdf#page=14
- Return early in `PDFViewerApplication._initializeJavaScript` for PDF documents without any `fieldObjects`, which is the vast majority of all documents, to prevent errors when trying to parse a non-existent object.
- Similar to the other `PDFViewerApplication._initialize*` methods, ignore the `fieldObjects` if the document was closed before the data resolved.
- Fix the JSDoc comment for the `generateRandomStringForSandbox` helper function, since there's currently a bit too much copy-and-paste going on :-)
- Change `FirefoxScripting` to a class with static methods, which is consistent with the surrounding code in `web/firefoxcom.js`.
For an invalid Annotation, there's one code-path where `undefined` is returned from `AnnotationFactory._create`. That'd currently, incorrectly, trigger an error during the `PDFDocument._collectFieldObjects` parsing which thus seem good to avoid.
Along these lines, the filtering in `PDFDocument.fieldObjects` is also updated to handle both `null` and `undefined` the same way.
Some pdf softwares don't remove highlight annotations but make the QuadPoints array empty.
And the Rect for the annotation can be [-32768, -32768, 32768, 32768] so it leads to have a giant div which catches all the mouse events and make the pdf unusable when there are some forms elements.
There's no compelling reason to update this property *manually* in multiple places, since that's error-prone with any future code changes, given that `_updateInternalState` is always called just before anyway.
While the referenced issue could very well be seen as an edge-case, this patch adds support for updating of the browser history when interacting with the thumbnails in the sidebar (assuming we want to do this).
The main reason for adding the history implementation in the first place, was to simplify navigating back to a previous position in the document when named/explicit destinations are used (e.g. when clicking on "links" or when using the outline in the sidebar).
As such, it never really crossed by mind to update the browser history when the thumbnails are used. However, a user clicking on thumbnails could be regarded as a pretty strong indication of user-intent w.r.t. navigation in the document, hence I suppose that updating the browser history in this particular case probably won't hurt.
This modernizes and improves the code, by using `async`/`await` and by extracting the helper function to its own method.
To hopefully avoid confusion, given the next patch, the method is also re-named to `goToDestination` to make is slightly clearer what it actually does.
Given that *all* fonts are, ever since PR 7347, now cached in the "normal" `fontCache` there's actually no reason for the special `font.translated` construction. (Given how Objects in JavaScript are references, rather than raw values, the old code shouldn't have caused any significant memory overhead.)
Instead we can simply store the `cacheKey`, which is a simple string, on only the Font `Dict`s where it's needed and thus look-up all fonts using the `fontCache` instead.
If this method is ever passed invalid/unexpected data, or if during the course of parsing (since it's used recursively) such data is found, it will fail in a non-graceful way.
Hence this patch, which ensures that we don't attempt to access non-existent properties and also that errors such as the one fixed in PR 12479 wouldn't have occured.
*This patch is based on a couple of smaller things that I noticed when working on PR 12479.*
- Don't store the /Fields on the `formInfo` getter, since that feels like overloading it with unintended (and too complex) data, and utilize a `hasFields` boolean instead.
This functionality was originally added in PR 12271, to help determine what kind of form data a PDF document contains, and I think that we should ensure that the return value of `formInfo` only consists of "simple" data.
With these changes the `fieldObjects` getter instead has to look-up the /Fields manually, however that shouldn't be a problem since the access is guarded by a `formInfo.hasFields` check which ensures that the data both exists and is valid. Furthermore, most documents doesn't even have any /AcroForm data anyway.
- Determine the `hasFields` property *first*, to ensure that it's always correct even if there's errors when checking e.g. the /XFA or /SigFlags entires, since the `fieldObjects` getter depends on it.
- Simplify a loop in `fieldObjects`, since the object being accessed is a `Map` and those have built-in iteration support.
- Use a higher logging level for errors in the `formInfo` getter, and include the actual error message, since that'd have helped with fixing PR 12479 a lot quicker.
- Update the JSDoc comment in `src/display/api.js` to list the return values correctly, and also slightly extend/improve the description.
The last unit-test didn't work correctly, since an error was thrown in `PDFDocument._hasOnlyDocumentSignatures` because the mocked `XRef`-instance wasn't actually being set correctly.
Also, updates the `XRefMock` to use `async` methods where appropriate.
This allows us to make a slight simplification in `PartialEvaluator.loadFont`, which thus removes an old TODO-comment from the method.
Furthermore, in `PartialEvaluator.translateFont`, the CMap-handling is now limited to only *composite* fonts to avoid having to wait for a "dummy"-Promise for most fonts.
- Actually register/unregister the `WorkerTask`s, used when saving each page, correctly.
To prevent issues when terminating the Worker, we purposely wait for all running `WorkerTask`s to complete first. Hence we need to actually handle `WorkerTask`s the same way in "SaveDocument" as in the rest of this file, see e.g. "GetOperatorList" and "GetTextContent".
- Access `PDFDocument` properties in a generally safe/consistent way.
While the current code works fine, given how the PDF document is being loaded, it still seems like a very good idea to be *consistent* in how we access these kind of properties (since in general you need to avoid `MissingDataException` everywhere in this file).
- Change a variable name, since there's essentially no precedent in the code-base for *local* variable names to start with an underscore.
Support for the `scope` parameter, in `MessageHandler.on`, was removed in PR 11110 however this particular case was unused/unnecessary for years prior to that change. (From a quick look through the history, I'm not even sure if it was actually needed in the first place.)
Given that we're no longer using SystemJS to load the `web/` files, see PR 11919, there's nothing that prevents us from using standard `ìmport` statements in this file.
Obviously it's still necessary to load part of the code conditionally on the build type, however this still allows us to clean-up and simplify at least some of this file.
The only noticeable changes are that the built files are now *slightly* smaller, and that Webpack now supports optional chaining and nullish coalescing without the need for Babel plugins.
In practice it's not uncommon for PDF documents to re-use the same TilingPatterns more than once, and parsing them is essentially equal to parsing of a (small) page since a `getOperatorList` call is required.
By caching the internal TilingPattern representation we can thus avoid having to re-parse the same data over and over, and there's also *less* asynchronous parsing required for repeated TilingPatterns.
Initially I had intended to include (standard) benchmark results with this patch, however it's not entirely clear that this is actually necessary here given the preliminary results.
When testing this manually in the development viewer, using `pdfBug=Stats`, the following (approximate) reduction in rendering times were observed when comparing `master` against this patch:
- http://pubs.usgs.gov/sim/3067/pdf/sim3067sheet-2.pdf (from issue 2765): `6800 ms` -> `4100 ms`.
- https://github.com/mozilla/pdf.js/files/1046131/stepped.pdf (from issue 8473): `54000 ms` -> `13000 ms`
- https://github.com/mozilla/pdf.js/files/1046130/proof.pdf (from issue 8473): `5900 ms` -> `2500 ms`
As always, whenever you're dealing with documents which are "slow", there's usually a certain level of subjectivity involved with regards to what's deemed acceptable performance.
Hence it's not clear to me that we want to regard any of the referenced issues as fixed, however the improvements are significant enough to warrant caching of TilingPatterns in my opinion.
I've run `gulp mozcentral`, `gulp generic`, and `gulp generic-es5` with `master` respectively this patch and then diffed the build output. With the (obvious) exception of increased version/build numbers, there were no actual changes from the updated Acorn version.
The `debugger`-statement would only, potentially, make sense during development and we thus want to prevent it from being accidentally included when landing code.
The `alert`, `confirm`, and `prompt` functions should generally be avoided, with the few intended cases manually allowed.
Please find additional details about the ESLint rules at:
- https://eslint.org/docs/rules/no-debugger
- https://eslint.org/docs/rules/no-alert
The changelogs of those dependencies showed no breaking changes for us.
Most of the time the major version bump was done to remove compatibility
with very outdated Node.js versions.
Only for `autoprefixer` and `gulp-postcss` a change was required, which
is including `postcss` in our `package.json` explicitly since it's now
a peer dependency of those packages.
Now only `acorn`,`systemjs`, `terser` and `yargs` are not the latest
versions because they require more work.
In the rest of the viewer code-base, we purposely don't treat `RenderingCancelledException`s as actual errors (since they aren't) and consequently we never log them.
Hence it makes sense, as far as I'm concerned, to simply treat `RenderingCancelledException`s the same way when printing in Firefox.
While I don't print a whole lot, I cannot remember seeing these "errors" logged when printing until *very* recently[1]. Given that the browser print functionality and UI, in Firefox, is under active development it's certainly possible that there's some recent changes to the related timings which make `RenderingCancelledException`s more likely now.
---
[1] Interestingly, only some PDF documents seem to be affected as well; I'm able to reproduce this pretty consistently by opening https://www.uni-muenster.de/imperia/md/content/ziv/pdf/printpay_flyer.pdf in Firefox and then repeating the following sequence:
Clicking on the PDF.js print button, and then cancelling printing.
This dependency hasn't been updated in two years and the only place that
uses it is the `externaltest` target in the Gulpfile. We can simply
replace `fancy-log` usage there with `console.log` like we do in all
other places in the Gulpfile because we're not interested in the
timestamps here. Gulp already prints timestamps and these tests finish
within a second anyway.
Note that it remains in `package-lock.json` because other Gulp-related
packages have it as a dependency, but at least we're no longer depending
on it directly anymore now.
This simplifies/consolidates the ESLint configuration slightly in the `src/` folder, and prevents the addition of any new files where `var` is being used.[1]
Hence we no longer need to manually add `/* eslint no-var: error */` in files, which is easy to forget, and can instead disable the rule in the `src/core/` files where `var` is still in use.
---
[1] Obviously the `no-var` rule can, in the same way as every other rule, be disabled on a case-by-case basis where actually necessary.
This should be helpful to easily determine the *exact* version of the viewer itself, when looking at a *built* `web/viewer.js` file.
Note that we're already including this information in other built files, such as e.g. `pdf.js`, `pdf.worker.js`, `pdf_viewer.js`, and `pdf.image_decoders.js`.
Previously this rule has been enabled in the `web/` folder, and in select files in the `src/` sub-folders.
In this case, enabling of this rule didn't actually require any further code changes.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/no-var
Previously this rule has been enabled in the `web/` folder, and in select files in the `src/` sub-folders.
Note that a number of the files in the `src/display/` folder were already enforcing the `no-var` rule, and thanks to Prettier the necessary re-writing will be (mostly) handled automatically.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/no-var
This adds a new `PDFViewerApplication.triggerPrinting` method, which takes care of checking that printing is actually supported before calling `window.print`, to remove the need to duplicate that code in multiple places.
Also, removes the `PDFViewerApplication.printing` getter since it's not really necessary any more.
For now we need to use a Babel-plugin, since part of our build system doesn't support this fully (e.g. Babel-loader, Webpack 4.x, and SystemJS).
While the `?.` operator will thus always be transpiled by Babel, even in modern builds, simply supporting it for development purposes seems like a step in the right direction.
https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Operators/Optional_chaining
This release contains support for additional modern ECMAScript features, such as e.g. the nullish coalescing operator `??` and the optional chaining operator `?.`.
Given how those are used, there *should* not be any situation in which e.g. `undefined` is ever returned. However, actually checking that the pageView/thumbnail is defined cannot hurt.
Also, re-factor `webViewerPageRendered` slightly since the `pageView` is no longer unconditionally necessary after the previous patches; note in particular that the thumbnails will only be updated when the sidebar *and* the thumbnailView is visible.
Finally, fixes a bug in `webViewerPageChanging` whereby an empty string would not be treated as a valid pageLabel and instead be replaced by `null`.
Given that the default viewer only uses the "page stats" when debugging is enabled, it seems much simpler and more straightforward to simply query the API *directly* when this information is actually required. That way, there's a bit less information that needs to be stored/updated on each `PDFPageView`-instance.
Finally, since the `EventBus` now exists, we no longer need to handle the "page stats"-case in the regular listeners in `web/app.js`, but can instead add special "page stats"-listeners only when debugging is enabled.
The way that rendering errors are handled in `PDFPageView` is *very* old, and predates e.g. the introduction of the `EventBus` by several years.
Hence we should be able to simplify things a bit here, by including the Error (when it exists) in the "pagerendered" event and thus avoid having to reach into `PDFPageView` for it.
Note that a `RenderingCancelledException` *should* never actually reach this method, but better safe than sorry I suppose, considering that both `PDFPageView` and `PDFThumbnailView` are already catching `RenderingCancelledException`s since those are *not* Errors in the normal sense of the word.
This changes the `transformOrigin` calculations in `AnnotationElement._createContainer` and `PopupAnnotationElement.render`, to ensure that e.g. the clickable area of annotations and/or popups are both positioned correctly.
The problem occurs for *negative* values, since they're not negated correctly because of how the `transformOrigin` strings were build; see issue 12406 for a more in-depth explanation. Previously, for negative values, the `transformOrigin` strings would thus be ignored since they're not valid.
Given that the long-standing `webpack-stream` issue 201 was recently fixed in PR 207, and a new version released, we should now finally be able to update the dependency.
However, depending on if/when `webpack-stream` gets support for Webpack 5 (which is currently in beta) we may still want remove our `webpack-stream` dependency.
- Check that the "Info"-entry, in the XRef-trailer, is actually a dictionary before accessing it. This is similar to the `PDFDocument.documentInfo` method and follows the general principal of validating data carefully before accessing it, given how often PDF-software may create corrupt PDF files.
- Slightly simplify the "XFA"-lookup, since there's no point in trying to fetch something from the empty dictionary.
Currently there's nothing that prevents modification of the `Dict.empty` primitive, which obviously needs to be *truly* empty to prevent any future (hard to find) bugs.
This patch contains a possible approach for fixing issue 12294, which compared to other PRs is purposely limited to the affected `WidgetAnnotation` code.
As mentioned elsewhere, considering that we're (at least for now) trying to fix *one specific* case, I think that we should avoid modifying the `Dict` primitive[1] and/or avoid a solution that (indirectly) modifies an existing `Dict`-instance[2].
This patch simply fixes the issue at hand, since that seems easiest for now, and I'd suggest that we worry about a more general approach if/when that actually becomes necessary.
Hence the solution implemented here, for `WidgetAnnotation`, is to simply use a combination of the local *and* AcroForm /DR resources during OperatorList-parsing to ensure that things work correctly regardless of where a particular /Font resource is found.
For saving of form-data, on the other hand, we want to avoid increasing the file-size unnecessarily and need to be smarter than just merging all of the available resources. To achive this, a new `WidgetAnnotation._getSaveFieldResources` method will when necessary produce a combined resources `Dict` with only the minimum amount of data from the AcroForm /DR resources included.
---
[1] You want to avoid anything that could cause the general `Dict` implementation to become slower, or more complex, just for handling an edge-case in my opinion.
[2] If an existing `Dict`-instance is modified unexpectedly, that could very easily lead to problems elsewhere since e.g. `Dict`-instances created during parsing are not expected to be changed.
This will allow `makeref` to run "successfully" on the bots, since in the current state testing/makeref is just overall broken.
Obviously we still need to figure what's causing the intermittent failures, and fix them, but let's at least unblock things for now; see issue 12371.
We were correctly finishing the SMask group but not restoring all the extra
transformations applied in stateStack, so if somebody ends up drawing to the
same context after canceling mid-draw we'd get artifacts.
This re-lands #12363 and fixes Mozilla bug 1664178[1].
[1]: https://bugzilla.mozilla.org/show_bug.cgi?id=1664178
This fixes the issue that caused #12363 to get reverted, see #12367.
When we end the SMask group and stateStack.length is zero, nothing updates
this.current to reflect it.
We were correctly finishing the SMask group but not restoring all the extra
transformations applied in stateStack, so if somebody ends up drawing to the
same context after canceling mid-draw we'd get artifacts.
This fixes Mozilla bug 1664178[1].
[1]: https://bugzilla.mozilla.org/show_bug.cgi?id=1664178
For these streams, compared to `Stream` and `ChunkedStream`, there's no well defined concept of length and consequently no `length` getter.[1] However, attempting to access the non-existent `length` won't currently error, but just return `undefined`, which could thus easily lead to bugs elsewhere in the code-base.
---
[1] However, note that *all* stream implementations have an `isEmpty` getter which can be used instead.
For years the loadingBar and sidebarContainer has had a slightly annoying and unfortunate dependency, since the loadingBar width follows the main toolbar width[1].
To prevent the loadingBar from obscuring part of the sidebarContainer, especially the buttons, the sidebarContainer is thus moved down when the loadingBar is visible. This has always annoyed me[2], since it means that the buttons in the sidebar may thus move vertically which seems bad from a UX perspective.
Now that CSS variables are available in all supported browsers[3] however, fixing the loadingBar/sidebarContainer overlap issues are finally easy. The solution is simply to let the sidebarContainer, when visible, control the loadingBar left position (right in RTL locales) in the same way that the viewerContainer is handled. Hence the sidebarContainer can now have a *consistent* vertical postition, without the loadingBar overlapping it.
---
[1] Obviously the right position (left in RTL locales) of the loadingBar is, potentially, reduced to account for a scrollbar.
[2] I've tried to fix this a few times, but it always seemed like more trouble than it's worth.
[3] https://developer.mozilla.org/en-US/docs/Web/CSS/Using_CSS_custom_properties#Browser_compatibility
The `getVisibleElements` helper function currently requires the viewerContainer to be absolutely positioned; possibly fixing this is tracked in issue 11626.
Without `position: absolute;` set, in the CSS, there's a number of things that won't work correctly such as e.g.
- Determining which pages are currently visible, thus forcing all of them to render on load and increasing resource usage significantly; note https://github.com/mozilla/pdf.js/wiki/Frequently-Asked-Questions#allthepages
- Scrolling pages into view, by using the `BaseViewer.currentPageNumber` setter or similar.
Based on the number of opened issues over the years, the fact that `position: absolute;` is required has shown to be something that users can very easily overlook unless they follow e.g. the `simpleviewer` example to the letter.
Hence, to improve things until such a time that issue 11626 is fixed, we'll now refuse to initialize a `BaseViewer` instance unless the `container` has the required CSS set. (Forcibly setting `position: absolute;` on the viewerContainer element is bound to cause significantly more issues/confusion, hence the current approach of throwing an Error.)
This reverts commit 9e4552d7920fcba13b3d74e1e180fc49180f4436 for causing the sidebar to become too narrow when the entire viewer is resized.
**Steps to reproduce:**
1. Load the viewer.
2. Open the sidebar.
3. Resize the sidebar, making it wider.
4. Resize the entire viewer, i.e. the browser window, making it *narrower* than 400 pixels.
**Expected result:**
The sidebar width is clamped at 200 pixels.
**Actual result:**
The sidebar becomes too narrow.
The cause of this bug is, in hindsight, quite obvious since the `clamp` helper function implicitly assumes that the `min`/`max` arguments are correctly sorted. At viewer widths *below* 400 pixels, that assumption is broken which explains the bug.
* Move display/xml_parser.js in shared to use it in worker
* Save form data in XFA datasets when pdf is a mix of acroforms and xfa
Co-authored-by: Brendan Dahl <brendan.dahl@gmail.com>
Given that the outlineView/attachmentsView/layersView all share a common base-class and CSS rules, see PRs 12169 and 12170, the names of the CSS variables in question feels slightly strange now.
This patch purposely starts small, by removing IE-specific code from various JS/CSS files in the `web/` folder.
There's obviously lots of potential for additional clean-up, especially the removal of no longer necessary polyfills in `src/shared/compatibility.js`, however that will require some care considering that certain polyfills may also be necessary for e.g. Node.js or the Chromium-extension as well.
Generally speaking, once we start removing polyfills it's probably a good idea to consult the compatibility information on https://developer.mozilla.org/ and also https://caniuse.com/ first. (Deciding on the lowest supported Chromium version, for the extension, would also seem like a good idea.)
This reverts commit 2a0de0b66b41c345cd357cbb828b3ca6f49fd395.
I can no longer reproduce these issues locally, and if ad blockers are still interfering with this functionality we really ought to pursue a mozilla-central solution to the problem instead. (Also, I'm no longer getting an "Open with Firefox"-option in the "Open with"-dialog making the PDF attachments experience worse for all users.)
This should help prevent future issues, caused by the user omitting the `viewer` option and/or providing an incorrect `container` option, when initializing a `BaseViewer`-instance.
This fixes a set of issues described in Mozilla bug 1662426[1].
In particular, once the print callback fails once (because the printing
operation has been canceled in Gecko / replaced by a newer one, for example) it
can't be re-invoked.
This patch fixes it by properly cancelling the render task if it throws, or if
the print callback is called again while ongoing.
[1]: https://bugzilla.mozilla.org/show_bug.cgi?id=1662426
All of the core/display functionality needed to support this already exists, we simply need to handle these named actions in the viewer and the buttons will "just" work.
Unfortunately there's not really any good way of testing this, but given the size and scope of the patch that's hopefully OK.
This allows for merging of dictionaries one level deeper than previously. This could be useful e.g. for /Resources dictionaries, where you want to e.g. merge their respective /Font dictionaries (and other) together rather than picking just the first one.
This is *similar* to the existing linting for JavaScript files, but covers CSS files instead.
While there's a lot of rules that could potentially be used, the main advantage of using Stylelint is that it has Prettier integration which means that we can automatically enforce a *consistent* style for our CSS files as well.
As a proof of concept, this patch is purposely limited to:
- Adding a simple rule, here `block-no-empty` is chosen; see https://stylelint.io/user-guide/rules/block-no-empty
- Adding Prettier integration, to unify the style of our CSS files.
Please find additional information at https://stylelint.io/
*Besides, obviously, adding viewer support:* This patch attempts to improve the general API for Optional Content Groups slightly, by adding a couple of new methods for interacting with the (more complex) data structures of `OptionalContentConfig`-instances. (Thus allowing us to mark some of the data as "private", given that it probably shouldn't be manipulated directly.)
By utilizing not just the "raw" Optional Content Groups, but the data from the `/Order` array when available, we can thus display the Layers in a proper tree-structure with collapsible headings for PDF documents that utilizes that feature.
Note that it's possible to reset all Optional Content Groups to their default visibility state, simply by double-clicking on the Layers-button in the sidebar.
(Currently that's indicated in the Layers-button tooltip, which is obviously easy to overlook, however it's probably the best we can do for now without adding more buttons, or even a dropdown-toolbar, to the sidebar.)
Also, the current Layers-button icons are a little rough around the edges, quite literally, but given that the viewer will soon have its UI modernized anyway they hopefully suffice in the meantime.
To give users *full* control of the visibility of the various Optional Content Groups, even those which according to the `/Order` array should not (by default) be toggleable in the UI, this patch will place those under a *custom* heading which:
- Is collapsed by default, and placed at the bottom of the Layers-tree, to be a bit less obtrusive.
- Uses a slightly different formatting, compared to the "regular" headings.
- Is localizable.
Finally, note that the thumbnails are *purposely* always rendered with all Optional Content Groups at their default visibility state, since that seems the most useful and it's also consistent with other viewers.
To ensure that this works as intended, we'll thus disable the `PDFThumbnailView.setImage` functionality when the Optional Content Groups have been changed in the viewer. (This obviously means that we'll re-render thumbnails instead of using the rendered pages. However, this situation ought to be rare enough for this to not really be a problem.)
The `/Order` array is used to improve the display of Optional Content groups in PDF viewers, and it allows a PDF document to e.g. specify that Optional Content groups should be displayed as a (collapsable) tree-structure rather than as just a list.
Note that not all available Optional Content groups must be present in the `/Order` array, and PDF viewers will often (by default) hide those toggles in the UI.
To allow us to improve the UX around toggling of Optional Content groups, in the default viewer, these hidden-by-default groups are thus appended to the parsed `/Order` array under a *custom* nesting level (with `name == null`).
Finally, the patch also slightly tweaks an `OptionalContentConfig` related JSDoc-comment in the API.
This patch:
- Removes the :hover effect from the `findMsg` element, since it's a simple span and clicking it *obviously* does nothing.
- Given the way that the checkboxes are visually hidden, with `opacity: 0;` and absolute positioning, they are unfortunately still focusable (fixed by adding `pointer-events: none;`). To reproduce this, in `master`: Place the mouse pointer over the upper left-hand corner of the "Highlight all"-option, and notice that the :hover effect vanishes and clicking toggles the "Match case"-option instead.
This special progressBar is only used in the (fortunately) rare case when a server doesn't provide a valid `Content-Length` header. Since this progressBar isn't normally seen, when testing the default viewer, it's certainly very easy to see why these CSS rules were missed during review.
Furthermore, this patch also makes a couple of *small* progressBar CSS tweaks not related to the colours.
With the changes in PR 11077, these panels are no longer aligned exactly with the *center* of the corresponding toolbar buttons. This is especially noticeable for the `findbar` at narrow viewer width.
In issue 12120, the font has a 1,0 cmap and is marked symbolic which
according to the spec means we should directly use the cmap instead of
the extra steps that are defined in 9.6.6.4.
However, just fixing that caused bug 1057544 to break. The font in bug
1057544 has a 0,1 cmap (Unicode 1.1) which we were not using, but is
easy to support. We're also easily able to support some of the other
unicode cmaps, so I added those as well.
There was also a second issue with bug 1057544, the cmap doesn't have
a mapping for the "quoteright" glyph, but it is defined in the post
table. To handle this, I've moved post table as a fallback for any
font that has an encoding.
Unfortunately the work-around implemented in PR 12286 didn't actually work in all cases, please refer to the previous commit messages.
To prevent opening of PDF attachments from being completely broken for some users, we'll simply force-download them for now in MOZCENTRAL-builds to unbreak things. (Given that the "Open with" dialog now features a "Open with Firefox"-option, this is less bad than it previously would've been.)
This should provide better filetype detection when downloading PDF attachments in the viewer.
Also, to avoid creating the "is PDF file" regular expression more than once it's extracted into a global constant instead.
This reverts commit 1e5d4b6a8025477c3b15b3a320b1cc41ccd7c2bc, since it unfortunately doesn't work in all situations.
Please note that I did *successfully* test the patch in a local Firefox build, obviously with an ad blocker installed.
However, I've now tested the *latest* Nightly-build with my default profile, and unfortunately I can still reproduce the bug there!?
Unfortunately e.g. ad blockers can interfere with `window.open` calls, thus preventing PDF attachments from being opened/viewed. For the MOZCENTRAL-build, we can work-around this problem by using a (hidden) link instead.
Without these changes, clicking on the "Open With Different Viewer"-button on the Firefox fallback bar won't actually do anything and the following is printed in the web-console:
```
Uncaught TypeError: (destructured parameter) is undefined
download resource://pdf.js/web/viewer.js:956
response resource://pdf.js/web/viewer.js:1054
listener resource://pdf.js/web/viewer.js:11891
viewer.js:956:1
```
Furthermore, this patch also fixes `PDFViewerApplication.fallback` to pass in an explicit `sourceEventType` when triggering downloading. While this, on its own, would obviously have been sufficient to fix the bug described above, it seems wrong to outright break backwards compatibility of any older `PDFViewerApplication.download` calls.
Now that the `parse` method is simplified we can inline the `setup`
method in the `parse` method since it's only two lines of code. This
avoids some indirection.
Good form type detection is important to get reliable telemetry and to
only show the fallback bar if a form cannot be filled out by the user.
PDF.js only supports AcroForm data, so XFA data is explicitly unsupported
(tracked in issue #2373). However, the previous form type detection
couldn't separate AcroForm and XFA well enough, causing form type
telemetry to be incorrect sometimes and the fallback bar to be shown for
forms that could in fact be filled out by the user.
The solution in this commit is found by studying the specification and
the form documents that are available to us. In a nutshell the rules are:
- There is XFA data if the `XFA` entry is a non-empty array or stream.
- There is AcroForm data if the `Fields` entry is a non-empty array and
it doesn't consist of only document signatures.
The document signatures part was not handled in the old code, causing a
document with only XFA data to also be marked as having AcroForm data.
Moreover, the old code didn't check all the data types.
Now that AcroForm and XFA can be distinguished, the viewer is configured
to only show the fallback bar for documents that only have XFA data. If
a document also has AcroForm data, the viewer can use that to render the
form. We have not found documents where the XFA data was necessary in
that case.
Finally, we include unit tests to ensure that all cases are covered and
move the form type detection out of the `parse` function so that it's
only executed if the document information is actually requested
(potentially making initial parsing a tiny bit faster).
Not only is `catDict` never accessed anymore outside of this file, it
should also never happen since it's internal to the catalog. If data
from it is needed elsewhere, the catalog should provide a getter for it
that can do basic data integrity checks and abstract away any
unnecessary details.
The `AcroForm` entry is part of the catalog, not of the document, so its
logic should be placed there instead. The document should look in the
catalog to fetch it, and not have knowledge of `catDict`, which is a
member internal to the catalog.
Moreover, make the AcroForm member private on the document instance. It's
only used internally and was also never intended to be public. For users
it's exposed by the `getMetadata` API endpoint as `IsAcroFormPresent`.
Only a boolean is exposed, so we now also only store the boolean on the
document instance.
Finally, the annotation code needs access to the full AcroForm
dictionary, so it's updated to fetch the data from the catalog instead
of the document that now only holds the boolean.
The `Collection` entry is part of the catalog, not of the document, so
its logic should be placed there instead. The document should look in the
catalog to fetch it, and not have knowledge of `catDict`, which is a
member internal to the catalog.
Moreover, remove the collection member from the document instance. It's
only used internally and was also never intended to be public. For users
it's exposed by the `getMetadata` API endpoint as `IsCollectionPresent`.
Moving this out of the `parse` function makes sure that the getter is
only executed if the document information is actually requested
(potentially making initial parsing a tiny bit faster).
The `Version` entry is part of the catalog, not of the document, so its
logic should be placed there instead. The document should look in the
catalog to fetch it, and not have knowledge of `catDict`, which is a
member internal to the catalog.
Moreover, make the version member private on the document instance. It's
only used internally and was also never intended to be public. For users
it's exposed by the `getMetadata` API endpoint as `PDFFormatVersion`.
Finally, clarify how the version from the header and the version from
the catalog are treated using a comment.
Even though the code obviously works as-is, given that we have unit-tests for it, it still feels incorrect to just *assume* that the `Catalog`-instance has all of its properties immediately available. Especially when (almost) all of the other handlers, in `src/core/worker.js`, protect their data accesses with appropriate `pdfManager.ensure` calls.
Even though the code obviously works as-is, given that we have unit-tests for it, it still feels incorrect to just *assume* that the `XRef`-instance has all of its properties immediately available. Especially when (almost) all of the other handlers, in `src/core/worker.js`, protect their data accesses with appropriate `pdfManager.ensure` calls.
In `display/canvas.js` the accent offsets must be multiplied by `fontSize` to make the offsets large enough. Another problem is in `core/type1_parser.js` when the Type1 command `seac` is handled. There is an error in the Adobe Type1 spec. See chapter 6 in Type1 Font Format Supplement, which provides an errata: The arguments of `seac` specify the offset of the left side bearing (LSB) points, not the offset of origins. This can be fixed in `core/type1_parser.js` by adding the difference of the LSB values.
Currently there's enough leading padding that the `numPages` span feels somewhat "disconnected" from the `pageNumber` input, which seems unfortunate when they contain related state.
In addition to the unit tests these reference tests make sure that this
document, that triggered some edge cases in our code, can be rendered
and printed successfully now.
The down appearance (`D`) is optional and not available in the document
from #12233, so the checkboxes are never saved/printed as checked
because the checked appearance is based on the export value that is
missing because the `D` entry is not available.
Instead, we should use the normal appearance (`N`) since that one is
required and therefore always available.
Finally, the /Off appearance is optional according to section 12.7.4.2.3
of the specification, so that needs to be taken into account to match
the specification and to fix reference test failures for the
`annotation-button-widget-print` test. That is a file that doesn't
specify an /Off appearance in the normal appearance dictionary.
The helper method `_decodeFormValue` is used to ensure that it happens
in one place. Note that form values are field values, display values
and export values.
Obviously it doesn't make sense to call that method without providing an `AnnotationStorage`-instance, however we should ensure that doing so won't cause errors.
Hence we need to check that `annotationStorage` is actually defined, before attempting to call its `resetModified` method.
This solution is obviously *not* perfect, since printing being cancelled will thus remove the warning as well. However, a similar problem already exists for saving, since the user may cancel that one as well.
All-in-all, since way cannot really detect with absolute certainty that either saving or printing actually finished, this seems good enough for now.
This commit makes the following improvements:
- The code is similar to the other interactive form widgets now, with a
clear note for the only difference.
- Calling `getOrCreateValue` unconditionally ensures that choice widgets
always have a value in the annotation storage. Previously we only
inserted a value in the annotation storage when an option matched or
when a selection was changed. However, this causes breakage when
saving/printing because comboboxes, which we don't fully support yet
but are rendered, might not have a value in storage at all. Their
field value might not match any option since it allows the user to
enter a custom value.
- Calling `getOrCreateValue` unconditionally ensures that forms with
choice widgets no longer always trigger a warning when the user
navigates away from the page. This fixes
https://github.com/mozilla/pdf.js/pull/12241#discussion_r474279654
Given that `renderInteractiveForms` is now enabled by default in "full" viewer, it seems reasonable to enable it by default in the viewer components as well.
Especially considering that it's simple to disable, when creating the affected components, for anyone implementing their own viewer.
The export value is used when the document is saved, so it should also
be used when the document is opened to determine which choice widget
option is selected. The display value is, as the name implies, only to
be used for viewer display purposes and not for other logic.
This makes sure that in the document from #12233 the "Favourite colour"
choice widget is correctly initialized with "Red" instead of "Black"
because the field value is equal to the export value (always the case),
but not the display value (not always the case). Moreover, saving now
also correctly uses the export value and not the display value.
Related to https://bugzilla.mozilla.org/show_bug.cgi?id=1659753
This allows Firefox trigger a "save" event from ctrl/cmd+s or the "Save
Page As" context menu, which in turn lets pdf.js generate a new PDF if
there is form data to save.
I also now use `sourceEventType` on downloads so Firefox can determine if
it should launch the "open with" dialog or "save as" dialog.
With the changes in PR 11077, the zoom dropdown now looks "squashed" in locales with longer than average zoom-strings[1]. The reason is that the zoom-value and the dropdown-icon are too close together, which doesn't look good in affected locales.
To fix this, the following changes are made:
- Increase the calculated dropdown width, in `Toolbar._adjustScaleWidth`, to account for the much wider icon (7 px -> 16 px) and the increased padding.
- Move the dropdown-icon *slightly* outwards, and also *slightly* reduce the left (right in RTL locales) padding of the dropdown-contents.
- Finally, remove the right (left in RTL locales) padding to reduce the chance of the *default* browser dropdown-icon being visible.
---
[1] This affects e.g. the `de` and `nl` locales, but there's probably other examples as well.
The specification states that the field value is `null` if no item is
selected and we didn't handle this case properly. Even though this did
not break the rendering because we always convert the value to an array
and the `includes` check in the display layer would simply not match,
the field value would be `[null]` which is not expected and strange from
an API perspective.
This commit fixes that by ensuring that we return an empty array in
case the field value is `null`. The API therefore still always gives an
array for the field value, but now the code is more specific so that the
value is either an empty array or an array of strings.
This commit follows the same pattern as another unit test in this file
and both reduces existing and future code duplication (since the next
commit will extend this test with an additional input).
This commit includes support for rendering pages in printing mode,
which, when combined with annotation storage data, is useful for testing
if form data is correctly rendered onto the printed canvas.
This is needed for some smoke tests in mozilla central for testing forms
in pdf.js.
Note: AnnotationLayerBuilder.render() doesn't really need to be async, but
we're talking of making the annotation's render functions async, so this
will make that switch easier.
Note that this will increase the run-time of `gulp dist` and `gulp dist-install`, but that's unavoidable given that there's now additional building happening.
This is *similar* to the existing transfer function support for SMasks, but extended to simple image data.
Please note that the extra amount of data now being sent to the worker-thread, for affected /ExtGState entries, is limited to *at most* 4 `Uint8Array`s each with a length of 256 elements.
Refer to https://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#G9.1658137 for additional details.
I completely overlooked the fact that `PartialEvaluator.handleSetFont` also updates the current `state`, which means that currently we're not actually handling font data correctly for cached /ExtGState data. (Thankfully, using /ExtGState to set a font is somewhat rare in practice.)
Currently using a touchscreen with pdf.js doesn't work so well. In Firefox,
with apz.allow_zooming = false (default on current release/beta), it does a
reflow zoom which makes the UI elements bigger. And with apz.allow_zooming = true
(default on current Firefox nightly), or in Chrome, it does a smooth pinch-zoom
but that also scales up the entire UI. Neither of these is a particularly good
experience, so this patch just disables any multi-touch gestures. Touch-based
panning (which involves a single touch point) is left unaffected.
I obviously missed this during review, but currently `PDFViewerApplication._saveInProgress` is reset *synchronously* in `PDFViewerApplication.save`.
That was probably not intended, since it essentially renders the `PDFViewerApplication._saveInProgress` check pointless given that the actual saving is an *asynchronous* operation.
While this will obviously increase the size of the output of `gulp minified`/`gulp minified-es5` *slightly*, the resulting files are still a lot smaller than the non-minified builds.
See https://github.com/terser/terser#minify-options for information about various Terser options.
The original code would get a long sequence of miniscule "tick" values while
pinch-zooming, and each tick value would cause a 1.1x zoom. So even the smallest
pinch gesture on a trackpad would cause high amounts of zoom. This patch
accumulates the wheel deltas until they reach an integer threshold (with a
tweak of the scaling factor to make it feel more natural) at which point it
triggers the zoom based on the integer component of the accumulated delta. The
fractional part is retained in the accumulator.
Some fonts have loca tables that aren't sorted or use 0 as an offset to
signal a missing glyph. This fixes the bad loca tables by sorting them
and then rewriting the loca table and potentially re-ordering the glyf
table to match.
Fixes#11131 and bug 1650302.
- Initialize the `AnnotationStorage`-instance, on `PDFDocumentProxy`, lazily.
- Change the `AnnotationStorage` to use a `Map` internally, rather than a regular Object (simplifies the following points).
- Let `AnnotationStorage.getAll` return `null` when there's no data stored, to avoid unnecessary parsing on the worker-thread. This ought to "just work", since the worker-thread code *should* already handle the `!annotationStorage` case everywhere.
- Add a new `AnnotationStorage.size` getter, to be able to easily tell if there's any data stored.
Prior to PR 11601, the `disableCreateObjectURL` option was present on `getDocument` in the API, since it was (potentially) used when decoding JPEG images natively in the browser. Hence setting this option, which was done automatically using compatibility-code, were in some browsers necessary in order for e.g. JPEG images to be correctly rendered.
The downside of the `disableCreateObjectURL` option is that memory usage increases significantly, since we're forced to build and use `data:` URIs (rather than `blob:` URLs).
However, at this point in time the `disableCreateObjectURL` option is only necessary for *some* (non-essential) functionality in the default viewer; in particular:
- The openfile functionality, used only when manually opening a new file in the default viewer.
- The download functionality, used when downloading either the PDF document itself or its attached files (if such exists).
- The print functionality, in the generic `PDFPrintService` implementation.
Hence neither the general PDF.js library, nor the *basic* functionality of the default viewer, depends on the `disableCreateObjectURL` option any more; which is why I'm thus proposing that we remove the option since using it is a performance footgun.
*Please note:* To not outright break currently "supported" browsers, which lack proper `URL.createObjectURL` support, this patch purposely keeps the compatibility-code to explicitly disable `URL.createObjectURL` usage *only* for browsers which are known to not work correctly.[1]
While it's certainly possible that there's additional, likely older, browsers with broken `URL.createObjectURL` support, the last time that these types of problems were reported was over *three* years ago.[2]
Hence in the *very* unlikely event that additional problems occur, as a result of these changes, we can either add a new case in the compatibility-code or simply declare the affected browser as unsupported.
---
[1] Which are IE11 (see issue 3977), and Google Chrome on iOS (see PR 8081).
[2] Given that `URL.createObjectURL` is used by default, you'd really expect more reports if these problems were widespread.
Issue 4398 was fixed by PR 4437, however a test-case wasn't included as far as I can tell. Given that PR 12186 is now in the process of re-factoring that code, adding a test-case cannot hurt as far as I'm concerned.
With recent changes, these event handlers are now essentially identical. Hence a new helper function is added, to reduce unnecessary duplication (will also be helpful with upcoming changes).
These two classes are unsurprisingly quite similar, and with upcoming changes[1] the amount of (essentially) duplicated code will increase even further.
Notable changes:
- Collect shared functionality in the `BaseTreeViewer` class, reducing both current and future code-duplication.
- Reduce unnecessary duplication in the CSS rules, which will be particularly useful with upcoming changes.
- Tweak the attachmentsView to use links, rather than buttons, to simplify (primarily) the CSS rules.
---
[1] Once API support for "Optional Content" lands, I've got more-or-less finished patches to add viewer support as well.
While the parameter name (clearly) suggests that an `AnnotationStorage`-instance is expected, looking at the only call-sites that include the parameter (i.e. the `PDFPrintServiceFactory` instances) it actually contains just a normal Object.
Hence it seems much more reasonable to actually pass a valid `AnnotationStorage`-instance, as the name suggests, and simply have `PDFPageProxy.render` do the `annotationStorage.getAll()` call. (Since we cannot send an `AnnotationStorage`-instance as-is to the worker-thread, given the "structured clone algorithm".)
Do the `AppOptions.get("printResolution")` lookup once in `web/app.js `, when initializing `PDFPrintServiceFactory`-instances, rather than for every printed page
- Fix the `gulp types` task to run on Windows. Currently this fails, and the solution was to "borrow" the same formatting as used in the `gulp jsdoc` task.
- Place the TypeScript definitions in their own `types` directory, when building `pdfjs-dist`. These should *not* be cluttering the main `build` directory, especially since the generated TypeScript definitions consists of *multiple folders*. (Only if the TypeScript definitions would be concatenated into *a single file*, would placing them directly in `pdfjs-dist/build` be acceptable.)
Over time we used multiple different formats for JSDoc comments. This
commit standardizes those formats to the one we used most often.
Moreover, this removes the example in the outline endpoint documentation
since it now has a proper type definition and it didn't render correctly
in JSDoc.
This commit:
- formats the documentation block according to the standards;
- replaces the callback definitions with the `function` type (we have
that for other definitions already and the callback type was not
rendered correctly by JSDoc);
- synchronizes the type documentation and the class documentation;
- fixes the documentation by making it easier to read and making sure
that all optional properties are indicated as such;
- uses the `@link` tag to indicate links to other code.
The `typestest` still passes and JSDoc now renders this class correctly.
This commit:
- moves the preparation work to a new `typestest-pre` target similar to
how the other targets work;
- moves the `TYPESTEST_DIR` definition to the top of the file like we
did for all other directory variables;
- renames the `TYPES_BUILD_DIR` variable to `TYPES_DIR` since it's
shorter and the naming scheme then corresponds to the other directory
variables;
- switches to `const`/template strings in the types targets where needed;
- converts the `if (err !== null)` check to `if (err)` similar to other
targets.
This patch should *hopefully* remove the intermittent unit-test failure, by using the *same* `optionalContentConfigPromise` for both `renderTask`s and thus get more predictable timing behaviour.
Add a new method to the API to get the optional content configuration. Add
a new render task param that accepts the above configuration.
For now, the optional content is not controllable by the user in
the viewer, but renders with the default configuration in the PDF.
All of the test files added exhibit different uses of optional content.
Fixes#269.
Fix test to work with optional content.
- Change the stopAtErrors test to ensure the operator list has something,
instead of asserting the exact number of operators.
Since the attachment fetching/parsing is already asynchronous, possibly delaying the dispatching of an "attachmentsloaded" event should thus not be a problem in general.
Note that in some cases, i.e. PDF documents with no "regular" attachments and only FileAttachment annotations, we'll thus no longer dispatch an "attachmentsloaded" event when `attachmentsCount === 0` and instead wait for the FileAttachment parsing to finish. (The use of a timeout still guarantees that an "attachmentsloaded" event is *eventually* dispatched though.)
This patch *considerably* simplifies the "attachmentsloaded" event handler, in `PDFSidebar`, since the details are now abstracted away from the consumer.
Finally, add a check in `_appendAttachment` to ensure (indirectly) that the FileAttachment annotation is actually relevant for the active document.
This approach is already used in other parts of the code-base, see e.g. `PDFOutlineViewer`, and has the advantage of only invalidating the DOM once rather than for every attachment item.
This is possible now that most intermittent unit test failures have been
resolved by other patches. There is only one remaining, but it's very
rare and doesn't have to block this update anymore.
This might make debugging intermittent failures a bit easier in the
future because it allows us to spot unexpected differences in the number
of tests being run and allows us to run the tests locally in the same
order in case of intermittent failures.
This reverts commit 50f73092e1f74e0fb1ea8d40763a5fba709c00c0. This
causes an inconsistency with the integrated find bar that should be
discussed more before moving on with this (refer to PR #12141).
These errors can/will occur if data is still loading when the document is destroyed, which is the case in the API unit-tests that load the `tracemonkey.pdf` file.
While this patch prevents these kind of problems, and thus allows us to update Jasmine again, I cannot help but thinking that it's slightly "hacky". Basically, we'll simply catch and ignore (some) rejected promises once the document is destroyed and/or its data loading is aborted. However, I don't *think* that these changes should cause issues in general, since we don't really care about errors once document destruction has started (note e.g. the fair number of `catch` handlers ignoring `AbortException`s already).
*The [api-minor] label probably ought to have been added to the original PR, given the changes to the `createAnnotationLayerBuilder` signature (if nothing else).*
This patch fixes the following things:
- Let the `AnnotationLayer.render` method create an `AnnotationStorage`-instance if none was provided, thus making the parameter *properly* optional. This not only fixes the reference tests, it also prevents issues when the viewer components are used.
- Stop exporting `AnnotationStorage` in the official API, i.e. the `src/pdf.js` file, since it's no longer necessary given the change above. Generally speaking, unless absolutely necessary we probably shouldn't export unused things in the API.
- Fix a number of JSDocs `typedef`s, in `src/display/` and `web/` code, to actually account for the new `annotationStorage` parameter.
- Update `web/interfaces.js` to account for the changes in `createAnnotationLayerBuilder`.
- Initialize the storage, in `AnnotationStorage`, using `Object.create(null)` rather than `{}` (which is the PDF.js default).
The f1040-annotations reftest started hanging after #12107. We traced
this to `TypeError: can't access property "getOrCreateValue", storage is
undefined`.
We essentially need to add `annotationStorage` to the parameters in
test/driver.js.
This PR adds typescript definitions from the JSDoc already present.
It adds a new gulp-target 'types' that calls 'tsc', the typescript
compiler, to create the definitions.
To use the definitions, users can simply do the following:
```
import {getDocument, GlobalWorkerOptions} from "pdfjs-dist";
import pdfjsWorker from "pdfjs-dist/build/pdf.worker.entry";
GlobalWorkerOptions.workerSrc = pdfjsWorker;
const pdf = await getDocument("file:///some.pdf").promise;
```
Co-authored-by: @oBusk
Co-authored-by: @tamuratak
These changes improves the existing search functionality triggered when
the URL contains a `#search` hash.
In addition to performing the actual search immediately like before,
this will ensure the search text field inside the find bar gets
populated with the text currently being searched for.
Besides making general sense, as far as I can tell, this patch should also prevent *one* source of `Uncaught (in promise) ...` exceptions.
Unfortunately `reason instanceof AbortError` doesn't work here, since `AbortError` isn't actually defined in browsers; note how even the DOM specification contains an example using the `name` property: https://dom.spec.whatwg.org/#aborting-ongoing-activities
This patch prevents the following errors from being logged in the console, when the unit-tests are running:
- Firefox: `Uncaught (in promise) DOMException: The operation was aborted.`
- Chrome: `Uncaught (in promise) DOMException: The user aborted a request.`
Looking carefully at this code, you'll notice that the `loadDocument` function has no less than *three* Promise handling functions. This obviously makes no sense, since a Promise can only have one resolve and one reject handler.
Hence the final `onFailure`-case is unreachable, which only serves to add confusion when reading the code. Note that this code has been re-factored more than once over the years, but it seems as if this may even have been incorrect already in PR 3310 (and no-one have noticed for seven years :-).
There's quite frankly no particular reason to special-case Type3-fonts with image resources, which are very rare anyway, now that we have a general mechanism for sending/receiving images globally.
While the `CharProcs` streams of Type3-fonts *usually* don't rely on dependencies, such as e.g. images, it does happen in some cases.
Currently any dependencies are simply appended to the parent operatorList, which in practice means *only* the operatorList of the *first* page where the Type3-font is being used.
However, there's one thing that's slightly unfortunate with that approach: Since fonts are global to the PDF document, we really ought to ensure that any Type3 dependencies are appended to the operatorList of *all* pages where the Type3-font is being used. Otherwise there's a theoretical risk that, if one page has its rendering paused, another page may try to use a Type3-font whose dependencies are not yet fully resolved. In that case there would be errors, since Type3 operatorLists are executed synchronously.
Hence this patch, which ensures that all relevant pages will have Type3 dependencies appended to the main operatorList. (Note here that the `OperatorList.addDependencies` method, via `OperatorList.addDependency`, ensures that a dependency is only added *once* to any operatorList.)
Finally, these changes also remove the need for the "waiting for the main-thread"-hack that was added to `PartialEvaluator.buildPaintImageXObject` as part of fixing issue 10717.
Jasmine >= 3.6.0 causes intermittent test failures because of random
task abortions.
Puppeteer >= 4.0.0 causes ENOTEMPTY/EBUSY errors during shutdown on the
Windows bot.
Moreover, `jasmine-core` is a dependency of `jasmine` so it doesn't have
to be required separately.
This patch should *hopefully* remove the `Unhandled promise rejection: ...` errors, by returning the "final" promise. Also, by pausing/delaying of rendering slightly the likelihood of the test failing in the first place should thus be reduced.
This uses a similar format to the reference-test logging, and will help determine in *exactly* which browser the failure occurred (since the tests run concurrently).
- Given the `DefaultExternalServices` implementation, the `PDFViewerApplication.supportsDocumentFonts` getter is guaranteed to be defined and we can thus remove some (now) unnecessary `PDFJSDev` checks from the `webViewerInitialized` function.
- By slightly tweaking the "pdfBugEnabled" definition in `web/app_options`, similar to the existing ones for "workerSrc" and "cMapUrl", we can remove some `PDFJSDev` checks from the `PDFViewerApplication._parseHashParameters` method.
The current behavior for `getPagesOverview` assumes we want to only
auto-rotate if:
- `enablePrintAutoRotate` is `true`
- `isFirstPagePortrait !== isPortraitOrientation(size)`
This second check is what is breaking #9297. The two PDFs linked have a
landscape orientation first page, as well as subsequent pages. Since
`false === false`, we print portrait.
Let's drop the comparison with `isFirstPagePortrait`, and print
landscape if `!isPortraitOrientation(size)`.
Fixes#9297.
When the old `Dict.getAll()` method was removed, it was replaced with a `Dict.getKeys()` call and `Dict.get(...)` calls (in a loop).
While this pattern obviously makes a lot of sense in many cases, there's some instances where we actually want the *raw* `Dict` values (i.e. `Ref`s where applicable). In those cases, `Dict.getRaw(...)` calls are instead used within the loop. However, by introducing a new `Dict.getRawValues()` method we can reduce the number of (strictly unnecessary) function calls by simply getting the *raw* `Dict` values directly.
Using a `Map` instead of an `Object` provides some advantages such as
cheaper ways to get the size of the cache, to find out if an entry is
contained in the cache and to iterate over the cache. Moreover, we can
clear and re-use the same `Map` object now instead of creating a new
one.
Since this method calls `Dict.get` to fetch data, there could thus be `Error`s thrown in corrupt PDF documents when attempting to resolve an indirect object.
To ensure that this won't ever become a problem, we change the method to be `async` such that a rejected Promise would be returned and general OperatorList parsing won't break.
- Replace the existing loops with `for...of` variants instead.
- Make use of `continue`, to reduce indentation and to make the code (slightly) easier to follow, when checking `/Resources` entries.
This case should no longer happen, given the `instanceof Ref` branch just above (added in PR 6971).
Also, I've run the entire test-suite locally with `continue` replaced by `throw new Error(...)` and didn't find any problems.
Given that this method is used during what's essentially a *pre*-parsing stage, before the actual OperatorList parsing occurs, on second thought it doesn't seem at all necessary to warn and trigger fallback in cases where there's lookup errors.
*Please note:* Any any errors will still be either suppressed or thrown, according to the `ignoreErrors` option, during the *actual* OperatorList parsing.
It turns out that `getTextContent` suffers from *similar* problems with repeated GStates as `getOperatorList`; please see the previous patch.
While only `/ExtGState` resources containing Fonts will actually be *parsed* by `PartialEvaluator.getTextContent`, we're still forced to fetch/validate repeated `/ExtGState` resources even though *most* of them won't affect the textContent (since they mostly contain purely graphical state).
With these changes we also no longer need to immediately reset the current text-state when encountering a `setGState` operator, which may thus improve text-selection in some cases.
This patch will help pathological cases the most, with issue 2813 being a particularily problematic example. While there's only *four* `/ExtGState` resources, there's a total `29062` of `setGState` operators. Even though parsing of a single `/ExtGState` resource is quite fast, having to re-parse them thousands of times does add up quite significantly.
For simplicity we'll only cache "simple" `/ExtGState` resource, since e.g. the general `SMask` case cannot be easily cached (without re-factoring other code, which may have undesirable effects on general parsing).
By caching "simple" `/ExtGState` resource, we thus improve performance by:
- Not having to fetch/validate/parse the same `/ExtGState` data over and over.
- Handling of repeated `setGState` operators becomes *synchronous* during the `OperatorList` building, instead of having to defer to the event-loop/microtask-queue since the `/ExtGState` parsing is done asynchronously.
---
Obviously I had intended to include (standard) benchmark results with this patch, but for reasons I don't understand the test run-time (even with `master`) of the document in issue 2813 is *a lot* slower than in the development viewer (making normal benchmarking infeasible).
However, testing this manually in the development viewer (using `pdfBug=Stats`) shows a *reduction* of `~10 %` in the rendering time of the PDF document in issue 2813.
Stop special-casing the (very unlikely) "no `/XObject` found"-scenario, when parsing `OPS.paintXObject` operators, in `PartialEvaluator.{getOperatorList, getTextContent}`
Originally there weren't any (generally) good ways to handle errors gracefully, on the worker-side, however that's no longer the case and we can simply fallback to the existing `ignoreErrors` functionality instead.
Also, please note that the "no `/XObject` found"-scenario should be *extremely* unlikely in practice and would only occur in corrupt/broken documents.
Note that the `PartialEvaluator.getOperatorList` case is especially bad currently, since we'll simply (attempt to) send the data as-is to the main-thread. This is quite bad, since in a corrupt/broken document the data *could* contain anything and e.g. be unclonable (which would cause breaking errors).
Also, we're (obviously) not attempting to do anything with this "raw" `OPS.paintXObject` data on the main-thread and simply ensuring that we never send it definately seems like the correct approach.
This special-case was added in PR 1992, however it became unnecessary with the changes in PR 4824 since all of the ColorSpace parsing is now done on the worker-thread (with only RGB-data being sent to the main-thread).
Originally ColorSpaces were only *partially* parsed on the worker-thread, to obtain an IR-format which was sent to the main-thread. This had the somewhat unfortunate side-effect of causing the majority of the (potentially heavy) ColorSpace parsing to happen on the main-thread.
Hence PR 4824 which, among other things, changed ColorSpaces to be *fully* parsed on the worker-thread with only RGB-data being sent to the main-thread.
While it thus originally was necessary to have `ColorSpace.{parseToIR, fromIR}` methods, to handle the worker/main-thread split, that's no longer the case and we can thus reduce all of the ColorSpace parsing to one method instead.
Currently, when parsing a ColorSpace, we call `ColorSpace.parseToIR` which parses the ColorSpace-data from the document and then creates the IR-format. We then, immediately, call `ColorSpace.fromIR` which parses the IR-format and then finally creates the actual ColorSpace.[1]
All-in-all, this leads to a fair amount of unnecessary indirection which also (in my opinion) makes the code less clear.
Obviously these changes are not really expected to have a significant effect on performance, especially with the recently added caching of ColorSpaces, however there'll now be strictly fewer function calls, less memory allocated, and overall less parsing required during ColorSpace-handling.
---
[1] For ICCBased ColorSpaces, given the validation necessary, this currently even leads to parsing an /Alternate ColorSpace *twice*.
Given the dummy-methods on `DefaultExternalServices`, there's no longer any compelling reason not to (attempt to) report telemetry unconditionally.
The only larger change consists of moving the `KNOWN_VERSIONS` and `KNOWN_GENERATORS` arrays ouf of the `PDFViewerApplication._initializeMetadata` method.
*Please note:* Most of this patch consists of whitespace-only changes.
There's a few things that could be improved in the current implementation, such as:
- It's currently necessary to *both* manually track the `featureId`s which should trigger delayed fallback, as well as manually report telemetry in affected cases.
Obviously there's only two call-sites as of now (forms and javaScript), but it still feels somewhat error-prone especially if more cases were to be added in the future. To address this, this patch adds a new (private) method which abstracts away these details from the call-sites.
- Generally, it also seems nice to reduce *and* simplify the amount of state we need to store on `PDFViewerApplication` in order to support the "delayedFallback" functionality.
Also, having to *manually* work with the "delayedFallback"-array in multiple places feels cumbersome and makes e.g. the `PDFViewerApplication.fallback` method less clear as to its behaviour.
- Having code *outside* of `PDFViewerApplication`, i.e. in the event handlers, directly access properties which are marked as "private" via a leading underscore doesn't seem that great in general.
Furthermore, having the event handlers directly deal with that should be internal state also seem unfortunate. To address this, the patch will instead make use of a new `PDFViewerApplication.triggerDelayedFallback` callback.
- There's at least one code-path in the viewer, see `PDFViewerApplication.error`, where `fallback` can be called without an argument.
It's currently possible (although maybe somewhat unlikely) that such a call *could* be overridden by the `featureId` of a pending "delayedFallback" call, thus not reporting the *correct* fallback reason.
- The "delayedFallback"-state weren't being reset on document close (which shouldn't affect Firefox, but nonetheless it ought to be fixed).
Note how the `getFontID`-method in `src/core/fonts.js` is *completely* global, rather than properly tied to the current document. This means that if you repeatedly open and parse/render, and then close, even the *same* PDF document the `fontID`s will still be incremented continuously.
For comparison the `createObjId` method, on `idFactory`, will always create a *consistent* id, assuming of course that the document and its pages are parsed/rendered in the same order.
In order to address this inconsistency, it thus seems reasonable to add a new `createFontId` method on the `idFactory` and use that when obtaining `fontID`s. (When the current `getFontID` method was added the `idFactory` didn't actually exist yet, which explains why the code looks the way it does.)
*Please note:* Since the document id is (still) part of the `loadedName`, it's thus not possible for different documents to have identical font names.
This removes additional `// eslint-disable-next-line no-shadow` usage, which our old pseudo-classes necessitated.
Most of the re-formatting changes, after the `class` definitions and methods were fixed, were done automatically by Prettier.
*Please note:* I'm purposely not doing any `var` to `let`/`const` conversion here, since it's generally better to (if possible) do that automatically on e.g. a directory basis instead.
This will simplify the `class` conversion in the next patch, and with modern JavaScript the moved code is still limited to the current module scope.
*Please note:* For improved consistency with our usual formatting, the `TILING_PATTERN`/`SHADING_PATTERN` constants where re-factored slightly.
This removes additional `// eslint-disable-next-line no-shadow` usage, which our old pseudo-classes necessitated.
*Please note:* I'm purposely not doing any `var` to `let`/`const` conversion here, since it's generally better to (if possible) do that automatically on e.g. a directory basis instead.
*Yet another instalment in the never-ending series of things that you think of __after__ a patch has landed.*
Since `Function`s are only cached by reference, we thus don't need to allocate storage for names in `LocalFunctionCache` instances. Obviously the effect of these changes are *really tiny*, but it seems reasonable in principle to avoid allocating data structures that are guaranteed to be unused.
Note that compared other structures, such as e.g. Images and ColorSpaces, `Function`s are not referred to by name, which however does bring the advantage of being able to share the cache for an *entire* page.
Furthermore, similar to ColorSpaces, the parsing of individual `Function`s are generally fast enough to not really warrant trying to cache them in any "smarter" way than by reference. (Hence trying to do caching similar to e.g. Fonts would most likely be a losing proposition, given the amount of data lookup/parsing that'd be required.)
Originally I tried implementing this similar to e.g. the recently added ColorSpace caching (and in a couple of different ways), however it unfortunately turned out to be quite ugly/unwieldy given the sheer number of functions/methods where you'd thus need to pass in a `LocalFunctionCache` instance. (Also, the affected functions/methods didn't exactly have short signatures as-is.)
After going back and forth on this for a while it seemed to me that the simplest, or least "invasive" if you will, solution would be if each `PartialEvaluator` instance had its *own* `PDFFunctionFactory` instance (since the latter is already passed to all of the required code). This way each `PDFFunctionFactory` instances could have a local `Function` cache, without it being necessary to provide a `LocalFunctionCache` instance manually at every `PDFFunctionFactory.{create, createFromArray}` call-site.
Obviously, with this patch, there's now (potentially) more `PDFFunctionFactory` instances than before when the entire document shared just one. However, each such instance is really quite small and it's also tied to a `PartialEvaluator` instance and those are *not* kept alive and/or cached. To reduce the impact of these changes, I've tried to make as many of these structures as possible *lazily initialized*, specifically:
- The `PDFFunctionFactory`, on `PartialEvaluator` instances, since not all kinds of general parsing actually requires it. For example: `getTextContent` calls won't cause any `Function` to be parsed, and even some `getOperatorList` calls won't trigger `Function` parsing (if a page contains e.g. no Patterns or "complex" ColorSpaces).
- The `LocalFunctionCache`, on `PDFFunctionFactory` instances, since only certain parsing requires it. Generally speaking, only e.g. Patterns, "complex" ColorSpaces, and/or (some) SoftMasks will trigger any `Function` parsing.
To put these changes into perspective, when loading/rendering all (14) pages of the default `tracemonkey.pdf` file there's now a total of 6 `PDFFunctionFactory` and 1 `LocalFunctionCache` instances created thanks to the lazy initialization.
(If you instead would keep the document-"global" `PDFFunctionFactory` instance and pass around `LocalFunctionCache` instances everywhere, the numbers for the `tracemonkey.pdf` file would be instead be something like 1 `PDFFunctionFactory` and 6 `LocalFunctionCache` instances.)
All-in-all, I thus don't think that the `PDFFunctionFactory` changes should be generally problematic.
With these changes, we can also modify (some) call-sites to pass in a `Reference` rather than the actual `Function` data. This is nice since `Function`s can also be `Streams`, which are not cached on the `XRef` instance (given their potential size), and this way we can avoid unnecessary lookups and thus save some additional time/resources.
Obviously I had intended to include (standard) benchmark results with these changes, but for reasons I don't really understand the test run-time (even with `master`) of the document in issue 2541 is quite a bit slower than in the development viewer.
However, logging the time it takes for the relevant `PDFFunctionFactory`/`PDFFunction ` parsing shows that it takes *approximately* `0.5 ms` for the `Function` in question. Looking up a cached `Function`, on the other hand, is *one order of magnitude faster* which does add up when the same `Function` is invoked close to 2000 times.
This moves, and slightly simplifies, code that's currently residing in the unit-test utils into the actual library, such that it's bundled with `GENERIC`-builds and used in e.g. the API-code.
As an added bonus, this also brings out-of-the-box support for CMaps in e.g. the Node.js examples.
The default viewer, and thus Firefox, depends on the `RenderTask.onContinue` functionality to pause/continue rendering (such that the most visible page always renders first).
Despite this functionality thus being very important, it has however never actually been tested *at all* as far as I can tell. Hence this patch which adds a new boolean `renderTaskOnContinue` parameter (`false` by default), that can be used to force a reference-test to use the `RenderTask.onContinue` code-path in the `InternalRenderTask` class.
Note that I purposely made this new reference-test behaviour *optional*, since I didn't want to negatively affect the general runtime of the tests (given that there's a slight delay added to the rendering). Also, for e.g. benchmarking you'd most likely want to stay away from the `RenderTask.onContinue` functionality for similar reasons.
This should reduce the possibility of accidentally truncating some inline images, while *not* causing the "EI" detection to become significantly slower.[1]
There's obviously a possibility that these added checks are not sufficient to catch *every* single case of "EI" sequences within the actual inline image data, but without specific test-cases I decided against over-engineering the solution here.
*Please note:* The interpolation issues are somewhat orthogonal to the main issue here, which is the truncated image, and it's already tracked elsewhere.
---
[1] I've looked at the issue a few times, and this is the first approach that I was able to come up with that didn't cause *unacceptable* performance regressions in e.g. issue 2618.
Versions of `needle` prior to `2.5.0` cannot cope with redirects (as documented: https://github.com/tomas/needle/issues/312).
This prevents prebuilt `canvas` binaries from being downloaded on MacOS,
requiring the global install of its dependencies.
Updating Needle restores it to functionality, addressing this. It also avoids
the need to add `request` to `package.json`; it also obsoletes https://github.com/mozilla/pdf.js/pull/12018.
With the changes made in PR 9659, `ColorSpace.fromIR` no longer takes a second `pdfFunctionFactory` parameter and there's thus one call-site that can be simplified.
This patch contains the following *notable* improvements:
- Changes the `ColorSpace.parse` call-sites to, where possible, pass in a reference rather than actual ColorSpace data (necessary for the next point).
- Adds (local) caching of `ColorSpace`s by `Ref`, when applicable, in addition the caching by name. This (generally) improves `ColorSpace` caching for e.g. the SMask code-paths.
- Extends the (local) `ColorSpace` caching to also apply when handling Images and Patterns, thus further reducing unneeded re-parsing.
- Adds a new `ColorSpace.parseAsync` method, almost identical to the existing `ColorSpace.parse` one, but returning a Promise instead (this simplifies some code in the `PartialEvaluator`).
This will allow caching of ColorSpaces by either `Name` *or* `Ref`, which doesn't really make sense for images, thus allowing (better) caching for ColorSpaces used with e.g. Images and Patterns.
Defining this *inline* in the "constructor" looks slightly weird (I really don't know why I wrote it like that originally), and it can simply be changed to a regular method instead.
Tweak the `QueueOptimizer` to recognize `OPS.paintImageMaskXObject` operators as *repeated* when the "skew" transformation matrix elements are non-zero (issue 8078)
As can be seen in the code there's a handful of places where this structure needs to be iterated, something that becomes cumbersome when dealing with `Object`s. Hence, by changing this to a `Map` instead we can both simplify the code and avoid creating unnecessary closures.
Particularily the `PDFPageProxy._tryCleanup` method becomes a lot more readable, at least in my opinion.
Finally, since this property is intended to be "private" the name is adjusted to reflect that.
This patch aims to simplify the `PDFPageProxy._destroy` method, by:
- Replacing the unnecessary `forEach` with a "regular" `for`-loop instead.
- Use a more appropriate variable name, since `intentState.renderTasks` contain instances of `InternalRenderTask`.
- Move the "is rendering completed"-handling to a new `InternalRenderTask.completed` getter, to abstract away some (mostly) internal `InternalRenderTask` state.
In hindsight, using the `for (let [key, value] of myMap) { ... }`-format when we don't care about the `key` probably wasn't such a great idea. Since `Map`s have explicit support for iterating either `key`s or `value`s, we should probably use that instead here.
*First of all, I should mention that my understanding of the finer details of the `QueueOptimizer` (and its related `CanvasGraphics` methods) is somewhat limited.*
Hence I'm not sure if there's actually a very good reason for *only* considering ImageMasks where the "skew" transformation matrix elements are zero as *repeated*, however simply looking at the code I just don't see why these elements cannot be non-zero as long as they are *all identical* for the ImageMasks.
Furthermore, looking at the *group* case (which is what we're currently falling back to), there's no particular limitation placed upon the transformation matrix elements.
While this patch obviously isn't enough to *completely* fix the issue, since there should be a visible Pattern rendered as well[1], it seem (at least to me) like enough of an improvement that submitting this is justified.
With these changes the referenced PDF document will no longer hang the *entire* browser, and rendering also finishes in a *reasonable* time (< 10 seconds for me) which seem fine given the *huge* number of identical inline images present.[2]
---
[1] Temporarily changing the Pattern to a solid color *does* render the correct/expected area, which suggests that the remaining problem is a pre-existing issue related to the Pattern-handling itself rather than the `QueueOptimizer` functionality.
[2] The document isn't exactly rendered immediately in e.g. Adobe Reader either.
This removes multiple instances of `// eslint-disable-next-line no-shadow`, which our old pseudo-classes necessitated.
*Please note:* I'm purposely not doing any `var` to `let`/`const` conversion here, since it's generally better to (if possible) do that automatically on e.g. a directory basis instead.
This removes one instance of `// eslint-disable-next-line no-shadow`, which our old pseudo-classes necessitated.
*Please note:* I'm purposely not doing any `var` to `let`/`const` conversion here, since it's generally better to (if possible) do that automatically on e.g. a directory basis instead.
The `isNaN` check is obviously redundant, since `NaN` is the only value that isn't equal to itself; see https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/NaN#Examples
The `parseFloat`/`parseInt` comparison would make sense if the `value` ever contains a String, which however is never actually the case. Besides looking through the code, I've also run the entire test-suite locally with `assert(typeof value === "number", "encodeNumber");` added at the top of the method and there were no failures.
Hence we can simplify the "is integer" check a bit in the `CFFCompiler.encodeNumber` method.
Because of a really stupid `Promise`-related mistake on my part, when re-factoring `PDFImage.buildImage` during the `NativeImageDecoder` removal, we're no longer re-throwing errors occuring during image parsing/decoding as intended.
The result is that some (fairly) corrupt documents will never finish loading, and unfortunately there were apparently no sufficiently corrupt images in the test-suite to catch this.
For now we need to use a Babel-plugin, since Webpack 4.x doesn't seem to support it yet. (Most likely we'll have to update to Webpack 5, once that becomes available, in order for this to be directly supported. This is thus also blocked on removing the `webpack-stream` package.)
While the `??` operator will thus always be transpiled by Babel, even in modern builds, simply supporting it for development purposes seems like a step in the right direction.
https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Operators/Nullish_coalescing_operator
Since this is completely internal functionality, and furthermore limited to the worker-thread, this change should thus not have any observable effect for e.g. an API-user.
Apparently I completely overlooked the fact that with the changes in PR 11069 these properties became *completely* unused, and consequently they thus ought to be removed.
Since there's (essentially) no tests for the SVG-backend, these changes didn't make in into PR 11912 when the code in the `src/display/canvas.js` file was modified.
Since this helper function is no longer used anywhere in the main code-base, but only in a couple of unit-tests, it's thus being moved to a more appropriate spot.
Finally, the implementation of `isEmptyObj` is also tweaked slightly by removing the manual loop.
Compared to regular `Object`s, `Map`s have a number of advantageous properties: Of particular importance in this case is the built-in iteration support, and that determining if the structure is empty is easy.
Compared to regular `Object`s, `Map`s have a number of advantageous properties: Of particular importance in this case is the built-in iteration support, and that determining if the structure is empty is easy.
Compared to regular `Object`s, `Map`s (and `Set`s) have a number of advantageous properties: Of particular importance in this case is the built-in iteration support, and that determining if the structure is empty is easy.
The `RefSet` primitive predates ES6, so that most likely explains why an
object is used internally to track the entries. However, nowadays we can
use built-in JavaScript sets for this purpose. Built-in types are often
more efficient/optimized and using it makes the code a bit more clear
since we don't have to assign `true` to keys anymore just to indicate
their presence.
Both of the removed methods were added in PR 2719, however they are no longer used:
- It appears that `hasPendingRequests` was never used at all, even from the beginning.
- The only general PDF.js library usage of `abortAllRequests` was removed in PR 6879, which is now four years ago. (Originally the Firefox-specific network implementation, see https://searchfox.org/mozilla-central/source/browser/extensions/pdfjs/content/PdfJsNetwork.jsm, was shared with the `src/display/network.js` file and *there* this method is used. However, since all of the Firefox-specific code now lives directly in mozilla-central, that's not relevant for the removal in this patch.)
After PRs 10727 and 11912, the code responsible for sending the decoded image data to the main-thread has now become a fair bit more involved the previously.
To reduce the amount of duplication here, the actual code responsible for sending the data is thus extracted into a new helper method instead.
Avoid calling Math.pow if possible when calculating the transfer
function of the CalRGB color space since calling Math.pow is expensive.
If the value of color is larger than the threshold, 0.99554525,
the final result of the transform is larger that 254.5
since ((1 + 0.055) * 0.99554525 ** (1 / 2.4) - 0.055) * 255 === 254.50000003134699
In the old code the use of an Array meant that we had to *manually* track the `numChunksLoaded` property, given that simply using the Array `length` wouldn't have worked since there's no guarantee that the data is loaded in order when e.g. range requests are in use.
Tracking closely related state *separately* in this manner never seem like a good idea, and we can now instead utilize a Set to avoid that.
On ISO/IEC 10918-6:2013 (E), section 6.1: (http://www.itu.int/rec/T-REC-T.872-201206-I/en)
"Images encoded with three components are assumed to be RGB data encoded as YCbCr unless the image contains an APP14 marker segment as specified in 6.5.3, in which case the colour encoding is considered either RGB or YCbCr according to the application data of the APP14 marker segment"
But common jpeg libraries consider RGB too if components index are ASCII R (0x52), G (0x47) and B (0x42): https://stackoverflow.com/questions/50798014/determining-color-space-for-jpeg/50861048
Issue #11931
Using `require.resolve("worker-loader")` to check if `worker-loader` is installed causes webpack to include `worker-loader` in the output bundle, which is not the intended effect. Aside from increasing the bundle size unnecessarily, it also causes errors for webpack configs with targets that don't have node's built-in modules.
These errors can be fixed by configuring webpack `externals` to exclude `worker-loader`, but it's more difficult to figure out this solution than to figure out that `worker-loader` needs to be installed (even without this explicit error message).
To solve this, the explicit check for `worker-loader` has been removed. An alternative solution would be to use webpack's `resolveWeak`. Documentation has also been added in `examples/webpack` to help users.
Since *inline* images, i.e. those defined inside of `/Contents` streams, are by their very definition page-specific it thus seem like a good idea to actually enforce that they won't accidentally end up in the `GlobalImageCache`.
When converting this file to use standard `import`/`export` statements, I sorted the exports in the same order as the imports to simplify things.
However, looking at the list of `export`ed properties it probably doesn't hurt to add a couple of comments to clarify from where specifically the `export`s originated.
By updating to the new major version of Acorn, we'll get support for newer ECMAScript features as they become available (although some features are currently also blocked by ESLint support and/or SystemJS usage).
Please see https://github.com/acornjs/acorn/releases/tag/7.2.0 for details.
It turns out that `getTextContent` suffers from *similar* problems with repeated images as `getOperatorList`; please see the previous patch.
While only `/XObject` resources of the `Form`-type will actually be *parsed* in `PartialEvaluator.getTextContent`, since those are the only ones that may contain text, we're still forced to fetch repeated image resources where the name differs (but not the reference).
Obviously it's less bad in this case, since we're not actually parsing `/XObject`s of e.g. the `Image`-type. However, you still want to avoid even fetching the data whenever possible, since `Stream`s are not cached on the `XRef` instance (given their potential size) and the lookup can thus be somewhat expensive in general.
To address these issues, we can simply replace the exiting name-only caching in `PartialEvaluator.getTextContent` with a new cache backed by `LocalImageCache` instead.
Currently the local `imageCache`, as used in `PartialEvaluator.getOperatorList`, will miss certain cases of repeated images because the caching is *only* done by name (usually using a format such as e.g. "Im0", "Im1", ...).
However, in some PDF documents the `/XObject` dictionaries many contain hundreds (or even thousands) of distinctly named images, despite them referring to only a handful of actual image objects (via the XRef table).
With these changes we'll now cache *local* images using both name and (where applicable) reference, thus improving re-usage of images resources even further.
This patch was tested using the PDF file from [bug 857031](https://bugzilla.mozilla.org/show_bug.cgi?id=857031), i.e. https://bug857031.bmoattachments.org/attachment.cgi?id=732270, with the following manifest file:
```
[
{ "id": "bug857031",
"file": "../web/pdfs/bug857031.pdf",
"md5": "",
"rounds": 250,
"lastPage": 1,
"type": "eq"
}
]
```
which gave the following results when comparing this patch against the `master` branch:
```
-- Grouped By browser, page, stat --
browser | page | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ---- | ------------ | ----- | ------------ | ----------- | --- | ----- | -------------
firefox | 0 | Overall | 250 | 2749 | 2656 | -93 | -3.38 | faster
firefox | 0 | Page Request | 250 | 3 | 4 | 1 | 50.14 | slower
firefox | 0 | Rendering | 250 | 2746 | 2652 | -94 | -3.44 | faster
```
While this is certainly an improvement, since we now avoid re-parsing ~1000 images on the first page, all of the image resources are small enough that the total rendering time doesn't improve that much in this particular case.
In pathological cases, such as e.g. the PDF document in issue 4958, the improvements with this patch can be very significant. Looking for example at page 2, from issue 4958, the rendering time drops from ~60 seconds with `master` to ~30 seconds with this patch (obviously still slow, but it really showcases the potential of this patch nicely).
Finally, note that there's also potential for additional improvements by re-using `LocalImageCache` instances for e.g. /XObject data of the `Form`-type. However, given that recent changes in this area I purposely didn't want to complicate *this* patch more than necessary.
In the PDF file from the issue below, the fill alpha (`ca`) is set
before drawing the circles using the `setGState` operator. Doing so
causes the global alpha to be set on the canvas' context for the canvas
back-end, but this was not handled in the SVG back-end. This patch fixes
that by taking the fill opacity into account when drawing shading
patterns in the same way as done elsewhere so it is only included if the
value is non-default.
Fixes#11812.
When "Cleanup" is triggered, you obviously need to remove all globally cached data on *both* the main- and worker-threads.
However, the current the implementation of the `GlobalImageCache.clear` method also means that we lose *all* information about which images were cached and not just their data. This thus has the somewhat unfortunate side-effect of requiring images, which were previously known to be "global", to *again* having to reach `NUM_PAGES_THRESHOLD` before being cached again.
To avoid doing unnecessary parsing after "Cleanup", we can thus let `GlobalImageCache.clear` keep track of which images were cached while still removing their actual data. This should not have any significant impact on memory usage, since the only extra thing being kept is a `RefSetCache` (essentially an Object) with a couple of `Set`s containing only integers.
With the changes made in the previous patch, the `web/app_options.js` file no longer depends on anything *except* files residing in the `web/` folder. Hence the `gulp default_preferences` task can now be further simplified and thus becomes even faster than before; see also PR 11724.
With the changes in previous patches, the `disableCreateObjectURL` option/functionality is no longer used for anything in the API and/or in the Worker code.
Note however that there's some functionality, mainly related to file loading/downloading, in the GENERIC version of the default viewer which still depends on this option.
Hence the `disableCreateObjectURL` option (and related compatibility code) is moved into the viewer, see e.g. `web/app_options.js`, such that it's still available in the default viewer.
Currently some JPEG images are decoded by the built-in PDF.js decoder in `src/core/jpg.js`, while others attempt to use the browser JPEG decoder. This inconsistency seem unfortunate for a number of reasons:
- It adds, compared to the other image formats supported in the PDF specification, a fair amount of code/complexity to the image handling in the PDF.js library.
- The PDF specification support JPEG images with features, e.g. certain ColorSpaces, that browsers are unable to decode natively. Hence, determining if a JPEG image is possible to decode natively in the browser require a non-trivial amount of parsing. In particular, we're parsing (part of) the raw JPEG data to extract certain marker data and we also need to parse the ColorSpace for the JPEG image.
- While some JPEG images may, for all intents and purposes, appear to be natively supported there's still cases where the browser may fail to decode some JPEG images. In order to support those cases, we've had to implement a fallback to the PDF.js JPEG decoder if there's any issues during the native decoding. This also means that it's no longer possible to simply send the JPEG image to the main-thread and continue parsing, but you now need to actually wait for the main-thread to indicate success/failure first.
In practice this means that there's a code-path where the worker-thread is forced to wait for the main-thread, while the reverse should *always* be the case.
- The native decoding, for anything except the *simplest* of JPEG images, result in increased peak memory usage because there's a handful of short-lived copies of the JPEG data (see PR 11707).
Furthermore this also leads to data being *parsed* on the main-thread, rather than the worker-thread, which you usually want to avoid for e.g. performance and UI-reponsiveness reasons.
- Not all environments, e.g. Node.js, fully support native JPEG decoding. This has, historically, lead to some issues and support requests.
- Different browsers may use different JPEG decoders, possibly leading to images being rendered slightly differently depending on the platform/browser where the PDF.js library is used.
Originally the implementation in `src/core/jpg.js` were unable to handle all of the JPEG images in the test-suite, but over the last couple of years I've fixed (hopefully) all of those issues.
At this point in time, there's two kinds of failure with this patch:
- Changes which are basically imperceivable to the naked eye, where some pixels in the images are essentially off-by-one (in all components), which could probably be attributed to things such as different rounding behaviour in the browser/PDF.js JPEG decoder.
This type of "failure" accounts for the *vast* majority of the total number of changes in the reference tests.
- Changes where the JPEG images now looks *ever so slightly* blurrier than with the native browser decoder. For quite some time I've just assumed that this pointed to a general deficiency in the `src/core/jpg.js` implementation, however I've discovered when comparing two viewers side-by-side that the differences vanish at higher zoom levels (usually around 200% is enough).
Basically if you disable [this downscaling in canvas.js](8fb82e939c/src/display/canvas.js (L2356-L2395)), which is what happens when zooming in, the differences simply vanish!
Hence I'm pretty satisfied that there's no significant problems with the `src/core/jpg.js` implementation, and the problems are rather tied to the general quality of the downscaling algorithm used. It could even be seen as a positive that *all* images now share the same downscaling behaviour, since this actually fixes one old bug; see issue 7041.
Currently image resources, as opposed to e.g. font resources, are handled exclusively on a page-specific basis. Generally speaking this makes sense, since pages are separate from each other, however there's PDF documents where many (or even all) pages actually references exactly the same image resources (through the XRef table). Hence, in some cases, we're decoding the *same* images over and over for every page which is obviously slow and wasting both CPU and memory resources better used elsewhere.[1]
Obviously we cannot simply treat all image resources as-if they're used throughout the entire PDF document, since that would end up increasing memory usage too much.[2]
However, by introducing a `GlobalImageCache` in the worker we can track image resources that appear on more than one page. Hence we can switch image resources from being page-specific to being document-specific, once the image resource has been seen on more than a certain number of pages.
In many cases, such as e.g. the referenced issue, this patch will thus lead to reduced memory usage for image resources. Scrolling through all pages of the document, there's now only a few main-thread copies of the same image data, as opposed to one for each rendered page (i.e. there could theoretically be *twenty* copies of the image data).
While this obviously benefit both CPU and memory usage in this case, for *very* large image data this patch *may* possibly increase persistent main-thread memory usage a tiny bit. Thus to avoid negatively affecting memory usage too much in general, particularly on the main-thread, the `GlobalImageCache` will *only* cache a certain number of image resources at the document level and simply fallback to the default behaviour.
Unfortunately the asynchronous nature of the code, with ranged/streamed loading of data, actually makes all of this much more complicated than if all data could be assumed to be immediately available.[3]
*Please note:* The patch will lead to *small* movement in some existing test-cases, since we're now using the built-in PDF.js JPEG decoder more. This was done in order to simplify the overall implementation, especially on the main-thread, by limiting it to only the `OPS.paintImageXObject` operator.
---
[1] There's e.g. PDF documents that use the same image as background on all pages.
[2] Given that data stored in the `commonObjs`, on the main-thread, are only cleared manually through `PDFDocumentProxy.cleanup`. This as opposed to data stored in the `objs` of each page, which is automatically removed when the page is cleaned-up e.g. by being evicted from the cache in the default viewer.
[3] If the latter case were true, we could simply check for repeat images *before* parsing started and thus avoid handling *any* duplicate image resources.
With these changes SystemJS is now only used, during development, on the worker-thread and in the unit/font-tests, since Firefox is currently missing support for worker modules; please see https://bugzilla.mozilla.org/show_bug.cgi?id=1247687
Hence all the JavaScript files in the `web/` and `src/display/` folders are now loaded *natively* by the browser (during development) using standard `import` statements/calls, thanks to a nice `import-maps` polyfill.
*Please note:* As soon as https://bugzilla.mozilla.org/show_bug.cgi?id=1247687 is fixed in Firefox, we should be able to remove all traces of SystemJS and thus finally be able to use every possible modern JavaScript feature.
As part of reducing our reliance on SystemJS in the development viewer, this patch replaces usage of `require` statements with modern standards `import`/`export` statements instead.
If we want to try and move forward with reducing usage of SystemJS, we don't have much choice but to make these kind changes (despite what prior test-results showed, however I'm no longer able to reproduce the issues locally).
As part of trying to reduce the usage of SystemJS in the development viewer, this patch is a necessary step that will allow removal of some `require` statements.
Currently this uses `SystemJS.import` in non-PRODUCTION mode, but it should be possible to replace those with standard *dynamic* `import` calls in the future.
Given that the `getPDFFileNameFromURL` helper function has a specific code-path for handling non-string inputs, this empty string fallback really isn't necessary at the call-site in `web/pdf_document_properties.js`.
Apparently the old link format used in MOZCENTRAL-builds, with the blob URL separated from the filename with a `?` character violates the specification; see https://bugzilla.mozilla.org/show_bug.cgi?id=1632644#c5
Obviously just removing the `?`-part of the URL would have worked, but that would also have meant that we'd no longer be able to provide the correct filename when the user attempts to download the opened PDF attachment.
To fix this we'll instead append the filename in the hash-part of the URL, which however required using a *custom* hash-parameter to avoid triggering the fallback "named destination" code-paths in the viewer.
Note that only changing the `web/pdf_attachment_viewer.js` file wasn't sufficient to fix the bug, and we also need to tweak the `webViewerInitialized` function in `web/app.js` since MOZCENTRAL-builds used to ignore *everything* in the URL hash.
This particular code is very old, but changing it *should* be completely safe given that the `PDFViewerApplication.setTitleUsingUrl` method since some time now stores both the original URL (in `this.url`) as well as one without the hash (in `this.baseUrl`). The latter one is already used everywhere where it matters, so this change seem fine to me.
This patch thus restores the original behaviour for PDF attachments in the MOZCENTRAL-build, by once again allowing them to be opened *directly* in the browser without downloading. (The fallback added in PR 11845 is obviously kept, since it seems generally useful to have.)
Originally the `default_preferences.json` file was checked into the repository, and we thus needed to load it in non-PRODUCTION mode (which was originally done asynchronously using `XMLHttpRequest`). Over the years a lot has changed and the `default_preferences.json` file is now built, by the `gulp default_preferences` task, from the `web/app_options.js` file. Hence it's no longer necessary, in non-PRODUCTION mode, to use SystemJS here since we can simply use a standard `import` statement instead.
Note how e.g. `web/app.js` already imports from `web/app_options.js` in the same exact way that `web/preferences.js` now does, hence this patch will *not* result in any significant changes in the built/bundled viewer file.
This is another (small) part in trying to reduce usage of SystemJS, with the goal of hopefully getting rid of it completely. (I've started working on this, and doing so has identified a number of problem areas; this patch addresses one of them.)
This replaces some additional `require`/`exports` usage with standard `import`/`export` statements instead.
Hence another, small, part in the effort to reduce the reliance on SystemJS-specific functionality in the development viewer.
By re-factoring the existing gulp tasks, most of the code can be re-used for both the existing `gulp minified` as well as the new `gulp minified-es5` task.
While working on PR 11872, it occurred to me that it probably wouldn't be a bad idea to change the `_parsedAnnotations` getter to handle errors individually for each annotation. This way, one broken/corrupt annotation won't prevent the rest of them from being e.g. fetched through the API.
Given that the `PDFLinkService.setHash` method itself if completely synchronous, moving the handling of "nameddest" to occur last *shouldn't* cause any problems (famous last words).
This way the destination will still override any previous parameter, such as e.g. the "page", as expected. Furthermore, given that the `PDFLinkService.navigateTo` method is asynchronous that should provide additional guarantees that the "nameddest" parameter is always respected.
As sort-of expected, this fairly innocent looking change also required some tweaks in the `PDFHistory` to prevent dummy history entires upon document load (only an issue when both "page" *and* "nameddest" parameters are provided in the hash).
This rule complements the existing `accessor-pairs` nicely, and ensures that a getter/setter pair is always consistently ordered.
Please find additional details about this rule at https://eslint.org/docs/rules/grouped-accessor-pairs
Having `assert` calls without a message string isn't very helpful when debugging, and it turns out that it's easy enough to make use of ESLint to enforce better `assert` call-sites.
In a couple of cases the `assert` calls were changed to "regular" throwing of errors instead, since that seemed more appropriate.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/no-restricted-syntax
This should ensure that a page will always render successfully, even if there's errors during the Annotation fetching/parsing.
Additionally the `OperatorList.addOpList` method is also adjusted to ignore invalid data, to make it slightly more robust.
All of the other *similar* helper functions only target one file per function, and there's no particular reason for this one to be different.
This patch will simplify future changes, e.g. experimenting with using `gulp watch` instead of SystemJS for the development viewer.
Not only is there no code depending on it now, the actual task itself doesn't even make sense as-is. Note that it uses the default `DEFINES` configuration *unaltered*, which is neither useful nor correct since the resulting build thus won't make sense without an actual built target set.
Given that none of the PDF.js contributors know React, maintaining and/or providing supporting for the example isn't really feasible unfortunately.
Even something as simple as running/testing the example becomes difficult for anyone completely unfamiliar with React, and furthermore:
- It's very difficult to tell if the example demonstrates React best-practices, since the PDF.js contributors don't know React.
- We also have no reasonable way of keeping the example up-to-date with changes in React.
- The React example, in its current form, is even *hard-coding* the PDF.js version to a now unsupported version.
- The example is currently triggering "fake worker" usage, see issue 11729, which is really *really* bad. Note that the "fake worker" functionality is *only* intended as a fallback, and it should absolutely *not* under any circumstances be advertised and certainly shouldn't be triggered in official PDF.js examples.
By preserving the exception type, more fine-grained error handling can be performed via client-side logic (e.g. redirect to a search page if a PDF is not found, or to a ticket system in case of invalid PDF files).
The original exception is now re-thrown.
Fixes#11658
We already rendered the name for radio buttons, but it was missing for
all other interactive form elements. This commit adds that so that
values entered in form elements can be read based on the element name.
To save time or resources during development it can be useful to run
tests only in Firefox. Previously this could be done by editing the
browser manifest file, but since that file is no longer used for
Puppeteer, this command line flag replaces it. For example, executing
`gulp unittest --noChrome` will only run the unit tests in Firefox.
This commit replaces our own infrastructure for handling browsers during
tests with Puppeteer. Using our own infrastructure for this had a few
downsides:
- It has proven to not always be reliable, especially when closing the
browser, causing failures on the bots because browsers were still
running even though they should have been stopped. Puppeteer should do
a better job with this because it uses the browser's test built-in
instrumentation tools for this (the devtools protocol) which our code
didn't. This also means that we don't have to pass
parameters/preferences to tweak browser behavior anymore.
- It requires the browsers under test to be installed on the system,
whereas Puppeteer downloads the browsers before the test. This means
that setup is much easier (no more manual installations and browser
manifest files) as well as testing with different browser versions
(since they can be provisioned on demand). Moreover, this ensures that
contributors always run the tests in both Firefox and Chrome,
regardless of which browsers they have installed locally.
- It's all code we have to maintain, so Puppeteer abstracts away how the
browsers start/stop for us so we don't have to keep that code.
By default, Puppeteer only installs one browser during installation,
hence the need for a post-install script to install the second browser.
This requires `cross-env` to make passing the environment variable work
on both Linux and Windows.
The other testing code already uses the name of the browser as the
unique identifier, so I don't see a good reason to not use that for
identifying browsers to quit as well. Doing so simplifies the (already
somewhat complex) testing logic and ensures that we can use existing
functionality (such as the `getSession` function) to retrieve sessions.
Given that the `NativeImageDecoder.{isSupported, isDecodable}` methods require both dictionary lookups *and* ColorSpace parsing, in hindsight it actually seems more reasonable to the `JpegStream.maybeValidDimensions` checks *first*.
This is a simple work-around for https://bugzilla.mozilla.org/show_bug.cgi?id=1632644 which was caused by platform changes in Firefox. Ideally the Firefox bug should still be fixed, but these PDF.js changes seem generally useful to prevent both current and future issues here.
There's no particular reason for using the PDF.js helper function `createObjectURL` here, given that the relevant code-path is already guarded by multiple "disableCreateObjectURL" option checks.
Given that `URL.createObjectURL` is assumed to always be available in MOZCENTRAL builds, note the existing usage in the file, there's no reason to depend on the PDF.js helper function `createObjectURL` at all here.
Furthermore this patch also changes `DownloadManager.downloadData` to actually revoke the `blobUrl` after downloading has completed, which is similar to the existing code in `DownloadManager.download`.
*This patch implements https://github.com/mozilla/pdf.js/pull/11777#issuecomment-609741348*
This extends the work from PR 11773 and 11777 further, by immediately releasing the `font.data` property once the font been attached to the DOM. By not unnecessarily holding onto this data on the main-thread, we'll thus reduce the memory usage of fonts even further (especially beneficial in longer documents with composite fonts).
The new behaviour is controlled by the recently added `fontExtraProperties` API option (adding a new option just for this patch didn't seem necessary), since there's one edge-case in the SVG renderer where the `font.data` property is necessary (see the `pdf2svg` example).
Note that while the default viewer does run clean-up with an idle timeout, that timeout will be reset whenever rendering occurs *or* when scrolling happens in the viewer. In practice this means that unless the user doesn't interact with the viewer in *any* way during an extended period of time, currently set to 30 seconds, the `PDFDocumentProxy.cleanup` method will never be called and font resources will thus not be cleaned-up.
This is necessary in order to support cases where the default viewer is embedded in a *dynamically* created <iframe> element.
In order to also support a use-case where there's *multiple* <iframe> elements (containing default viewers) on the same page, the "webviewerloaded" event now includes a `source` detail parameter such that it's possible to associate an event with the relevant DOM element.
As evident from the code, `PageViewport` only supports[1] `rotation` values which are a multiple of 90 degrees. Besides it being somewhat difficult to imagine meaningful use-cases for a non-multiple of 90 degrees `rotation`, the code also becomes both simpler and more efficient by not having to consider arbitrary `rotation` values.
However, any invalid rotation will *silently* fallback to assume zero `rotation` which probably isn't great for e.g. `PDFPageProxy.getViewport` in the API. Hence this patch, which will now enforce that only valid `rotation` values are accepted.
---
[1] As far as I can tell, from looking through the history, nothing else has ever been supported either.
Hopefully we don't need the *latest* Node.js releases for the unit-tests to work on Travis, and by using "long-term support" releases instead we should be able to avoid these types of sudden failures in the future as well.
Change the "download" keyboard shortcut (Ctrl+S) handling, in GENERIC/CHROME builds, to utilize the `EventBus` (issue 11657); add a new "openfile" keyboard shortcut (Ctrl+O), in GENERIC builds
Somewhat surprisingly, despite the GENERIC viewer implementing "openfile" support, there's never been a keyboard shortcut available. Similar to the previous patch, this utilizes the `EventBus` for consistency with the `Toolbar`/`SecondaryToolbar` buttons.
*Please note:* This patch should NOT be construed as carte blanche to simply convert all of the code in `webViewerKeyDown`, or elsewhere, to make use of the `EventBus` instead of direct function calls.
Any further changes, along the lines in this patch, would need to be evaluated on a case-by-case basis to determine if they are actually wanted, given that many/most existing cases in `webViewerKeyDown` should already be *indirectly* observable through the `EventBus` instance.
This improves the consistency of the "download" handling, in the default viewer, such that the `Toolbar`/`SecondaryToolbar` buttons *and* the keyboard shortcut are now handled in the same way (using the `EventBus`).
Given that the "download" keyboard shortcut handling is limited to GENERIC/CHROME builds and that the issue does raise a valid point about only being able to observe *some* downloads, these changes seem acceptable in this particular case.
Finally the pre-processor condition is adjusted to *explicitly*, rather than implicitly, list the affected build targets.
*Please note:* This patch should NOT be construed as carte blanche to simply convert all of the code in `webViewerKeyDown`, or elsewhere, to make use of the `EventBus` instead of direct function calls.
Any further changes, along the lines in this patch, would need to be evaluated on a case-by-case basis to determine if they are actually wanted, given that many/most existing cases in `webViewerKeyDown` should already be *indirectly* observable through the `EventBus` instance.
These two `AppOptions` are only defined in GENERIC builds, hence it's completely unnecessary to check them in the extension builds (e.g. MOZCENTRAL and CHROME).
Also, simply let the "printResolution" option be defined in all builds since it's being accessed in `web/firefox_print_service.js` as well.
Rather than creating a new `Stream` just to validate the OS/2 TrueType table, it's simpler/better to just pass in a reference to the font data and use that instead (similar to other TrueType helper functions).
There's a handful of cases in the code where the intention is simply to advance the `Stream` position, but rather than only doing that the code instead fetches/computes a Uint16 value (and without using the result for anything).
There's a handful of cases in the code where the intention is simply to advance the `Stream` position, but rather than only doing that the code instead fetches the bytes in question (and without using the result for anything).
*Please note:* These changes were done automatically, using the `gulp lint --fix` command.
This rule is already enabled in mozilla-central, see https://searchfox.org/mozilla-central/rev/567b68b8ff4b6d607ba34a6f1926873d21a7b4d7/tools/lint/eslint/eslint-plugin-mozilla/lib/configs/recommended.js#103-104
The main advantage, besides improved consistency, of this rule is that it reduces the size of the code (by 3 bytes for each case). In the PDF.js code-base there's close to 8000 instances being fixed by the `dot-notation` ESLint rule, which end up reducing the size of even the *built* files significantly; the total size of the `gulp mozcentral` build target changes from `3 247 456` to `3 224 278` bytes, which is a *reduction* of `23 178` bytes (or ~0.7%) for a completely mechanical change.
A large number of these changes affect the (large) lookup tables used on the worker-thread, but given that they are still initialized lazily I don't *think* that the new formatting this patch introduces should undo any of the improvements from PR 6915.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/dot-notation
What the user did:
Open the PDF Viewer in Chrome;
Mouse click into the “Page number” input field;
What they saw:
A pop-up list with seemingly random numbers
What you were expecting to see:
Nothing
What they see is the Chrome “Autofill” feature at work – that is suggesting values that you have previously entered into number fields in forms, as possible values you may want to enter into this field. The list has nothing to do with the PDF currently open but the user does not know this.
- Add a reduced test-case for issue 11768, to prevent future regressions.
(Given that PR 11769 is only a work-around, rather than a proper solution, it may not be entirely accurate for the issue to be closed as fixed.)
- Add more validation of the charCode, as found by the heuristics, in `PartialEvaluator._buildSimpleFontToUnicode` to prevent future issues.
Some of the code in `src/core/jpg.js` is fairly old, and has with time become unnecessary when the surrounding code has been updated to handle various types of JPEG corruption.
In particular the `if (!marker || marker <= 0xff00) { ... }` branch is now dead code, since:
- The `!marker` case can no longer happen, since we would already have broken out of the loop thanks to the `!fileMarker` branch a handful of lines above.
- The `marker <= 0xff00` case can also no longer happen, since the `findNextFileMarker` function validate markers much more thoroughly (by checking `marker >= 0xffc0 && marker <= 0xfffe`). Hence we'd again have broken out of the loop via the `!fileMarker` branch above when no valid marker was found.
This patch fixes yet another instalment in the never-ending series of "what the *bleep* was I thinking", by changing the `PDFDocumentProxy.getViewerPreferences` method to return `null` by default.
Not only is this method now consistent with many other API methods, for the data not present case, but it also avoids having to e.g. loop through an object to check if it's actually empty (note the old unit-test).
Please note that these changes were done automatically, using `gulp lint --fix`.
Given that the major version number was increased, there's a fair number of (primarily whitespace) changes; please see https://prettier.io/blog/2020/03/21/2.0.0.html
In order to reduce the size of these changes somewhat, this patch maintains the old "arrowParens" style for now (once mozilla-central updates Prettier we can simply choose the same formatting, assuming it will differ here).
- Since the Gitpod contributing work-flow is, from the PDF.js project perspective, completely unsupported we don't want to mention it in such a highly visible way as the README file.
- Since Gitpod is a commercial service, we probably want to avoid *potentially* being seen as advertising and/or endorsing it by mentioning it (somewhat prominently) in the PDF.js README file.
- By leaving the actual Gitpod files in the repository, for now, this should thus avoid outright breaking things for any existing users.
With two kind of builds now being produced, with/without translation/polyfills, it's unfortunately somewhat easy for users to accidentally pick the wrong one.
In the case where a user would attempt to use a modern build of PDF.js in an older browser, such as e.g. IE11, the failure would be immediate when the code is loaded (given the use of unsupported ECMAScript features).
However in some browsers/environments, a modern PDF.js build may load correctly and thus *appear* to function, only to fail for e.g. certain API calls. To hopefully lessen the support burden, and to try and improve things overall, this patch adds additional checks to ensure that a modern build of PDF.js cannot be used in browsers/environments which lack native support for `Promise.allSettled`.[1] Hence we'll fail early, with an error message telling users to pick an ES5-compatible build instead.
*Please note:* While it's probably too early to tell if this will be a widespread issue, it's possible that this is the sort of patch that *may* warrant being `git cherry-pick`ed onto the current beta version (v2.4.456).
---
[1] This was a fairly recent addition to the web platform, see https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Promise/allSettled#Browser_compatibility
*Please note:* Most of the necessary API work was done in PR 10033, and the only remaining thing to do here was to implement it in the viewer.
The new preference should thus allow e.g. enterprise users to disable copying in the viewer, for PDF documents whose permissions specify that.
In order to simplify things the "copy"-permission was implemented using CSS, as suggested in https://bugzilla.mozilla.org/show_bug.cgi?id=792816#c55, which should hopefully suffice.[1]
The advantage of this approach, as opposed to e.g. disabling the `textLayer` completely, is first of all that it ensures that searching still works correctly even in copy-protected documents. Secondly this also greatly simplifies the overall implementation, since it doesn't require a lot of code for something that's disabled by default.
---
[1] As the discussion in the bug shows, this kind of copy-protection is not very strong and is also generally easy to remove/circumvent in various ways. Hence a simple solution, targeting "regular"-users rather than "power"-users is hopefully deemed acceptable here.
At this point in time, compared to when the "ignore single-char" code was added, we *should* generally be doing a much better job of combining text into as few chunks as possible.
However, there's still bad cases where we're not able to combine text as much as one would like, which is why I'm *not* proposing to simply measure/scale all text. Instead this patch will to only measure/scale single-char text in cases where the horizontal/vertical scale is off significantly, since that's were you'd expect bad text-selection behaviour otherwise.
Note that most of the movement caused by this patch is with Type3 fonts, which is a somewhat special font type and one where our current text-selection behaviour is probably the least good.
[api-minor] Change `Font.exportData` to, by default, stop exporting properties which are completely unused on the main-thread and/or in the API (PR 11773 follow-up)
For years now, the `Font.exportData` method has (because of its previous implementation) been exporting many properties despite them being completely unused on the main-thread and/or in the API.
This is unfortunate, since among those properties there's a number of potentially very large data-structures, containing e.g. Arrays and Objects, which thus have to be first structured cloned and then stored on the main-thread.
With the changes in this patch, we'll thus by default save memory for *every* `Font` instance created (there can be a lot in longer documents). The memory savings obviously depends a lot on the actual font data, but some approximate figures are: For non-embedded fonts it can save a couple of kilobytes, for simple embedded fonts a handful of kilobytes, and for composite fonts the size of this auxiliary can even be larger than the actual font program itself.
All-in-all, there's no good reason to keep exporting these properties by default when they're unused. However, since we cannot be sure that every property is unused in custom implementations of the PDF.js library, this patch adds a new `getDocument` option (named `fontExtraProperties`) that still allows access to the following properties:
- "cMap": An internal data structure, only used with composite fonts and never really intended to be exposed on the main-thread and/or in the API.
Note also that the `CMap`/`IdentityCMap` classes are a lot more complex than simple Objects, but only their "internal" properties survive the structured cloning used to send data to the main-thread. Given that CMaps can often be *very* large, not exporting them can also save a fair bit of memory.
- "defaultEncoding": An internal property used with simple fonts, and used when building the glyph mapping on the worker-thread. Considering how complex that topic is, and given that not all font types are handled identically, exposing this on the main-thread and/or in the API most likely isn't useful.
- "differences": An internal property used with simple fonts, and used when building the glyph mapping on the worker-thread. Considering how complex that topic is, and given that not all font types are handled identically, exposing this on the main-thread and/or in the API most likely isn't useful.
- "isSymbolicFont": An internal property, used during font parsing and building of the glyph mapping on the worker-thread.
- "seacMap": An internal map, only potentially used with *some* Type1/CFF fonts and never intended to be exposed in the API. The existing `Font.{charToGlyph, charToGlyphs}` functionality already takes this data into account when handling text.
- "toFontChar": The glyph map, necessary for mapping characters to glyphs in the font, which is built upon the various encoding information contained in the font dictionary and/or font program. This is not directly used on the main-thread and/or in the API.
- "toUnicode": The unicode map, necessary for text-extraction to work correctly, which is built upon the ToUnicode/CMap information contained in the font dictionary, but not directly used on the main-thread and/or in the API.
- "vmetrics": An array of width data used with fonts which are composite *and* vertical, but not directly used on the main-thread and/or in the API.
- "widths": An array of width data used with most fonts, but not directly used on the main-thread and/or in the API.
By slicing the Uint8Array directly, rather than using the prototype and a `call` invocation, the runtime of `decryptAscii` is decreased slightly (~30% based on quick logging).
The `decryptAscii` function is still less efficient than `decrypt`, however ASCII encoded Type1 font programs are sufficiently rare that it probably doesn't matter much (we've only seen *two* examples, issue 4630 and 11740).
The PDF document, in the referenced issue, actually contains ASCII-encoded Type1 data which we currently *incorrectly* identify as binary.
According to the specification, see https://www-cdf.fnal.gov/offline/PostScript/T1_SPEC.PDF#[{%22num%22%3A203%2C%22gen%22%3A0}%2C{%22name%22%3A%22XYZ%22}%2C87%2C452%2Cnull], the current checks are insufficient to decide between binary/ASCII encoded Type1 font programs.
In preparation for the next patch, this changes the signature of `TranslatedFont` to take an object rather than individual parameters. This also, in my opinion, makes the call-sites easier to read since it essentially provides a small bit of documentation of the arguments.
Finally, since it was necessary to touch `TranslatedFont` anyway it seemed like a good idea to also convert it to a proper `class`.
Given that the `vertical` property is always accessed on the main-thread, ensuring that the property is explicitly defined seems like the correct thing to do since it also avoids boolean casting elsewhere in the code-base.
With `Font.exportData` now only exporting white-listed properties, there should no longer be any reason to not use standard shadowing in the `Font.spaceWidth` method.
Furthermore, considering the amount of other changes to the code-base over the years it's not even clear to me that the special-case was necessary any more (regardless of the preceding patches).
A number of *internal* font properties, which only make sense on the worker-thread, were previously exported. Some of these properties could also contain potentially large Arrays/Objects, which thus unnecessarily increases memory usage since we're forced to copy these to the main-thread and also store them there.
This patch stops exporting the following font properties:
- "_shadowWidth": An internal property, which was never intended to be exported.
- "charsCache": An internal cache, which was never intended to be exported and doesn't make any sense on the main-thread. Furthermore, by the time `Font.exportData` is called it's usually `undefined` or a mostly empty Object as well.
- "cidEncoding": An internal property used with (some) composite fonts.
As can be seen in the `PartialEvaluator.translateFont` method, `cidEncoding` will only be assigned a value when the font dictionary has an "Encoding" entry which is a `Name` (and not in the `Stream` case, since those obviously cannot be cloned).
All-in-all this property doesn't really make sense on the main-thread and/or in the API, and note also that the resulting `cMap` property is (partially) available already.
- "fallbackToUnicode": An internal map, part of the heuristics used to improve text-selection in (some) badly generated PDF documents with simple fonts. This was never intended to be exposed on the main-thread and/or in the API.
- "glyphCache": An internal cache, which was never intended to be exported and which doesn't make any sense on the main-thread. Furthermore, by the time `Font.exportData` is called it's usually a mostly empty Object as well.
- "isOpenType": An internal property, used only during font parsing on the worker-thread. In the *very* unlikely event that an API consumer actually needs that information, then `fontType` should be a (generally) much better property to use.
Finally, in the (hopefully) unlikely event that any of these properties become necessary on the main-thread, re-adding them to the white-list is easy to do.
This patch addresses an existing, and very long standing, TODO in the code such that it's no longer possible to send arbitrary/unnecessary font properties to the main-thread.
Furthermore, by having a white-list it's also very easy to see *exactly* which font properties are being exported.
Please note that in its current form, the list of exported properties contains *every* possible enumerable property that may exist in a `Font` instance.
In practice no single font will contain *all* of these properties, and e.g. embedded/non-embedded/Type3 fonts will all differ slightly with respect to what properties are being defined. Hence why only explicitly set properties are included in the exported data, to avoid half of them being `undefined`, which however should not be a problem for any existing consumer (since they'd already need to handle those cases).
Since a fair number of these font properties are completely *internal* functionality, and doesn't make any sense to expose on the main-thread and/or in the API, follow-up patch(es) will be required to trim down the list. (I purposely included all properties here for brevity and future documentation purposes.)
- Use template strings when printing document/viewer information in `_initializeMetadata`, since the old format feels overly verbose.
Also, get the WebGL state from the `BaseViewer` instance[1] rather than the `AppOptions`. Since the `AppOptions` value could theoretically have been changed (by the user) after the viewer components were initialized, it seems much more useful to print the *actual* value that'll be used during rendering.
- Change `_initializePdfHistory` to actually do the "is embedded"-check first, in accordance with the comment and given that the "disableHistory" option usually shouldn't be set.
---
[1] Admittedly reaching into the `BaseViewer` instance and just grabbing the value perhaps isn't a great approach overall, but given that the WebGL-backend isn't even on by default this probably doesn't matter too much.
Over time, with more and more API-functionality added, the `PDFViewerApplication.load` method has become quite large and complex. In an attempt to improve the current situation somewhat, this patch moves the fetching and initialization of "auto print" out into its own (private) helper method instead.
Over time, with more and more API-functionality added, the `PDFViewerApplication.load` method has become quite large and complex. In an attempt to improve the current situation somewhat, this patch moves the fetching and initialization of "metadata" out into its own (private) helper method instead.
Over time, with more and more API-functionality added, the `PDFViewerApplication.load` method has become quite large and complex. In an attempt to improve the current situation somewhat, this patch moves the fetching and initialization of "page labels" out into its own (private) helper method instead.
I completely overlooked the fact that we had *one* occurrence of an asynchronous `mkdirp` call in the gulpfile, which thus breaks since the package now uses Promises rather than a callback function; sorry about that!
*Please note:* This patch on its own is *not* sufficient to address the underlying problem in the referenced issue, hence why no test-case is included since the *actual* bug still needs to be fixed.
As can be seen in the specification, https://tc39.es/ecma262/#sec-string.fromcodepoint, `String.fromCodePoint` will throw a RangeError for invalid code points.
In the event that a CMap, in a composite font, contains invalid data and/or we fail to parse it correctly, it's thus possible that the glyph mapping that we build end up with entires that cause `String.fromCodePoint` to throw and thus `Font.charToGlyph` to break.
If that happens, as is the case in issue 11768, significant portions of a page/document may fail to render which seems very unfortunate.
While this patch doesn't fix the underlying problem, it's hopefully deemed useful not only for the referenced issue but also to prevent similar bugs in the future.
With two kind of builds now being produced, with/without translation/polyfills, it's unfortunately somewhat easy for users to accidentally pick the wrong one.
In the case where a user would attempt to use a modern build of PDF.js in an older browser, such as e.g. IE11, the failure would be immediate when the code is loaded (given the use of unsupported ECMAScript features).
However in some browsers/environments, in particular Node.js, a modern PDF.js build may load correctly and thus *appear* to function, only to fail for e.g. certain API calls. To hopefully lessen the support burden, and to try and improve things overall, this patch adds checks to ensure that a modern build of PDF.js cannot be used in browsers/environments which lack native support for critical functionality (such as e.g. `ReadableStream`). Hence we'll fail early, with an error message telling users to pick an ES5-compatible build instead.
To ensure that we actually test things better especially w.r.t. usage of the PDF.js library in Node.js environments, the `gulp npm-test` task as used by Node.js/Travis was changed (back) to test an ES5-compatible build.
(Since the bots still test the code as-is, without transpilation/polyfills, this shouldn't really be a problem as far as I can tell.)
As part of these changes there's now both `gulp lib` and `gulp lib-es5` build targets, similar to e.g. the generic builds, which thanks to some re-factoring only required adding a small amount of code.
*Please note:* While it's probably too early to tell if this will be a widespread issue, it's possible that this is the sort of patch that *may* warrant being `git cherry-pick`ed onto the current beta version (v2.4.456).
This avoids unnecessary duplication of many images, thus reducing the size of PDF.js image resources slightly.
Note that since the images should only be flipped horizontally, this required specifying the horizontal/vertical scaling separately for the hiDPI-images.
This functionality was only added to the default viewer for backwards compatibility and to support the various PDF viewer tests in mozilla-central, with the intention to eventually remove it completely.
While the different mozilla-central tests cannot be *easily* converted from DOM events, it's however possible to limit that functionality to only MOZCENTRAL builds *and* when tests are running.
Rather than depending of the re-dispatching of internal events to the DOM, the default viewer can instead be used in e.g. the following way:
```javascript
document.addEventListener("webviewerloaded", function() {
PDFViewerApplication.initializedPromise.then(function() {
// The viewer has now been initialized, and its properties can be accessed.
PDFViewerApplication.eventBus.on("pagerendered", function(event) {
console.log("Has rendered page number: " + event.pageNumber);
});
});
});
```
The correct/intended way of working with the "viewer components" is by providing an `EventBus` instance upon initialization, and the `getGlobalEventBus` was only added for backwards compatibility.
Note, for example, that using `getGlobalEventBus` doesn't really work at all well with a use-case where there's *multiple* `PDFViewer` instances on a one page, since it may then be difficult/impossible to tell which viewer a particular event originated from.
All of the "viewer components" examples have been previously updated, such that there's no longer any code/examples which relies on the now removed `getGlobalEventBus` functionality.
The `sizes` property doesn't appear to have been used ever since the code was first split into main/worker-threads, which is so many years ago that I wasn't able to easily find exactly in which PR/commit it became unused.
The `encoding` property is always assigned the `properties.baseEncoding` value, however the `PartialEvaluator` doesn't actually compute/set that value any more. Again it was difficult to determine when it became unused, but it's been that way for years.
This looks entirely like something which was left-over from debugging, and that line hasn't been touched since PR 4515, especially considering that the corresponding branch in `FirefoxBrowser` doesn't print anything.
Given the way that "classes" were previously implemented in PDF.js, using regular functions and closures, there's a fair number of false positives when the `no-shadow` ESLint rule was enabled.
Note that while *some* of these `eslint-disable` statements can be removed if/when the relevant code is converted to proper `class`es, we'll probably never be able to get rid of all of them given our naming/coding conventions (however I don't really see this being a problem).
This rule is *not* currently enabled in mozilla-central, but it appears commented out[1] in the ESLint definition file; see https://searchfox.org/mozilla-central/rev/c80fa7258c935223fe319c5345b58eae85d4c6ae/tools/lint/eslint/eslint-plugin-mozilla/lib/configs/recommended.js#238-239
Unfortunately this rule is, for fairly obvious reasons, impossible to `--fix` automatically (even partially) and each case thus required careful manual analysis.
Hence this ESLint rule is, by some margin, probably the most difficult one that we've enabled thus far. However, using this rule does seem like a good idea in general since allowing variable shadowing could lead to subtle (and difficult to find) bugs or at the very least confusing code.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/no-shadow
---
[1] Most likely, a very large number of lint errors have prevented this rule from being enabled thus far.
Fixes#11718 in which the `ff` ligature glyph is at index zero in a CFF font. Beacuse this is a CIDFont, glyph names are CIDs, which are integers. Thus the string `".notdef"` is not correct. The rest of the charset data is already parsed correctly as integers when the boolean argument `cid` is true.
*This is part of a series of patches that will try to split PR 11566 into smaller chunks, to make reviewing more feasible.*
Once all the code has been fixed, we'll be able to eventually enable the ESLint no-shadow rule; see https://eslint.org/docs/rules/no-shadow
*This is part of a series of patches that will try to split PR 11566 into smaller chunks, to make reviewing more feasible.*
Once all the code has been fixed, we'll be able to eventually enable the ESLint no-shadow rule; see https://eslint.org/docs/rules/no-shadow
Please note that the `setPDFNetworkStreamFactory` functionality isn't exposed in the public API, i.e. not listed among the exports in the `src/pdf.js` file, and that even if it were it wouldn't really be useful considering that none of the `PDFNetworkStream`/`PDFFetchStream`/`PDFNodeStream` classes are exported either.
These methods were deprecated already in PDF.js version `2.1.266`, see PRs 10246 and 10369, and were converted to throw `Error`s upon invocation in PDF.js version `2.4.456`, see PR 11219.
Hence it ought to be possible to remove these methods now.
Given that none of the relevant options are marked as optional in the code/JSDocs, and that the `PDFDocumentProperties` class is specific to the default viewer (and not exposed as part of the viewer components), there's no good reason as far as I can tell for these checks.
The `BaseViewer.setDocument` method in particular is necessary for rendering to start when the viewer loads, hence you obviously want that to happen as soon as possible and without any unnecessary delays.
Unfortunately *some* API calls need to be done before that, note existing comments, however the `ViewHistory` initialization (and subsequent fetching of data) in particular can be moved slightly without any adverse effects.
As part of testing I've used logging with `performance.now()` inserted in various parts of this code, and there's *obviously* no discernible changes between `master` and this patch for e.g. rendering starting in the viewer.
*Note:* The vast majority of this patch is simple indentation changes, which were forced by Prettier (and done automatically with `gulp lint --fix`).
- Ensure that `database.files` actually contains an Array, rather than some arbitrary data.
- Only try to lookup an existing entry when the `database` existed on load, since there's obviously nothing to find when `database.files = []` was set (this case is very common in the MOZCENTRAL build since `sessionStorage` is being used there).
*This is part of a series of patches that will try to split PR 11566 into smaller chunks, to make reviewing more feasible.*
Once all the code has been fixed, we'll be able to eventually enable the ESLint no-shadow rule; see https://eslint.org/docs/rules/no-shadow
When JPEG images are decoded by the browser, on the main-thread, there's a handful of short-lived copies of the image data; see c3f4690bde/src/display/api.js (L2364-L2408)
That code thus becomes quite problematic for very big JPEG images, since it increases peak memory usage a lot during decoding. In the referenced issue there's a couple of JPEG images whose dimensions are `10006 x 7088` (i.e. ~68 mega-pixels), which causes the *peak* memory usage to increase by close to `1 GB` (i.e. one giga-byte) in my testing.
By letting the PDF.js JPEG decoder, rather than the browser, handle very large images the *peak* memory usage is considerably reduced and the allocated memory also seem to be reclaimed faster.
*Please note:* This will lead to movement in some existing `eq` tests.
Letting Travis update npm packages can lead to unexpected and completely unrelated failures for any PR and/or merge, see e.g. 11719, if there's ever backwards *incompatible* changes when a package is updated.
The *exact* package versions are specified in `package-lock.json`, and they should thus be used when running tests. Note that the bots won't update npm packages, and neither should Travis as far as I'm concerned.
This property has never been documented and/or *intentionally* exposed through the API, instead the `PDFPageProxy.pageNumber` property is the documented/intended API to use here.
Hence pageIndex is changed to a "private" property on `PDFPageProxy` instances, and internal API functionality is also updated to *consistently* use `this._pageIndex` rather than a mix of formats.
The viewer doesn't currently support executing of any JavaScript, as found in some PDF documents, for security reasons. (Although there's a small hack to at least provide basic support for automatic printing on document load, without running scripts.)
However, in the event that the browser doesn't support printing we're not run *any* of this code. In particular that means that we're also not displaying the "Warning: JavaScript is not supported" message, which seems strange since JavaScript found in a PDF document can really contain *anything* (and not only printing instructions).
It thus seem reasonable, as far as I'm concerned, to always display the JavaScript-warning *even* if printing happens to be unsupported.
Don't accidentally accept invalid glyphNames which *appear* to follow the Cdd{d}/cdd{d} format in `PartialEvaluator._buildSimpleFontToUnicode` (issue 11697)
The /Differences array of the problematic font contains a `/c.1` entry, which is consequently detected as a *possible* Cdd{d}/cdd{d} glyphName by the existing heuristics.
Because of how the base 10 conversion is implemented, which is necessary for the base 16 special case, the parsed charCode becomes `0.1` thus causing `String.fromCodePoint` to throw since that obviously isn't a valid code point.
To fix the referenced issue, and to hopefully prevent similar ones in the future, the patch adds *additional* validation of the charCode found by the heuristics.
*This is part of a series of patches that will try to split PR 11566 into smaller chunks, to make reviewing more feasible.*
Once all the code has been fixed, we'll be able to eventually enable the ESLint `no-shadow` rule; see https://eslint.org/docs/rules/no-shadow
Trying to enable the ESLint rule `no-shadow`, against the `master` branch, would result in a fair number of errors in the `Glyph` class in `src/core/fonts.js`.
Since the glyphs are exposed through the API, we can't very well change the `isSpace` property on `Glyph` instances. Thus the best approach seems, at least to me, to simply rename the `isSpace` helper function to `isWhiteSpace` which shouldn't cause any issues given that it's only used in the `src/core/` folder.
Rather than duplicating the lookup and caching in multiple files, it seems easier to simply move all of this functionality into `src/shared/util.js` instead.
This will also help avoid a bunch of ESLint errors once the `no-shadow` rule is eventually enabled.
While Firefox originally used transition effects for browser UI toolbar buttons, that was removed years ago in https://bugzilla.mozilla.org/show_bug.cgi?id=1393057
Since the PDF.js viewer toolbar transitions were likely based on the Firefox ones, it seems reasonable that these transition effects are removed from PDF.js as well. Besides removing a bunch of CSS, this also makes the toolbar feel ever so slightly more "snappy" without these delays on mouse interaction.
(In order to make it more feasible to modernize/improve the viewer UI, trying to clean-up/simplify existing rules iteratively seems like the most reasonble way to make any progress here w.r.t. being able to test/review things.)
When reftest analyzer shows magnified pixels, there is a seemingly random offset between the mouse position and the magnified position. The reason for this is that reftest analyzer assumes all images have 800 * 1000 pixels but actually the test images have varying sizes.
*This patch fixes something that's annoyed me every now and then over the years, when debugging/fixing corrupt PDF documents.*
For corrupt PDF documents where `Lexer.getHexString` encounters invalid characters, there's very rarely just a handful of them. In practice it's not uncommon for there to be many hundreds, or even many thousands, invalid hex characters found.
Not only is the resulting console warning spam utterly useless in these cases, there's often enough of it that performance may even suffer; hence this patch which limits the amount of messages that any one `Lexer.getHexString` invocation may print.
The PDF document in question is *corrupt*, since it contains an XObject with a truncated dictionary and where the stream contents start without a "stream" operator.
For reasons that I don't even pretend to understand, the `textLayerFactory` property is determined for *every single* page in the PDF document.
Given that the `TextLayerMode` should be consistent for *all* pages in a document, we obviously could/should define `textLayerFactory` just once instead.
There's a couple of issues with this functionality:
- The respective `PromiseCapability` instances are not being reset, in `BaseViewer._resetView`, when the document is closed which is inconsistent with all other state.
- While the default viewer depends on these promises, and they thus ought to be considered part of e.g. the `PDFViewer` API-surface, they're not really defined in a particularily user-visible way (being that they're attached to the `BaseViewer` instance *inline* in `BaseViewer.setDocument`).
- There's some internal `BaseViewer` state, e.g. `BaseViewer._pageViewsReady`, which is tracked manually and could instead be tracked indirectly via the relevant `PromiseCapability`, thus reducing the need to track state *twice* since that's always best to avoid.
*Please note:* In the existing implementation, these promises are not defined *until* the `BaseViewer.setDocument` method has been called.
While it would've been simple to lift that restriction in this patch, I'm purposely choosing *not* to do so since this ensures that any Promise handlers added inside of `BaseViewer.setDocument` are always invoked *before* any external ones (and keeping that behaviour seems generally reasonable).
The only reason for the `return undefined;` lines was to appease the ESLint `consistent-return` rule, but that's not actually necessary if you make use of the fact that the method is `async` and that we can thus await the Promise rather than returning it.
Note that `Dict.set` will only be called with values returned through `Parser.getObj`, and thus indirectly via `Lexer.getObj`. Since neither of those methods will ever return `undefined`, we can simply assert that that's the case when inserting data into the `Dict` and thus get rid of `in` checks when doing the data lookups.
In this case, since `Dict.set` is fairly hot, the patch utilizes an *inline check* and when necessary a direct call to `unreachable` to not affect performance of `gulp server/test` too much (rather than always just calling `assert`).
For very large and complex PDF files this will help performance *slightly*, since `Dict.{get, getAsync, has}` is called *a lot* during parsing in the worker.
This patch was tested using the PDF file from issue 2618, i.e. http://bugzilla-attachments.gnome.org/attachment.cgi?id=226471, with the following manifest file:
```
[
{ "id": "issue2618",
"file": "../web/pdfs/issue2618.pdf",
"md5": "",
"rounds": 250,
"type": "eq"
}
]
```
which gave the following results when comparing this patch against the `master` branch:
```
-- Grouped By browser, stat --
browser | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ------------ | ----- | ------------ | ----------- | --- | ----- | -------------
Firefox | Overall | 250 | 2838 | 2820 | -18 | -0.65 | faster
Firefox | Page Request | 250 | 1 | 2 | 0 | 11.92 | slower
Firefox | Rendering | 250 | 2837 | 2818 | -19 | -0.65 | faster
```
This patch deprecates the existing `getOpenActionDestination` API method, in favor of a better and more general `getOpenAction` method instead. (For now JavaScript actions, related to printing, are still handled as before.)
By clearly separating "regular" Print actions from the JavaScript handling, it's thus possible to get rid of the somewhat annoying and strictly incorrect warning when the viewer loads.
It occured to me that similar to the `getGlobalEventBus` function, it's probably a good idea to *also* notify users of the fact that `eventBusDispatchToDOM` is now deprecated.
Rather than depending of the re-dispatching of internal events to the DOM, the default viewer can instead be used in e.g. the following way:
```javascript
document.addEventListener("webviewerloaded", function() {
PDFViewerApplication.initializedPromise.then(function() {
// The viewer has now been initialized, and its properties can be accessed.
PDFViewerApplication.eventBus.on("pagerendered", function(event) {
console.log("Has rendered page number: " + event.pageNumber);
});
});
});
```
I overlooked this in PR 11631; sorry about that!
Also, ensure that `EventBus` instances *always* track "external" events using a boolean regardless of the actual option value.
This changes the dropdown icon from being set using the `background` CSS property, to being set with `::after` which is *similar* to all the other toolbar button icons (which use `::before`).
Also tweaks the dropdown `background-color` on `:hover` slightly, since the other changes made it too light.
Fixes#11477
The PDF draws many space characters but the embedded fonts don't have a glyph named `space`, so `.notdef` should be drawn instead. PDF.js assumed that Type1 fonts define `.notdef` as the first glyph (index 0). However, now the fonts have the glyph `A` at index 0 and `.notdef` is the last one, so `A` appears where spaces are expected.
Because the rest of the font machinery in `core/fonts.js` assumes `.notdef` is at index zero, it's easiest to modify `core/type1_parser.js` so that it "repairs" fonts and makes sure `.notdef` is at index 0.
The PDF document in question is *corrupt*, since it contains multiple instances of incorrect operators.
We obviously don't want to slow down parsing of *all* documents (since most are valid), just to accommodate a particular bad PDF generator, hence the reason for the inline check before calling the `ensureStateFont` method.
It's no longer necessary to special-case this getter in the `GenericFontLoader` case, since the GENERIC build hasn't been using `mozPrintCallback` for years now (furthermore Firefox 63 is really old as well).
Add a `ignoreDestinationZoom` option/preference to allow users to preserve the current zoom level when navigating to internal destinations (issue 5064, 11606)
Since the goal has always been, essentially since the `EventBus` abstraction was added, to remove all dispatching of DOM events[1] from the viewer components this patch tries to address one thing that came up when updating the examples:
The DOM events are always dispatched last, and it's thus guaranteed that all internal event listeners have been invoked first.
However, there's no such guarantees with the general `EventBus` functionality and the order in which event listeners are invoked is *not* specified. With the promotion of the `EventBus` in the examples, over DOM events, it seems like a good idea to at least *try* to keep this ordering invariant[2] intact.
Obviously this won't prevent anyone from manually calling the new *internal* viewer component methods on the `EventBus`, but hopefully that won't be too common since any existing third-party code would obviously use the `on`/`off` methods and that all of the examples shows the *correct* usage (which should be similarily documented on the "Third party viewer usage" Wiki-page).
---
[1] Looking at the various Firefox-tests, I'm not sure that it'll be possible to (easily) re-write all of them to not rely on DOM events (since getting access to `PDFViewerApplication` might be generally difficult/messy depending on scopes).
In any case, even if technically feasible, it would most likely add *a lot* of complication that may not be desireable in the various Firefox-tests. All-in-all, I'd be fine with keeping the DOM events only for the `MOZCENTRAL` target and gated on `Cu.isInAutomation` (or similar) rather than a preference.
[2] I wouldn't expect any *real* bugs in a custom implementation, simply based on event ordering, but it nonetheless seem like a good idea if any "external" events are still handled last.
To avoid outright breaking third-party usages of the "viewer components" the `getGlobalEventBus` functionality is left intact, but a deprecation message is printed if the function is invoked.
The various examples are updated to *explicitly* initialize an `EventBus` instance, and provide that when initializing the relevant viewer components.
This patch extends the existing heuristics, which are really the best that we can do in general for these kinds of non-embedded *and* non-standard fonts.
Furthermore, this patch also tries to improve the copy-and-paste behaviour for non-embedded Wingdings fonts by also using the `ZapfDingbatsEncoding` in this case.
*Note:* I'm not sure that adding additional tests for Wingdings fonts matters that much, given how limited our "support" for them really is.
Given that all of these primitives implement caching, to avoid unnecessarily duplicating those objects *a lot* during parsing, it would thus be good to actually enforce usage of `Cmd.get()`/`Name.get()`/`Ref.get()` in the code-base.
Luckily it turns out that there's an ESLint rule, which is fairly easy to use, that can be used to disallow arbitrary JavaScript syntax.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/no-restricted-syntax
*This whole patch feels somewhat arbitrary, and I'd be slightly worried about possibly breaking something else.*
To limit the impact of these changes, we only re-parse JPEG images using a reduced `scanLines` value if and only if: An unexpected EOI (End of Image) marker was encountered during decoding of Scan data *and* the "actual" `scanLines` value is at least one order of magnitude smaller than expected.
In some cases PDF documents can contain JPEG images that the native browser decoder cannot handle, e.g. images with DNL (Define Number of Lines) markers or images where the SOF (Start of Frame) marker contains a wildly incorrect `scanLines` parameter.
Currently, for "simple" JPEG images, we're relying on native image decoding to *fail* before falling back to the implementation in `src/core/jpg.js`. In some cases, note e.g. issue 10880, the native image decoder doesn't outright fail and thus some images may not render.
In an attempt to improve the current situation, this patch adds additional validation of the JPEG image SOF data to force the use of `src/core/jpg.js` directly in cases where the native JPEG decoder cannot be trusted to do the right thing.
The only way to implement this is unfortunately to parse the *beginning* of the JPEG image data, looking for a SOF marker. To limit the impact of this extra parsing, the result is cached on the `JpegStream` instance and this code is only run for images which passed all of the pre-existing "can the JPEG image be natively rendered and/or decoded" checks.
---
*Slightly off-topic:* Working on this *really* makes me start questioning if native rendering/decoding of JPEG images is actually a good idea.
There's certain kinds of JPEG images not supported natively, and all of the validation which is now necessary isn't "free". At this point, in the `NativeImageDecoder`, we're having to check for certain properties in the image dictionary, parse the `ColorSpace`, and finally read the actual image data to find the SOF marker.
Furthermore, we cannot just send the image to the main-thread and be done in the "JpegStream" case, but we also need to wait for rendering to complete (or fail) before continuing with other parsing.
In the "JpegDecode" case we're even having to parse part of the image on the main-thread, which seems completely at odds with the principle of doing all heavy parsing in the Worker, and there's also a couple of potentially large (temporary) allocations/copies of TypedArray data involved as well.
Given that this is completely unused, and that a "normal" function call may be a *tiny* bit more efficient, there's no good reason as far as I can tell to keep it.
Rather than adding two buttons, it seems easier to simply link to the relevant section of the README instead (since it also means fewer things to keep up-to-date).
This complements the existing `PDFViewerApplication.initialized` boolean property, and may be helpful for custom implementations of the default viewer. This will thus provide users of the default viewer an alternative to setting the preference to dispatch events to the DOM (and listen for the "localized" event), since they can instead use:
```javascript
document.addEventListener("webviewerloaded", function() {
PDFViewerApplication.initializedPromise.then(function() {
// The viewer has now been initialized.
})
});
```
Note that in order to avoid manually tracking the initialization state *twice*, this implementation purposely uses the `PromiseCapability` functionality to handle both `PDFViewerApplication.initialized` and `PDFViewerApplication.initializedPromise` internally.
*I've had this patch locally for awhile, but have apparently missed to upstream it.*
This simplifies enabling of new ESLint rules, since most of them support automatic fixing of errors, without having to edit `gulpfile.js` or manually invoke ESLint directly.
With the way that the `default_preferences.json` file is now generated at build time, the `gulp lint` task is now noticeably slower than before. This slowdown has been, and still is, somewhat annoying during the deployment of new ESLint rules.
Hence this patch, which moves the `chromium/preferences_schema.json` validation from `gulp lint` and into a new `gulp lint-chromium` task instead. *Obviously* this new task is run as part of the `gulp npm-test` task, and thus through `npm test` on Node.js/Travis, such that it's still being tested as before.
Please note that these changes do *not* affect the *public* interface of the `Metadata` class, but only touches internal structures.[1]
These changes were prompted by looking at the `getAll` method, which simply returns the "private" metadata object to the consumer. This seems wrong conceptually, since it allows way too easy/accidental changes to the internal parsed metadata.
As part of fixing this, the internal metadata was changed to use a `Map` rather than a plain Object.
---
[1] Basically, we shouldn't need to worry about someone depending on internal implementation details.
- Remove the "capturing group" in the regular expression that removes leading "junk" from the raw metadata, since it's not necessary here (it's simply a case of too much copy-pasting in a prior patch).
According to [MDN](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Guide/Regular_Expressions/Cheatsheet#Groups_and_ranges) you want to, for performance reasons, avoid "capturing groups" unless actually needed.
- Add inline comments to document a bunch of magic values in the code.
Given that all `TypedArray` polyfills were removed in PDF.js version `2.0`, since native support is now required, this branch has been dead code for awhile.
In practice it's extremely rare[1] for the padding to be zero in *all* components, hence it seems better to just set it directly rather than creating a temporary variable and checking for the "no padding"-case.
---
[1] In the `tracemonkey.pdf` file that only happens with `0.08%` of all text elements.
Over the years there's been a fair number of issues/PRs opened, where people have wanted to add `hasOwnProperty` checks in (hot) loops in the font parsing code. This has always been rejected, since we don't want to risk reducing performance in the Firefox PDF viewer simply because some users of the general PDF.js library are *incorrectly* extending the `Array.prototype` with enumerable properties.
With this patch the general PDF.js library will now fail immediately with a hopefully useful Error message, rather than having (some) fonts fail to render, when the `Array.prototype` is incorrectly extended.
Note that I did consider making this a warning, but ultimately decided against it since it's first of all possible to disable those (with the `verbosity` parameter). Secondly, even when printed, warnings can be easy to overlook and finally a warning may also *seem* OK to ignore (as opposed to an actual Error).
This patch contains some *much* needed clean-up of, and improvements to, this old code thus addressing a number of issues:
- Set more reasonable *default* values for the widths, in `web/viewer.css`, since the current ones are actually too small even for the (default) `en-US` locale.
This obviously result in a slightly larger zoom dropdown width for many locales, but the more consistent UI does look good to me.
- Stop setting the `min-width`/`max-width` and just use `width` instead.
- Set an explicit `height` of the zoom dropdown, in an attempt to get Google Chrome to display it with the same height as the toolbar buttons.
- Remove additional `Element.setAttribute("style", ...)` usage.
- Actually check *all* of the predefined l10n strings, since the old implementation (implicitly) assumed that the currently selected one was the longest (note e.g. the `ja-JP` locale where one string is considerably longer than the rest).
- Stop invalidating the DOM multiple times when doing the measurements. This was achieved by using a temporary in-memory `canvas`, and we now only need to query the DOM once in order to get the current font properties of the zoom dropdown.
This patch makes the following changes, to improve these API methods:
- Let `PDFPageProxy.cleanup` return a boolean indicating if clean-up actually happened, since ongoing rendering will block clean-up.
Besides being used in other parts of this patch, it seems that an API user may also be interested in the return value given that clean-up isn't *guaranteed* to happen.
- Let `PDFDocumentProxy.cleanup` return the promise indicating when clean-up is finished.
- Improve the JSDoc comment for `PDFDocumentProxy.cleanup` to mention that clean-up is triggered on *both* threads (without going into unnecessary specifics regarding what *exactly* said data actually is).
Add a note in the JSDoc comment about not calling this method when rendering is ongoing.
- Change `WorkerTransport.startCleanup` to throw an `Error` if it's called when rendering is ongoing, to prevent rendering from breaking.
Please note that this won't stop *worker-thread* clean-up from happening (since there's no general "something is rendering"-flag), however I'm not sure if that's really a problem; but please don't quote me on that :-)
All of the caches that's being cleared in `Catalog.cleanup`, on the worker-thread, *should* be re-filled automatically even if cleared *during* parsing/rendering, and the only thing that probably happens is that e.g. font data would have to be re-parsed.
On the main-thread, on the other hand, clearing the caches is more-or-less guaranteed to cause rendering errors, since the rendering code in `src/display/canvas.js` isn't able to re-request any image/font data that's suddenly being pulled out from under it.
- Last, but not least, add a couple of basic unit-tests for the clean-up functionality.
While it would be nice to change the `PDFFormatVersion` property, as returned through `PDFDocumentProxy.getMetadata`, to a number (rather than a string) that would unfortunately be a breaking API change.
However, it does seem like a good idea to at least *validate* the PDF header version on the worker-thread, rather than potentially returning an arbitrary string.
This should hopefully be useful in environments where restrictive CSPs are in effect.
In most cases the replacement is entirely straighforward, and there's only a couple of special cases:
- For the `src/display/font_loader.js` and `web/pdf_outline_viewer.js `cases, since the elements aren't appended to the document yet, it shouldn't matter if the style properties are set one-by-one rather than all at once.
- For the `web/debugger.js` case, there's really no need to set the `padding` inline at all and the definition was simply moved to `web/viewer.css` instead.
*Please note:* There's still *a single* case left, in `web/toolbar.js` for setting the width of the zoom dropdown, which is left intact for now.
The reasons are that this particular case shouldn't matter for users of the general PDF.js library, and that it'd make a lot more sense to just try and re-factor that very old code anyway (thus fixing the `setAttribute` usage in the process).
Having thought *briefly* about using `css-vars-ponyfill`, I'm no longer convinced that it'd be a good idea. The reason is that if we actually want to properly support CSS variables, then that functionality should be available in *all* of our CSS files.
Note in particular the `pdf_viewer.css` file that's built as part of the `COMPONENTS` target, in which case I really cannot see how a rewrite-at-the-client solution would ever be guaranteed to always work correctly and without accidentally touching other CSS in the surrounding application.
All-in-all, simply re-writing the CSS variables at build-time seems much easier and is thus the approach taken in this patch; courtesy of https://github.com/MadLittleMods/postcss-css-variables
By using its `preserve` option, the built files will thus include *both* a fallback and a modern `var(...)` format[1]. As a proof-of-concept this patch removes a couple of manually added fallback values, and converts an additional sidebar related property to use a CSS variable.
---
[1] Comparing the `master` branch with this patch, when using `gulp generic`, produces the following diff for the built `web/viewer.css` file:
```diff
@@ -408,6 +408,7 @@
:root {
--sidebar-width: 200px;
+ --sidebar-transition-duration: 200ms;
}
* {
@@ -550,27 +551,28 @@
position: absolute;
top: 32px;
bottom: 0;
- width: 200px; /* Here, and elsewhere below, keep the constant value for compatibility
- with older browsers that lack support for CSS variables. */
+ width: 200px;
width: var(--sidebar-width);
visibility: hidden;
z-index: 100;
border-top: 1px solid rgba(51, 51, 51, 1);
-webkit-transition-duration: 200ms;
transition-duration: 200ms;
+ -webkit-transition-duration: var(--sidebar-transition-duration);
+ transition-duration: var(--sidebar-transition-duration);
-webkit-transition-timing-function: ease;
transition-timing-function: ease;
}
html[dir='ltr'] #sidebarContainer {
-webkit-transition-property: left;
transition-property: left;
- left: -200px;
+ left: calc(-1 * 200px);
left: calc(-1 * var(--sidebar-width));
}
html[dir='rtl'] #sidebarContainer {
-webkit-transition-property: right;
transition-property: right;
- right: -200px;
+ right: calc(-1 * 200px);
right: calc(-1 * var(--sidebar-width));
}
@@ -640,6 +642,8 @@
#viewerContainer:not(.pdfPresentationMode) {
-webkit-transition-duration: 200ms;
transition-duration: 200ms;
+ -webkit-transition-duration: var(--sidebar-transition-duration);
+ transition-duration: var(--sidebar-transition-duration);
-webkit-transition-timing-function: ease;
transition-timing-function: ease;
}
```
Based on the PDF spec, with `v` operator, current point should be used as the first control point of the curve.
Do not overwrite current point before an SVG curve is built, so it can b actually used as first control point.
Since this rule is now enforced in the entire `/web` folder, there's no good reason for it to remain disabled in this file. Most likely, it's added to reduce the number of linting errors when we started enforcing block-scoped variables.
As can be seen in the code, the `xScaleBlockOffset` typed array doesn't depend on the actual image data but only on the width and x-scale. The width is obviously consistent for an image, and it turns out that in practice the `componentScaleX` is quite often identical between two (or more) adjacent image components.
All-in-all it's thus not necessary to *unconditionally* re-compute the `xScaleBlockOffset` when getting the JPEG image data.
While avoiding, in many cases, one or more loops can never be a bad thing these changes are unfortunately completely dominated by the rest of the JpegImage code and consequently doesn't really show up in benchmark results. *Hence I'd understand if this patch is ultimately deemed not necessary.*
*Note:* This is inspired by PR 5473, which made similar changes for another kind of JPEG data.
Since the implementation in `src/core/jpg.js` only supports 8-bit data, as opposed to similar code in `src/core/colorspace.js`, the computations can be further simplified since the `scale` is always constant.
By updating the coefficients, effectively inlining the `scale`, we'll thus avoid *four* multiplications for each loop iteration.
Unfortunately I wasn't able, based on a quick look through the test-files, to find a sufficiently *large* CMYK JPEG image in order for these changes to really show up in benchmark results. However, when testing the `cmykjpeg.pdf` manually there's a total of `120 000` fewer multiplication with this patch.
There's no particular reason for using the PDF.js helper function `createObjectURL`[1] in the debugger, instead of the native `URL.createObjectURL` directly, for a couple of reasons:
- The relevant code-path only applies to fonts loaded with the Font Loading API, and this isn't supported in IE anyway.
- The debugger can, since quite some time, not even be loaded in IE any more.
- General support for IE is now limited, and there's no guaratee that everything actually works.
---
[1] It provides a fallback for browsers with broken `Blob` support, which as usual means Internet Explorer :-P
In the current `AnnotationLayer` implementation, Popup annotations require that the parent annotation have already been rendered (otherwise they're simply ignored).
Usually the annotations are ordered, in the `/Annots` array, in such a way that this isn't a problem, however there's obviously no guarantee that all PDF generators actually do so. Hence we simply ensure, when rendering the `AnnotationLayer`, that the Popup annotations are handled last.
Interestingly the viewer already seem to work correctly as-is, with workers disabled and a non-standard `verbosity` level.
Hence this is possibly Node.js specific, but given that the issue is lacking *both* the PDF file in question and a runnable test-case, so this patch is essentially a best-effort guess at what the problem could be.
- Re-factor the "incorrect encoding" check, since this can be easily achieved using the general `findNextFileMarker` helper function (with a suitable `startPos` argument).
- Tweak a condition, to make it easier to see that the end of the data has been reached.
- Add a reference test for issue 1877, since it's what prompted the "incorrect encoding" check.
The other image decoders, i.e. the JBIG2 and JPEG 2000 ones, are using the common helper function `readUint16`. Most likely, the only reason that the JPEG decoder is doing it this way is because it originated *outside* of the PDF.js library.
Hence we can simply re-factor `src/core/jpg.js` to use the common `readUint16` helper function, which is especially nice given that the functionality was essentially *duplicated* in the code.
This covers cases that the `--fix` command couldn't deal with, and in a few cases (notably `src/core/jbig2.js`) the code was changed to use block-scoped variables instead.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/prefer-const
With the recent introduction of Prettier this sort of mass enabling of ESLint rules becomes a lot easier, since the code will be automatically reformatted as necessary to account for e.g. changed line lengths.
Note that this patch is generated automatically, by using the ESLint `--fix` argument, and will thus require some additional clean-up (which is done separately).
The debugging hash parameters[1] are intended to facilitate access to various tools/settings in PRODUCTION builds, protected by the `pdfBugEnabled` preference.
At this point, the remaining debugging hash parameters are mainly intended to allow access to the `PDFBug` tools and/or to quickly toggle certain larger features.
The "useOnlyCssZoom" functionality doesn't really seem to fit in with the rest of these hash parameters, since:
- This is, comparatively speaking, a minor viewer-specific feature.
- The zooming implementation will (almost) always fallback to CSS-only zooming, for any document, once the canvases becomes large enough. Hence, the majority of the CSS zooming feature can still be tested *directly* in any build of the viewer.
- After the initial implementation, years ago, the CSS-only zooming code in question hasn't changed much (or even at all), i.e. it doesn't seem like an active development target.[2]
- If the "useOnlyCssZoom" functionality was added today, it's unlikely that a hash parameter would've been added.
- Last, but not least, there's also a `useOnlyCssZoom` preference hence toggling this functionality shouldn't be too difficult (e.g. if someone needs to hack on it).
All in all, I'm thus suggesting that we remove the "useOnlyCssZoom" hash parameter.
---
[1] Originally these hash parameters could be used directly in any build, which was bad since it would allow any link to potentially disable functionality and/or reduce performance.
[2] If it had seen active development over the years, I'd be *much* more inclined to keep the hash parameter.
For people running e.g. Firefox with the `pdfBugEnabled` preference set, to allow quick access to debugging tools, this method will obviously run for every opened PDF file. However, in most cases the URL hash is empty and we can thus skip most of the parsing and simply return early instead.
These two cases should have been whitelisted prior to re-formatting respectively had the comments fixed afterwards, however I unfortunately missed them because of the massive size of the diff.
Fixes#11403
The PDF uses the non-embedded Type1 font Helvetica. Character codes 194 and 160 (`Â` and `NBSP`) are encoded as `.notdef`. We shouldn't show those glyphs because it seems that Acrobat Reader doesn't draw glyphs that are named `.notdef` in fonts like this.
In addition to testing `glyphName === ".notdef"`, we must test also `glyphName === ""` because the name `""` is used in `core/encodings.js` for undefined glyphs in encodings like `WinAnsiEncoding`.
The solution above hides the `Â` characters but now the replacement character (space) appears to be too wide. I found out that PDF.js ignores font's `Widths` array if the font has no `FontDescriptor` entry. That happens in #11403, so the default widths of Helvetica were used as specified in `core/metrics.js` and `.nodef` got a width of 333. The correct width is 0 as specified by the `Widths` array in the PDF. Thus we must never ignore `Widths`.
This removes a couple of, thanks to preceeding code, unnecessary `typeof PDFJSDev` checks, and also fixes a couple of incorrectly implemented (my fault) checks intended for `TESTING` builds.
After PR 9566, which removed all of the old Firefox extension code, the `FIREFOX` build flag is no longer used for anything.
It thus seems to me that it should be removed, for a couple of reasons:
- It's simply dead code now, which only serves to add confusion when looking at the `PDFJSDev` calls.
- It used to be that `MOZCENTRAL` and `FIREFOX` was *almost* always used together. However, ever since PR 9566 there's obviously been no effort put into keeping the `FIREFOX` build flags up to date.
- In the event that a new, Webextension based, Firefox addon is created in the future you'd still need to audit all `MOZCENTRAL` (and possibly `CHROME`) build flags to see what'd make sense for the addon.
While only the `MOZCENTRAL` builds will actually do anything meaningful with the telemetry data, none of the code in question actually runs *at all* in e.g. development mode.[1]
This seems bad since it essentially means that this code is completely untested, despite being quite important for the built-in Firefox PDF viewer, and this thus ought to be fixed.
In this case, the explanation for the current state of the code should be "for historical reasons". Before the viewer was split into the current components and before the pre-processor was improved, back when all code resided in the `web/viewer.js` file, the telemetry reporting was done with *direct* `FirefoxCom` calls. However, with the dummy `DefaultExternalServices.reportTelemetry` method there's nothing actually preventing attempting to report telemetry in any type of build.
NOTE: By running this code in GENERIC builds as well, in addition to just locally, the *viewer* part of telemetry reporting becomes tested e.g. in preview builds too which should help with reviewing.
---
[1] When fixing bug 1606566, I had to edit the relevant `PDFJSDev` checks to be able to actually test the changes locally.
With https://bugzilla.mozilla.org/show_bug.cgi?id=844349 now being fixed in Firefox, the textLayer will now actually stay hidden as intended regardless of the browser settings.
Hence it should no longer be necessary to display the fallback bar, nor print a warning in the console, for documents which contains a textLayer.
Besides removing the `supportsDocumentColors` methods in the default viewer, we can also remove a now unused l10n string.
This particular JSDoc comment is fairly old and it also contains some now unrelated/confusing information.
The only way to *guarantee* that the PDF.js library works as expected is to correctly set the global `workerSrc`[1], hence giving the impression that the option isn't strictly necessary is thus incorrect.
---
[1] Since advertising the fallbackWorkerSrc functionality definitely seems like the *wrong* thing to do.
This way we'll benefit from the existing font caching, and can thus avoid re-creating a fallback font over and over again during parsing.
(Thece changes necessitated the previous patch, since otherwise breakage could occur e.g. with fake workers.)
This is beneficial in situations where the Worker is being re-used, for example with fake workers, since it ensures that things like font resources are actually released.
This rule is already enabled in mozilla-central, and helps avoid some confusing formatting, see https://searchfox.org/mozilla-central/rev/9e45d74b956be046e5021a746b0c8912f1c27318/tools/lint/eslint/eslint-plugin-mozilla/lib/configs/recommended.js#209-210
With the recent introduction of Prettier some of the existing nested ternary statements became even more difficult to read, since any possibly helpful indentation was removed.
This particular ESLint rule wasn't entirely straightforward to enable, and I do recognize that there's a certain amount of subjectivity in the changes being made. Generally, the changes in this patch fall into three categories:
- Cases where a value is only clamped to a certain range (the easiest ones to update).
- Cases where the values involved are "simple", such as Numbers and Strings, which are re-factored to initialize the variable with the *default* value and only update it when necessary by using `if`/`else if` statements.
- Cases with more complex and/or larger values, such as TypedArrays, which are re-factored to let the variable be (implicitly) undefined and where all values are then set through `if`/`else if`/`else` statements.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/no-nested-ternary
Note that this will still allow the FBF tests to run locally, and also on the bots when invoked with `test`/`browsertest` (to not lose all the FBF test-coverage), but will no longer prevent `makeref` from running successfully on the bots.
As described in the issue, having a DOM element with `id=page2` (or any other number) will automatically cause that element to become linkable through the URL hash. That's currently leading to some confusing and outright wrong behaviour, since it obviously only works for pages that have been loaded and rendered.
For PDF documents the only officially supported way to reference a particular page through the URL hash is using the `#page=2` format, which also works for all pages regardless if they're loaded or not.
As far as I can tell there's nothing in the PDF.js default viewer that actually depends on the page/thumbnail `id` at this point in time, hence why I believe that this removal ought to be safe.
Just as a pre-caution this patch adds an `aria-label` to the page canvas, similar to the thumbnail canvas/image, to at least keep this information in the DOM.
Currently *all* users of `pdfjs-dist` are forced to install the `webpack` and `worker-loader` packages, despite the fact that they are *only* relevant if the `webpack.js` file is being used (with a custom Webpack build).
This really doesn't seem great, especially since those packages are the only remaining dependencies in the `pdfjs-dist` library, and it thus seem more reasonable overall that Webpack users handle those dependencies themselves.
To prevent unnecessarily cryptic runtime failures, when people update to newer `pdfjs-dist` versions, the `webpack.js` file was updated to explicitly check for the existence of the `worker-loader` package and error otherwise.
Furthermore, note that `webpack` was only listed as a dependency because of the `worker-loader` package itself (see issue 9248).
Obviously these changes may not be seen as great by Webpack users who rely on `pdfjs-dist`, since it forces them to handle the dependencies themselves, however it should improve things considerably for "general" users of `pdfjs-dist` by not burdening them with unnecessary dependencies.
These sort of changes are also in line with other recent changes, see PR 11418, which removed built-in fake worker loader code for specific JS builders/bundlers/frameworks. This work was prompted not only by a desire to simplify/clean-up old code, but also to lessen future support burden since the PDF.js contributors cannot be assumed to be experts in various JS bundlers.
In order to eventually get rid of SystemJS and start using native `import`s instead, we'll need to provide "complete" file identifiers since otherwise there'll be MIME type errors when attempting to use `import`.
When working on PR 11463 I couldn't help thinking that the `Array.prototype.some` callback function, used when determining the generator, was somewhat difficult to read with its partly unused and strangely named parameters.
This is not only useful to have one format for consistency, but also to
be able to quickly search for colors for e.g., finding duplicates or
when tweaking the CSS for custom deployments.
For reasons that I now cannot even begin to understand, the non-standard SegoeUISymbol font was placed in the `getStdFontMap`. That honestly makes no sense, hence this patch which does what I *should* have done from the start.
This covers the handful of cases that the `--fix` command couldn't deal with, and the changes aren't just fixing the linting errors but attempt to slightly improve the relevant code.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/prefer-const
Note that this patch is generated automatically, by using the ESLint `--fix` argument, and will thus require some additional clean-up (which is done separately).
Rather than having a copy of this regular expression in the `test/unit/api_spec.js` file, with a comment about keeping it up-to-date with the code in the viewer (note the incorrect file reference as well), we can just import it instead to simplify all of this.
The original issue did not contain a (reduced) test case that we could
include and linked test cases are not ideal for unit tests, so the
original PR could only be verified manually.
I found this a bit unfortunate considering that the print data is
exposed through the API, so I thought about how we could have an
automated test and managed to create a reduced test case with the
OpenAction dictionary from the file in the original issue.
Therefore, this commit includes a unit test for parsing OpenAction
dictionaries without `Type` entries. I verified that this PDF file
behaves the same as the original one, i.e., no print dialog is shown for
older viewers and the print dialog is shown for the most recent viewer.
This patch makes the follow changes:
- Remove no longer necessary inline `// eslint-disable-...` comments.
- Fix `// eslint-disable-...` comments that Prettier moved down, thus causing new linting errors.
- Concatenate strings which now fit on just one line.
- Fix comments that are now too long.
- Finally, and most importantly, adjust comments that Prettier moved down, since the new positions often is confusing or outright wrong.
Note that Prettier, purposely, has only limited [configuration options](https://prettier.io/docs/en/options.html). The configuration file is based on [the one in `mozilla central`](https://searchfox.org/mozilla-central/source/.prettierrc) with just a few additions (to avoid future breakage if the defaults ever changes).
Prettier is being used for a couple of reasons:
- To be consistent with `mozilla-central`, where Prettier is already in use across the tree.
- To ensure a *consistent* coding style everywhere, which is automatically enforced during linting (since Prettier is used as an ESLint plugin). This thus ends "all" formatting disussions once and for all, removing the need for review comments on most stylistic matters.
Many ESLint options are now redundant, and I've tried my best to remove all the now unnecessary options (but I may have missed some).
Note also that since Prettier considers the `printWidth` option as a guide, rather than a hard rule, this patch resorts to a small hack in the ESLint config to ensure that *comments* won't become too long.
*Please note:* This patch is generated automatically, by appending the `--fix` argument to the ESLint call used in the `gulp lint` task. It will thus require some additional clean-up, which will be done in a *separate* commit.
(On a more personal note, I'll readily admit that some of the changes Prettier makes are *extremely* ugly. However, in the name of consistency we'll probably have to live with that.)
There's a fair number of (primarily) `Array`s/`TypedArray`s whose formatting we don't want disturb, since in many cases that would lead to the code becoming much more difficult to read and/or break existing inline comments.
*Please note:* It may be a good idea to look through these cases individually, and possibly re-write some of the them (especially the `String` ones) to reduce the need for all of these ignore commands.
I recently noticed a couple of intermittent failures on Travis, hence this patch which changes the expectation to be identical to the 'Page Request' check in the preceding test-case.
https://npmjs.com/advisories/1084
webpack-stream between 4.0.3 and 5.0.0 added official support for Webpack 4. Which is nice since we ARE using Webpack 4... Also, dropped support for Node.js 4 which shouldn't be a big deal for us since we are already using packages that are incompatible with Node.js 4 (Webpack 4.x supports Node.js 6 and up).
Apparently Ghostscript can, in some cases, generate/include `Metadata` with incorrectly encoded characters.[1] This results in the viewer title looking wrong, which we thus attempt to avoid by falling back to the `Info` entry instead.
*Please note:* Obviously it would be better if this was fixed in the `Metadata` parser instead, rather than using a viewer work-around, but I'm just not sure how or even *if* that could actually be done given that the `Metadata` stream contains no trace of the *original* character.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1605526
---
[1] The problematic characters are from the Specials Unicode block, see https://en.wikipedia.org/wiki/Specials_(Unicode_block)
With these changes we'll always set the `pdfTitle` to the `Info` dictionary entry *first* (assuming it exists and isn't empty), before attempting to override it with the `Metadata` stream entry (assuming it exists, is non-empty *and* valid).
There should be no functional changes with this patch, but it will simplify the following patch somewhat.
Since bundlers, such as Webpack, cannot be told to leave `require` statements alone we are thus forced to jump through hoops in order to prevent these warnings in third-party deployments of the PDF.js library; please see [Webpack issue 8826](https://github.com/webpack/webpack) and libraries such as [require-fool-webpack](https://github.com/sindresorhus/require-fool-webpack).
*Please note:* This is based on the assumption that code running in Node.js won't ever be affected by e.g. Content Security Policies that prevent use of `eval`. If that ever occurs, we should revert to a normal `require` statement and simply document the Webpack warnings instead.
This patch reduces some duplication, by moving *all* fake worker loader code into the `setupFakeWorkerGlobal` function. Furthermore, the functions are simplified further by using `async`/`await` where appropriate.
There's no particularily good reason, as far as I can tell, to not support a custom worker path in Node.js environments (even if workers aren't supported). This patch thus make the Node.js fake worker loader code-path consistent with the fallback code-path used with *browser* fake worker loader.
Finally, this patch also deprecates[1] the `fallbackWorkerSrc` functionality, except in Node.js, since the user should *always* provide correct worker options since the fallback is nothing more than a best-effort solution.
---
[1] Although it probably shouldn't be removed until the next major version.
For performance reasons, and to avoid hanging the browser UI, the PDF.js library should *always* be used with web workers enabled.
At this point in time all of the supported browsers should have proper worker support, and Node.js is thus the only environment where workers aren't supported. Hence it no longer seems relevant/necessary to provide, by default, fake worker loaders for various JS builders/bundlers/frameworks in the PDF.js code itself.[1]
In order to simplify things, the fake worker loader code is thus simplified to now *only* support Node.js usage respectively "normal" browser usage out-of-the-box.[2]
*Please note:* The officially intended way of using the PDF.js library is with workers enabled, which can be done by setting `GlobalWorkerOptions.workerSrc`, `GlobalWorkerOptions.workerPort`, or manually providing a `PDFWorker` instance when calling `getDocument`.
---
[1] Note that it's still possible to *manually* disable workers, simply my manually loading the built `pdf.worker.js` file into the (current) global scope, however this's mostly intended for testing/debugging purposes.
[2] Unfortunately some bundlers such as Webpack, when used with third-party deployments of the PDF.js library, will start to print `Critical dependency: ...` warnings when run against the built `pdf.js` file from this patch. The reason is that despite the `require` calls being protected by *runtime* `isNodeJS` checks, it's not possible to simply tell Webpack to just ignore the `require`; please see [Webpack issue 8826](https://github.com/webpack/webpack) and libraries such as [require-fool-webpack](https://github.com/sindresorhus/require-fool-webpack).
*Please note:* This patch does *not* prevent the 'default_preferences' task from running more than once during `npm test`, but it does ensure that the tasks won't run *concurrently* by running the relevant tests in *series*.
While it would obviously still make sense to re-factor the gulpfile to account for changes in `gulp` version 4, by at least tweaking the `npm test` command the intermittent failures on Travis should at least go away.
During initial parsing of every PDF document we're currently creating a few `1 kB` strings, in order to find certain commands needed for initialization.
This seems inefficient, not to mention completely unnecessary, since we can just as well search through the raw bytes directly instead (similar to other parts of the code-base). One small complication here is the need to support backwards search, which does add some amount of "duplication" to this function.
The main benefits here are:
- No longer necessary to allocate *temporary* `1 kB` strings during initial parsing, thus saving some memory.
- In practice, for well-formed PDF documents, the number of iterations required to find the commands are usually very low. (For the `tracemonkey.pdf` file, there's a *total* of only 30 loop iterations.)
Looking at the `parseCurrentHash` function again it's now difficult for me to understand *what* I was thinking, since having a helper function that needs to be manually passed a `linkService` reference just looks weird.
This patch addresses a couple of smaller issues with the `PDFHistory` class:
- Most, if not all, other viewer components can be reset in one way or another, and there's no good reason for the `PDFHistory` implementation to be different here.
- Currently it's (technically) possible to keep adding entries to the browser history, via the `PDFHistory` instance, even after the document has been closed. That obviously makes no sense, and is caused by the lack of a `reset` method.
- The internal `this._isPagesLoaded` property was never actually reset, which would lead to it being temporarily wrong when a new document was opened in the default viewer.
Note that most (reasonably) modern browsers have supported this for a while now, see https://developer.mozilla.org/en-US/docs/Web/API/ReadableStream#Browser_compatibility
By moving the polyfill into `src/shared/compatibility.js` we can thus get rid of the need to manually export/import `ReadableStream` and simply use it directly instead.
The only change here which *could* possibly lead to a difference in behavior is in the `isFetchSupported` function. Previously we attempted to check for the existence of a global `ReadableStream` implementation, which could now pass (assuming obviously that the preceding checks also succeeded).
However I'm not sure if that's a problem, since the previous check only confirmed the existence of a native `ReadableStream` implementation and not that it actually worked correctly. Finally it *could* just as well have been a globally registered polyfill from an application embedding the PDF.js library.
Given that the error in question is surfaced on the API-side, this patch makes the following changes:
- Updates the wording such that it'll hopefully be slightly easier for users to understand.
- Changes the plain `Error` to an `InvalidPDFException` instead, since that should work better with the existing Error handling.
- Adds a unit-test which loads an empty PDF document (and also improves a pre-existing `InvalidPDFException` message and its test-case).
Given how this method is currently used there shouldn't be any fonts loaded at the point in time where it's called, but it does seem like a bad idea to assume that that's always going to be the case. Since `PDFDocument.checkFirstPage` is already asynchronous, it's easy enough to simply await `Catalog.cleanup` here.
(The patch also makes a tiny simplification in a loop in `Catalog.cleanup`.)
The `viewer` option was *only* used for checking that a document is loaded in `PDFPresentationMode.request`, however that's just as easy to do by simply utilizing `BaseViewer.pagesCount` instead and this way we can also avoid the DOM lookup.
In the PDF document in question, there's an ASCII85Decode inline image where the '>' part of EOD (end-of-data) marker is missing; hence the PDF document is corrupt.
This was a blatant oversight in PR 10217, since there's obviously no `this.pageNumber` property anywhere in the `BaseViewer`. Luckily this shouldn't have caused any bugs, since the only call-site is also validating the `pageNumber` (but correctly that time).
*This patch is simple enough that I almost feel like I'm overlooking some trivial reason why this would be a bad idea.*
Note how in `{BaseViewer, PDFThumbnailViewer}.setDocument` we're always getting the *first* `pdfPage` in order to initialize all pages/thumbnails.
However, once that's done the first `pdfPage` is simply ignored and rendering of the first page thus requires calling `PDFDocumentProxy.getPage` yet again. (And in the `BaseViewer` case, it's even done once more after `onePageRenderedCapability` is resolved.)
All in all, I cannot see why we cannot just immediately set the first `pdfPage` and thus avoid an early round-trip to the API in the `_ensurePdfPageLoaded` method before rendering can begin.
Note that the XRef cache will only hold objects returned through `Parser.getObj`, and indirectly via `Lexer.getObj`. Since neither of those methods will ever return `undefined`, we can simply `assert` that when inserting objects into the cache and thus get rid of one function call when doing cache lookups.
Obviously this won't have a huge effect on performance, however `XRef.fetch` is usually called *a lot* in larger documents and this patch thus cannot hurt.
I'm slightly surprised that this hasn't actually caused any (known) bugs, but that may be more luck than anything else since it fortunately doesn't seem common for Streams to be defined inside of an 'ObjStm'.[1]
Note that in the `XRef.fetchUncompressed` method we're *not* caching Streams, and that for very good reasons too.
- Streams, especially the `DecodeStream` ones, can become *very* large once read. Hence caching them really isn't a good idea simply because of the (potential) memory impact of doing so.
- Attempting to read from the *same* Stream more than once won't work, unless it's `reset` in between, since using any method such as e.g. `getBytes` always starts at the current data position.
- Given that even the `src/core/` code is now fairly asynchronous, see e.g. the `PartialEvaluator`, it's generally impossible to assert that any one Stream isn't being accessed "concurrently" by e.g. different `getOperatorList` calls. Hence `reset`-ing a cached Streams isn't going to work in the general case.
All in all, I cannot understand why it'd ever be correct to cache Streams in the `XRef.fetchCompressed` method.
---
[1] One example where that happens is the `issue3115r.pdf` file in the test-suite, where the streams in question are not actually used for anything within the PDF.js code.
- Change all occurences of `var` to `let`/`const`.
- Initialize the (temporary) Arrays with the correct sizes upfront.
- Inline the `isCmd` check. Obviously this won't make a huge difference, but given that the check is only relevant for corrupt documents it cannot hurt.
Having ran the entire test-suite locally with these `Number.isInteger` checks removed, there wasn't a single test failure anywhere; see also PR 8857.
Hence everything points to this being completely unnecessary now, and by removing this code there's thus fewer function calls being made in `XRef.fetchUncompressed`.
The contents of this comment hasn't been correct for *years*, ever since the library was properly split into main/worker-threads, so it's probably high time for this to be updated.
This code was originally added to support IE10 (and below), however with those browsers *explicitly* unsupported since PDF.js version `2.0` this code is now dead.
Obviously the `_pagesRequests` functionality is *mainly* used when `disableAutoFetch` is set, but it will also be used during ranged/streamed loading of documents.
However, the `_pagesRequests` property is currently an Array which seems a bit strange:
- Arrays are zero-indexed, but the first element will never actually be set in the code.
- The `_pagesRequests` Array is never cleared, unless a new document is loaded, and once the `PDFDocumentProxy.getPage` call has resolved/rejected the element is just replaced by `null`.
- Unless the document is browsed *in order* the resulting `_pagesRequests` Array can also be arbitrarily sparse.
All in all, I don't believe that an Array is an appropriate data structure to use for this purpose.
This patch simply restores the behaviour that existed prior to PR 7697, since I cannot imagine that that was changed other than by pure accident.
As mentioned by a comment in `BaseViewer.setDocument`: "Printing is semi-broken with auto fetch disabled.", and note that since triggering of printing is a synchronous operation there's generally no easy way to load the missing data.
https://github.com/mozilla/pdf.js/pull/7697/files#diff-529d1853ee1bba753a0fcb40ea778723L1114-L1118
Considering just how small/simple this code is, it doesn't seem necessary to have a separate method for it (even more so when there's only one call-site).
For documents with a Linearization dictionary the computed `startXRef` position will be relative to the raw file, rather than the actual PDF document itself (which begins with `%PDF-`).
Hence it's necessary to subtract `stream.start` in this case, since otherwise the `XRef.readXRef` method will increment the position too far resulting in parsing errors.
*Please note:* A a similar change was attempted in PR 5005, but it was subsequently backed out (in PR 5069) since other parts of the patch caused issues.
With these changes, it's possible to replace repeated function calls within a loop with just a single function call and subsequent assignment instead.
I've always disliked the solution in PR 10461, since it required changes to the `PDFHistory` code itself to deal with a bug in IE11.
Now that IE11 support is limited, it seems reasonable to remove these `pushState`/`replaceState` hacks from the main code-base and simply use polyfills instead.
For Popup annotation trigger elements consisting of an arbitrary polyline, you need to ensure that the 'stroke-width' is always non-zero since otherwise it's impossible to actually open/close the popup.
Unfortunately I don't believe that any of the test-suites can be used to test this, hence why no tests are included in the patch.
As we've seen in numerous other cases, avoiding unnecessary function calls is never a bad thing (even if the effect is probably tiny here).
With these changes we also avoid potentially two back-to-back `isDict` checks when evaluating possible Page nodes, and can also no longer accidentally pick a dictionary with an incorrect /Type.
For certain canvas-related errors (and probably others), the browser rendering exceptions may be propagated "as-is" to the PDF.js code. In this case, the exceptions are of the somewhat cryptic `NS_ERROR_FAILURE` type.
Unfortunately these aren't actual `Error`s, which thus ends up unintentionally triggering the `assert` in `PDFPageProxy._abortOperatorList`; sorry about that!
The bug report seem to suggest that we don't support UTF-16 strings with a BOM (byte order mark), which we *actually* do as evident by both the code and a unit-test.
The issue at play here is rather that we previously only supported big-endian UTF-16 BOM, and the `Title` string in the PDF document is using a *little-endian* UTF-16 BOM instead.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1593902
Rather than having to store a `PromiseCapability` on the `ObjectLoader` instances, we can simply convert `_walk` to be `async` and thus have the same functionality with native JavaScript instead.
I happened to look at this code and the way that the link target is set seems unecessarily convoluted, since we're using `Object.values` and `Array.prototype.includes` for *every* link being parsed.
Given that the number of link targets are so few, the easist solution honestly seem to be to just use a `switch` statement to do the link target mapping.
When the end of data has already been reached for the various Streams, the `getByte` methods will return `-1` to signal that to the caller. Note however that the current position obviously won't be incremented in this case, meaning that the `peekByte` methods will in this case *incorrectly* decrement the position.
Thankfully the corresponding `peekBytes` shouldn't be affected by this bug, since they decrement the current position with the *actually* returned number of bytes.
I'm not aware of any bugs caused by this blatant oversight, but that doesn't mean this shouldn't be fixed :-)
This will allow us to attempt to recover as much as possible of a page, rather than immediately failing, when a broken/unsupported ColorSpace is encountered. This patch thus extends the framework added in PRs such as e.g. 8240 and 8922, to also cover parsing of ColorSpaces.
The code in question is *only* relevant in non-`PRODUCTION` mode, i.e. the *development* version of the viewer run with `gulp server`, and has been completely unused at least since SystemJS was added.
I really cannot see any reason to keep this, since it's code which first of all isn't shipping and secondly isn't even being used in the development viewer.
When `ReadableStream` support was added to the `MessageHandler`, the `_onComObjOnMessage` function became more complex than previously.
All of the nested `if`/`else if`/`else` branches are now, at least in my opinion, making some of this code a bit difficult to follow. Hence this patch, which attempts to help readability by making use of early `return`s and `Error`s.
The patch also changes a couple of `var`/`let` occurences to `const`.
Note that using `in` leads to unnecessary stringification of the properties, which seems completely unnecessary here. To avoid future problems from these changes the `MessageHandler.on` method will now assert, in non-`PRODUCTION`/`TESTING` builds, that it's always called with a function as expected.
This patch also renames `callbacksCapabilities` to `callbackCapabilities`, note the removed "s", since using a double plural format looks a bit strange.
Given that the `isReply` property is an internal implementation detail, changing its type shouldn't be a problem. Note that by directly indicating if either data or an Error is sent, it's no longer necessary to use `in` when handling the callback.
Currently, for data in `ChunkedStream` instances, the `getMissingChunks` method is used in a couple of places to determine if data is already available or if it needs to be loaded.
When looking at how `ChunkedStream.getMissingChunks` is being used in the `ObjectLoader` you'll notice that we don't actually care about which *specific* chunks are missing, but rather only want essentially a yes/no answer to the "Is the data available?" question.
Furthermore, when looking at how `ChunkedStream.getMissingChunks` itself is implemented you'll notice that it (somewhat expectedly) always iterates over *all* chunks.
All in all, using `ChunkedStream.getMissingChunks` in the `ObjectLoader` seems like an unnecessary "heavy" and roundabout way to obtain a boolean value. However, it turns out there already exists a `ChunkedStream.allChunksLoaded` method, consisting of a *single* simple check, which seems like a perfect fit for the `ObjectLoader` use cases.
In particular, once the *entire* PDF document has been loaded (which is usually fairly quick with streaming enabled), you'd really want the `ObjectLoader` to be as simple/quick as possible (similar to e.g. loading a local files) which this patch should help with.
Note that I wouldn't expect this patch to have a huge effect on performance, but it will nonetheless save some CPU/memory resources when the `ObjectLoader` is used. (As usual this should help larger PDF documents, w.r.t. both file size and number of pages, the most.)
I completely overlooked this in PR 11281, but you obviously need to make similar changes in `PartialEvaluator.hasBlendModes` since it will otherwise ignore valid Blend Modes.
As can be seen in `PageViewport` only multiples of 90 degrees are really supported by the code, hence the unit-test doesn't really make sense.
(Possibly this should be enforced in the API, to avoid surprises, but given that this problem has always existed I'm passing on that for now.)
This argument is a left-over from older API code, where we unconditionally initialized `StatTimer` instances for every page. For quite some time that's only been done when `pdfBug` is set, hence it seems unnecessary to keep this functionality.
Even though the currect situation only results in six unnecessary function calls per page, it nonetheless seems completely unnecessary to call dummy functions when `pdfBug` is *not* set (i.e. the default behaviour).
For *very* long/large documents fetching all pages on load may cause quite bad performance, both memory and CPU wise. In order to at least slightly alleviate this, we can let the viewer treat these kind of documents[1] as if `disableAutoFetch` were set.
---
[1] One example of a really bad case is https://bugzilla.mozilla.org/show_bug.cgi?id=1588435, which this patch should at least help somewhat. In general, for these cases, we'd probably need to implement switching between `PDFViewer`/`PDFSinglePageViewer` (as already tracked on GitHub) and use the latter for these kind of long documents.
Unfortunately I forgot to test `gulp mozcentraldiff` with PR 11256, which is really bad since it will cause the en-US `chrome.properties` l10n file to be deleted at mozilla-central; sorry about breaking this!
In order to address this we'll have to re-add (only) the en-US `chrome.properties` l10n file, which is simple enough. Furthermore, since we're not doing any sort of build-specific parsing of the l10n files, we can just copy the en-US files as-is rather than having to run `gulp locale` during `gulp mozcentral`.
As can be seen in the API, there's a number of document loading Exception handlers which are both really simple and highly similar. Hence these are changed such that all the relevant Exceptions are sent via *one* message instead.
Furthermore, the patch also avoids unnecessarily re-creating `UnknownErrorException`s at the worker side and removes an unnecessary `bind` call.
This was done automatically, using the `npm audit fix` command.
*Please note:* There's two vulnerabilities left, however they couldn't be fixed automatically and may thus benefit from being handled separately.
Obviously this won't look exactly right, but considering that the PDF file doesn't bother embedding non-standard fonts this is the best that we can do here.
Originally only `skipPages` existed, but given that `firstPage`/`lastPage` has existed for a long time now using them whenever possible looks simpler overall.
- In the `ibwa-bad` case the sixteenth page contains corrupt/incomplete commands, but given that we're suppressing `Error`s by default now skipping hardly seems warranted any more.
- In the `geothermal.pdf` case the first page contains an unsupported ColourSpace, but again we're suppressing `Error`s by default now and skipping hardly seems warranted any more.
With the removal of the (standalone) Firefox building code in PR 9566 (a year and a half ago), these files are now completely unused in the GitHub repository[1].
Hence it doesn't really seem necessary to keep fetching them with `gulp importl10n`, and the existing files in the `l10n` folder can also be removed (thanks to version control, they're easy enough to restore should the need ever arise).
The patch also allows an additional simplification, for the `gulp locale` and `gulp mozcentral` commands, since it's now possible to stop writing `l10n` files to the `extensions/firefox/` folder and instead just copy them similar to other build targets.
---
[1] They're obviously still used in `mozilla-central`, for fallback messages displayed through `PdfStreamConverter.jsm`, but that doesn't make it necessary to keep them *here* as far as I'm concerned.
With the removal of the (standalone) Firefox building code in PR 9566 (a year and a half ago), these files are now completely unused.
Hence it doesn't really make sense to keep building them as part of `gulp locale`, and the existing files in the `l10n` folder can also be removed (thanks to version control, they're easy enough to restore should the need ever arise).
The issue has been open for years now, and has even been marked with `5-good-beginner-bug` for *months*, without any movement.
Considering just how simple the suggested solution is, I'm submitting this patch just to close out a long-standing issue.
The following is printed with every build, which gets kind of annoying when looking at the logs:
```
Replace Autoprefixer browsers option to Browserslist config.
Use browserslist key in package.json or .browserslistrc file.
Using browsers option cause some error. Browserslist config
can be used for Babel, Autoprefixer, postcss-normalize and other tools.
If you really need to use option, rename it to overrideBrowserslist.
Learn more at:
https://github.com/browserslist/browserslist#readmehttps://twitter.com/browserslist
```
Given how we're using Autoprefixer, with a number of build-specific configs, simply changing the option name seems like the easiest solution here.
(I'm also adding a couple of newlines at the `autoprefixer` call-sites, to aid readability.)
Obviously userAgent checks aren't that great, since it's very easy to spoof, but it probably doesn't hurt to attempt to extend this (since it's limited to `GENERIC` builds).
Besides, anyone using the default viewer can always set the `maxCanvasPixels` option to a value of their choosing.
This patch is making me somewhat worried about future regressions, since it's certainly easy to imagine this completely breaking certain kinds of corrupt/edited PDF documents while fixing others.[1]
Obviously it passes all existing reference tests (and even improves one), however compared to many other patches there's no telling how much it could break.
The only reason that I'm even submitting this patch, is because of the number of open issues that it would address.
Generally speaking though, the best course of action would probably be if `XRef.indexObjects` was re-written to be much more robust (since it currently feels somewhat hand-wavy in parts). E.g. by actually checking/validating more of the objects before committing to them.
---
[1] Especially given that it's reverting part of PR 5910, however in the case of issue 5909 it seems that other (more recent) changes have actually made that PR redundant.
Sometimes we also used `@return`, but `@returns` is what the JSDoc
documentation recommends. Even though `@return` works as an alias, it's
good to use the recommended syntax and to be consistent within the
project.
Sometimes we also used `@return` or `@returns`, but `@type` is what
the JSDoc documentation recommends. This also improves the documentation
because before this commit the types were not shown and now they are.
Rather than specifying certain build targets manually, it seems much more appropriate (and future-proof) to use the `SKIP_BABEL` build target instead.
Also, the patch adds a missing `/* eslint no-var: error */` line since I'm touch the file anyway and no code-changes were necessary for it.
As part of attempting to fix a number issues containing PDF documents with corrupt XRef tables, I'd like to improve the reference test-coverage slightly *first*.
Obviously this will increase the runtime of the tests a bit, however I'd rather "waste" resources on the bots instead of developer time fixing regressions which could have been avoided.
For badly generated PDF documents, with issue 6961 being one example, there's well over one hundred thousand function calls being made in total for just the *two* pages.
This handles the two different ways that fonts can be loaded, either by Name (which is the common case) or by Reference.
Furthermore, this also takes the `ignoreErrors` option into account when deciding whether to fallback or Error.
Finally, by creating a minimal but valid Font dictionary, there's no special-cases necessary in any of the font parsing code.
Co-authored-by: huzjakd <huzjakd@gmail.com>
Co-Authored-By: Jonas Jenwald <jonas.jenwald@gmail.com>
With the sole exception of `meh`[1], none of the locales removed here have even been updated since the last change was made to the default `en-US` locale.
Please note that if/when these locales would start shipping in Nightly again, they will automatically be re-added in PDF.js as well with the `gulp importl10n` command.
---
[1] The current translation is also somewhat incomplete, to put it mildly.
All of these methods have been marked as `deprecated` in *three* releases now, and I'd thus like to (slowly) move towards complete removal.
However rather than just removing the methods right away, which would cause somewhat cryptic failures, this patch tries to implement a hopefully reasonable middle ground by throwing `Error`s with (essentially) the same information as the previous warnings.
While the previous `deprecated` messages could perhaps be seen as optional, with these changes API consumers will now be forced to actually migrate their code.
Rather than having the script remove locales automatically, which seems like a heavy-handed approach at least initially, listing these for manual checking seems nice though.
Rather than having to manually maintain a static list of language codes, it's much easier to simply fetch the active ones from `mozilla-central` instead.
As part of this the code in `external/importL10n/locales.js` was modernized slightly, by using Promises/async functions to get rid of a bunch of annoying callbacks (which shouldn't be a problem for reasonably modern Node.js versions).
*Please note:* I've been thinking about possible ways of addressing this issue for a while now, but all of the solutions I came up with became too complicated and thus hurt readability of the code.
However, it occured to me that we're essentially trying to add a heuristic *on top* of another heuristic, and that it shouldn't matter how efficient the code is as long as it works.
In the PDF file in the issue the Encoding contains glyphNames of the `Cdd` format, which our existing heuristics will treat as base 10 values. However, in this particular file they actually contain base 16 values, which we thus attempt to detect and fix such that text-selection works.
I've absolutely no idea why I wrote the code that way originally, since the nested `if`s are not really helping readability one bit.
Hence this patch which changes things to use early `return`s instead, to help readability.
Rather than manually clamping the `width` here, we can just `export` an already existing helper function from `ui_utils.js` instead.
(Hopefully it will eventually be possible to replace this helper function with a native `Math.clamp` function, given that there exists a "Stage 1 Proposal" for adding such a thing to the ECMAScript specification.)
By utilizing a base "class", things become significantly simpler. Unfortunately the new `BaseException` cannot be a proper ES6 class and just extend `Error`, since the SystemJS dependency doesn't seem to play well with that.
Note also that we (generally) need to keep the `name` property on the actual `...Exception` object, rather than on its prototype, since the property will otherwise be dropped during the structured cloning used with `postMessage`.
By default, i.e. with workers enabled, it's *purposely* not possible to send `Dict`s and `Stream`s from the worker-thread. This is achieved by defining a `function` on every `Dict` instance, since that ensures that [the structured clone algoritm](https://developer.mozilla.org/en-US/docs/Web/API/Web_Workers_API/Structured_clone_algorithm) will throw an Error on `postMessage`.
However, with workers *disabled* we fall-back to the `LoopbackPort` implementation which just ignores any `function`s, thus incorrectly allowing sending of data which *should* be unclonable.
The following changes were made:
- Remove unnecessary `typeof` checks in the `get`/`getAsync` methods.
- Reduce unnecessary code duplication in the `get`/`getAsync` methods.
- Inline the `Ref` checks in the `get`/`getAsync`/`getArray` methods, since it helps avoid many unnecessary functions calls. I.e. this way it's possible to directly call `XRef.{fetch, fetchAsync)` only when necessary, rather than always having to call `XRef.{fetchIfRef, fetchIfRefAsync)`.
This patch was tested using the PDF file from issue 2618, i.e. http://bugzilla-attachments.gnome.org/attachment.cgi?id=226471, using the following manifest file:
```
[
{ "id": "issue2618",
"file": "../web/pdfs/issue2618.pdf",
"md5": "",
"rounds": 250,
"type": "eq"
}
]
```
This gave the following results when comparing this patch against the `master` branch:
```
-- Grouped By browser, stat --
browser | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ------------ | ----- | ------------ | ----------- | --- | ----- | -------------
Firefox | Overall | 250 | 2821 | 2790 | -32 | -1.12 | faster
Firefox | Page Request | 250 | 2 | 2 | 0 | 6.68 |
Firefox | Rendering | 250 | 2820 | 2788 | -32 | -1.13 | faster
```
Compared to the recently replaced `URL` polyfill, the new `ReadableStream` polyfill isn't being exported globally for two reasons:
- We're currently checking for the existence of a global `ReadableStream` implementation when determining if the Fetch API will be used; please see `isFetchSupported` in the src/display/display_utils.js file.
- Given that it's much newer functionality (compared to `URL`) and that not all browsers may implement all parts of the specification yet, not exposing the `ReadableStream` globally seems safer for now.
Having these methods fallback to returning `null` in only *one* particular case seems outright wrong, since a "falsy" value will thus be handled incorrectly.
The only reason that this hasn't caused issues in practice is that there's only one call-site passing in three keys, and in that case we're trying to read a font file where falling back to `null` isn't a problem.
This only required changing the import paths. The `es` folder contains
all polyfills we need now. If we want to import everything, we need to
explicitly require the `index` file.
Hopefully this patch makes sense, and in order to reduce the regression risk the implementation ensures that only completely missing widths are being replaced.
Given that there's only a couple of call-sites, and that the helper function is really simple, it doesn't seem entirely necessary to keep it around. While fewer function calls is always a good thing, in this case the performance impact is small enough to be unmeasurable.
With *one* single exception the code in `MessageHandler` is using `reason` when passing around various Errors, hence this patch also renames an `error` key for consistency.
The `streamId` short-hand in `MessageHandler._processStreamMessage` was only used partially througout the method, which seemed kind of strange, hence that's fixed in this patch.
Furthermore, always giving the `streamController` object a constant shape in `MessageHandler.sendWithStream` cannot hurt either.
With this patch we're finally able to abort worker-thread parsing of the `OperatorList`, rather than *only* aborting the main-thread rendering itself, when the `RenderTask.cancel` method is being called.
This will help improve perceived performance in the default viewer, especially when reading longer and more complex documents, since pages that've been scrolled out-of-view (and thus evicted from the cache) will no longer compete for parsing resources on the worker-thread.
*Please note:* With the implementation in this patch we're *not* aborting worker-thread parsing immediately on `RenderTask.cancel`, since that would lead to *worse* performance in many cases. For example: When zoom/rotation occurs in the viewer, while parsing/rendering is still ongoing, a `cancel` call will usually be (almost) immediately folled by a new `PDFPageProxy.render` call. In that case you obviously don't want to abort parsing on the worker-thread, since that would risk throwing away a partially parsed `OperatorList` and thus force unnecessary re-parsing which will regress perceived performance (especially for more complex documents).
When choosing a reasonable delay, before cancelling `getOperatorList` on the worker-thread when `RenderTask.cancel` is called, two different positions need to be considered:
1. The delay needs to be short enough, since a timeout in the multiple seconds range would essentially make this entire functionality meaningless (by always allowing most/all pages enough time to finish parsing).
2. The delay cannot be *too* short, since that would actually *reduce* performance in the zoom/rotation case outlined above. Furthermore, the time between `RenderTask.cancel` and `PDFPageProxy.render` calls will obviously be affected by both general computer performance and current CPU load.
It's certainly possible that the timeout may require some further tweaks, however the value settled on in this patch was easily *one order* of magnitude larger than the delta between cancel/render in my tests.
Currently this leads to confusing behaviour with `eventBusDispatchToDOM = true` set, since then there's *two* different "localized" events being fired.
There's no good reason for calling this helper function without a `url` parameter, and this way we can prevent that from happening.
Note how the `PDFOutlineViewer` call-site was already doing the right thing here, and only the `LinkAnnotationElement` call-site needed a small adjustment to make it work.
This rule is in the process of being rolled out in mozilla-central, and it helps avoid unnecessary `async` functions together with `new Promise(...)`.
Please see https://eslint.org/docs/rules/no-async-promise-executor for additional information.
With the changes made in PR 11069, it's no longer necessary to include the `pageIndex`/`intent` parameters when sending 'GetOperatorList' data. In the previous implementation these properties were used to associate the `OperatorList` with the correct `RenderTask`, however now that `ReadableStream`s are used that's handled automatically and it's thus dead code at this point.
Besides avoiding errors during loading, this also ensures that the document will be correctly scrolled/zoomed into view once the viewer becomes visible.
This "new" behaviour was always intended, see PR 2613, however various re-factoring over the years seem to have broken this (and I'm probably at least somewhat responsible for that).
By transfering, rather than copying, `ArrayBuffer`s between the main- and worker-threads, you can avoid unnecessary allocations by only having *one* copy of the same data.
Hence manually setting `postMessageTransfers: false`, when calling `getDocument`, is a performance footgun[1] which will do nothing but waste memory.
Given that every reasonably modern browser supports `postMessage` transfers[2], I really don't see why it should be possible to force-disable this functionality.
Looking at the browser support, for `postMessage` transfers[2], it's highly unlikely that PDF.js is even usable in browsers without it. However, the feature testing of `postMessage` transfers is kept for the time being just to err on the safe side.
---
[1] This is somewhat similar to the, now removed, `disableWorker` parameter which also provided API users a much too simple way of reducing performance.
[2] See e.g. https://developer.mozilla.org/en-US/docs/Web/API/MessagePort/postMessage#Browser_compatibility and https://developer.mozilla.org/en-US/docs/Web/API/Transferable#Browser_compatibility
Note how the sent values have inconsistent types, with a boolean in one case and an object in the other (normal) case.
Furthermore, explicitly sending a `supportTypedArray: true` property seems superfluous at least to me.
This check was added in PR 2445, however it's no longer necessary since all data[1] is now loaded on the main-thread (and then transferred to the worker-thread).
Furthermore, by default the Fetch API is now (usually) used rather than `XMLHttpRequest`.
All in all, while these checks *were* necessary at one point that's no longer the case and they can thus be removed.
---
[1] This includes both the actual PDF data, as well as the CMap data.
It recently occurred to me that the CMap data should be an excellent candidate for transfering.
This will help reduce peak memory usage for PDF documents using CMaps, since transfering of data avoids duplicating it on both the main- and worker-threads.
Unfortunately it's not possible to actually transfer data when *returning* data through `sendWithPromise`, and another solution had to be used.
Initially I looked at using one message for requesting the data, and another message for returning the actual CMap data. While that should have worked, it would have meant adding a lot more complexity particularly on the worker-thread.
Hence the simplest solution, at least in my opinion, is to utilize `sendWithStream` since that makes it *really* easy to transfer the CMap data. (This required PR 11115 to land first, since otherwise CMap fetch errors won't propagate correctly to the worker-thread.)
Please note that the patch *purposely* only changes the API to Worker communication, and not the API *itself* since changing the interface of `CMapReaderFactory` would be a breaking change.
Furthermore, given the relatively small size of the `.bcmap` files (the largest one is smaller than the default range-request size) streaming doesn't really seem necessary either.
Having recently worked with this code, it struck me that most of the `postMessage` calls where `Error`s are involved have never been correctly implemented (i.e. missing `wrapReason` calls).
There's only three call-sites and one of them doesn't even need the complete functionality of `resolveCall`, hence it seems reasonable to just inline this code.
An additional benefit of this is that the `Function.prototype.apply()` instance can also be converted into "normal" function calls, which should be a tiny bit more efficient.
The patch also replaces a number of unnecessary arrow functions, in relevant parts of the `MessageHandler` code, with "normal" functions instead.
Finally, all `Promise.resolve().then(...)` calls are replaced with `new Promise(...)` instead since the latter is a tiny bit more efficient. This also explains the test failures on the Linux bot, with a prior version of the patch, since the `Promise.resolve().then(...)` format essentially creates two Promises thus causing additional delay.
At this point in time it's easy to convert the `MessageHandler.on` call-sites to use arrow functions, and thus let the JavaScript engine handle scopes for us, rather than having to manually keep references to the relevant scopes in `MessageHandler`.[1]
An additional benefit of this is that a couple of `Function.prototype.call()` instances can now be converted into "normal" function calls, which should be a tiny bit more efficient.
All in all, I don't see any compelling reason why it'd be necessary to keep supporting custom `scope`s in the `MessageHandler` implementation.
---
[1] In the event that a custom scope is ever needed, simply using `bind` on the handler function when calling `MessageHandler.on` ought to work as well.
Since `wrapReason` and `makeReasonSerializable` are essentially functionally equivalent it doesn't seem necessary to keep both of them around, especially when `makeReasonSerializable` only has a *single* call-site.
Given that the `stream` property is an internal implementation detail, changing its type shouldn't be a problem. By using Numbers instead, we can avoid unnecessary String allocations when creating/processing Streams.
With PR 11069 we're now using Streams for OperatorList parsing (in addition to just TextContent parsing), which brings the nice benefit of being able to easily abort parsing on the worker-thread thus saving resources.
However, since we're now creating many more `ReadableStream` there appears to be a tiny bit more overhead because of it (giving ~1% slower runtime of `browsertest` on the bots). In this case we're just going to have to accept such a small regression, since the benefits of using Streams clearly outweighs it.
What we *can* do here, is to try and make the Streams part of the `MessageHandler` implementation slightly more efficient by e.g. removing unnecessary function calls (which has been helpful in other parts of the code-base). To that end, this patch makes the following changes:
- Actually support `transfers` in `MessageHandler.sendWithStream`, since the parameter was being ignored.
- Inline the `sendStreamRequest`/`sendStreamResponse` helper functions at their respective call-sites. Obviously this causes some amount of code duplication, however I still think this change seems reasonable since for each call-site:
- It avoids making one unnecessary function call.
- It avoids allocating one temporary object.
- It avoids sending, and thus structure clone, various undefined object properties.
- Inline objects in the `MessageHandler.{send, sendWithPromise}` methods.
- Finally, directly call `comObj.postMessage` in various methods when `transfers` are *not* present, rather than calling `MessageHandler.postMessage`, to further reduce the amount of function calls.
One of the motivations for using `setAttribute` in the first place was to support more efficient DOM updates in the `expandTextDivs` method, since performance of the `enhanceTextSelection` mode can be somewhat bad when there's a lot of `textDivs` on the page.
With recent `TextLayer` changes/optimizations it's no longer necessary to store a complete `style`-string for every `textDiv`, and we can thus re-visit the `setAttribute` usage.
Note that with the current code, in `appendText`, there's only *one* string per `textDiv` which avoids a bunch of temporary strings. While the changes in this patch means that there's now *three* strings per `textDiv` instead, the total length of these strings are now quite a bit shorter (42 characters to be exact).
*This should obviously have been done in PR 11097, but for some reason I completely overlooked it; sorry about that.*
There's no good reason to update the font unless you're actually going to measure the width of the textContent. This can reduce unnecessary font switching a fair bit, even for documents which are somewhat simple/short (in e.g. the `tracemonkey.pdf` file this cuts the amount of font switches almost in half).
For performance reasons single-char text divs aren't being scaled, as outlined in a comment in `appendText`. Hence it doesn't seem necessary, or even a good idea, to unconditionally measuring the width of the text in `_layoutText`.
This major version bump required two changes:
- The global line in the mobile viewer example should be removed because
the `.eslintrc` file already defines these globals and with the new
`eslint` version we otherwise get an error saying "'pdfjsLib' is already
defined as a built-in global variable".
- The ECMA version for the examples must be set to 6 since we're using
modules, otherwise we get an error saying "sourceType 'module' is not
supported when ecmaVersion < 2015". It turns out that the previous
version of `eslint` already used ECMA version 6 silently even though
we set 5, see https://github.com/eslint/eslint/issues/9687#issuecomment-432413384,
so in terms of our code nothing really changes.
These functions aren't returning anything, now that they're using `ReadableStream`s, and it thus doesn't seem necessary to re-throw errors (also given the console message that's caused by it).
*Please note:* The majority of this patch was written by Yury, and it's simply been rebased and slightly extended to prevent issues when dealing with `RenderingCancelledException`.
By leveraging streams this (finally) provides a simple way in which parsing can be aborted on the worker-thread, which will ultimately help save resources.
With this patch worker-thread parsing will *only* be aborted when the document is destroyed, and not when rendering is cancelled. There's a couple of reasons for this:
- The API currently expects the *entire* OperatorList to be extracted, or an Error to occur, once it's been started. Hence additional re-factoring/re-writing of the API code will be necessary to properly support cancelling and re-starting of OperatorList parsing in cases where the `lastChunk` hasn't yet been seen.
- Even with the above addressed, immediately cancelling when encountering a `RenderingCancelledException` will lead to worse performance in e.g. the default viewer. When zooming and/or rotation of the document occurs it's very likely that `cancel` will be (almost) immediately followed by a new `render` call. In that case you'd obviously *not* want to abort parsing on the worker-thread, since then you'd risk throwing away a partially parsed Page and thus be forced to re-parse it again which will regress perceived performance.
- This patch is already *somewhat* risky, given that it touches fundamentally important/critical code, and trying to keep it somewhat small should hopefully reduce the risk of regressions (and simplify reviewing as well).
Time permitting, once this has landed and been in Nightly for awhile, I'll try to work on the remaining points outlined above.
Co-Authored-By: Yury Delendik <ydelendik@mozilla.com>
Co-Authored-By: Jonas Jenwald <jonas.jenwald@gmail.com>
Furthermore, it's possible to re-use the same Array for all `textDiv`s on the page and the resulting padding string also becomes a lot more compact.
Please note that the `paddingLeft` branch was moved, since the padding values need to be ordered as `top, right, bottom, left`.
Finally, with this re-factoring it's no longer necessary to cache the original `style` string for every `textDiv` when `enhanceTextSelection` is enabled.
Given that browsers will reject padding values smaller than zero (which may be caused by limited numerical precision during calculations in the `expand` code), it makes no sense to include those when expanding the `textDiv`s.
Given that the different types of `Stream`s will never be cached, this thus implies that the `XRef.cache` Array will *always* be more-or-less sparse.
Generally speaking, the longer the document the more sparse the `XRef.cache` will thus become. For example, looking at the `pdf.pdf` file from the test-suite: The length of the `XRef.cache` Array will be a few hundred thousand elements, with approximately 95% of them being empty.
Hence it seems pretty clear that an Array isn't really the best data-structure for this kind of cache, and this patch thus changes it to a Map instead.
This patch-series was tested using the PDF file from issue 2618, i.e. http://bugzilla-attachments.gnome.org/attachment.cgi?id=226471, with the following manifest file:
```
[
{ "id": "issue2618",
"file": "../web/pdfs/issue2618.pdf",
"md5": "",
"rounds": 200,
"type": "eq"
}
]
```
which gave the following results when comparing this patch-series against the `master` branch:
```
-- Grouped By browser, stat --
browser | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ------------ | ----- | ------------ | ----------- | --- | ----- | -------------
Firefox | Overall | 200 | 2736 | 2736 | 1 | 0.02 |
Firefox | Page Request | 200 | 2 | 2 | 0 | -8.26 | faster
Firefox | Rendering | 200 | 2733 | 2734 | 1 | 0.03 |
```
The relevant methods are usually not hot enough for these changes to have an easily measurable effect, however there's been a lot of other cases where similiar inlining has helped performance. (And these changes may help offset the changes made in the next patch.)
For very large and complex PDF files this will help performance *slightly*, since `Parser.getObj` is called *a lot* during parsing in the worker.
This patch was tested using the PDF file from issue 2618, i.e. http://bugzilla-attachments.gnome.org/attachment.cgi?id=226471, with the following manifest file:
```
[
{ "id": "issue2618",
"file": "../web/pdfs/issue2618.pdf",
"md5": "",
"rounds": 200,
"type": "eq"
}
]
```
which gave the following results when comparing this patch against the `master` branch:
```
-- Grouped By browser, stat --
browser | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ------------ | ----- | ------------ | ----------- | --- | ----- | -------------
Firefox | Overall | 200 | 2847 | 2830 | -17 | -0.60 | faster
Firefox | Page Request | 200 | 2 | 2 | 0 | -7.14 |
Firefox | Rendering | 200 | 2844 | 2827 | -17 | -0.60 | faster
```
Looking at this again, it struck me that added functionality in `Util.intersect` is probably more confusing than helpful in general; sorry about the churn in this code!
Based on the parameter name you'd probably expect it to only match when the intersection is `[0, 0, 0, 0]` and not when only one component is zero, hence the `skipEmpty` parameter thus feels too tightly coupled to the `Page.view` getter.
This was added in PR 4470, but doesn't appear to have been used since.
While it's certainly easy to understand how this was helpful during development of that PR, actually providing this hash parameter isn't going to work anymore given that the original CMap files were also removed from the repository.
I suppose that the hash parameter *could* be useful if you'd attempt to update the BCMap files, however that hasn't been attempted even once in over *five* years time. Furthermore, at this point using the `AppOptions` directly in that situation should also work fine.
All in all, this seems like a piece of old and unused code which we can simply remove now.
This is based on a real-world PDF file I encountered very recently[1], although I'm currently unable to recall where I saw it.
Note that different PDF viewers handle these sort of errors differently, with Adobe Reader outright failing to render the attached PDF file whereas PDFium mostly handles it "correctly".
The patch makes the following notable changes:
- Refactor the `cropBox` and `mediaBox` getters, on the `Page`, to reduce unnecessary duplication. (This will also help in the future, if support for extracting additional page bounding boxes are added to the API.)
- Ensure that the page bounding boxes, i.e. `cropBox` and `mediaBox`, are never empty to prevent issues/weirdness in the viewer.
- Ensure that the `view` getter on the `Page` will never return an empty intersection of the `cropBox` and `mediaBox`.
- Add an *optional* parameter to `Util.intersect`, to allow checking that the computed intersection isn't actually empty.
- Change `Util.intersect` to have consistent return types, since Arrays are of type `Object` and falling back to returning a `Boolean` thus seem strange.
---
[1] In that case I believe that only the `cropBox` was empty, but it seemed like a good idea to attempt to fix a bunch of related cases all at once.
The current code will only consider the `cropBox` and `mediaBox` as equal when they both point to the *same* underlying Array. In the case where a PDF file actually specifies both boxes independently, with the exact same values in each, the comparison will currently fail and lead to an unneeded intersection computation.
By using the same internal formatting here as in the `Ref.toString` method, in `src/core/primitives.js`, all cache-keys will become at least two bytes shorter (and most three bytes shorter).
Obviously this won't have a huge effect on memory since there's only one cache entry per page, but it nonetheless seems wasteful to use longer keys than strictly required.
With the changes to the `StreamType`/`FontType` "enums" in PR 11029, one unfortunate result is that `getStats` now *always* returns empty Arrays. Something that everyone, myself included, apparently missed is that you obviously cannot index an Array with Strings :-)
I wrongly assumed that the unit-tests would catch any bugs, but they apparently suffered from the same issue as the code in `src/core/`.
Another possible option could perhaps be to use `Set`s, rather than objects, but that will require larger changes since `LoopbackPort` (in `src/display/api.js`) doesn't support them.
Firefox telemetry supports using string labels now. Convert our integers
that we used for categories to just use strings.
The upstream work will happen in:
https://bugzilla.mozilla.org/show_bug.cgi?id=1566882
There's a number of spots in the current code, and tests, where `cancel` methods are not called with appropriate arguments (leading to Promises not being rejected with Errors as intended).
In some cases the cancel `reason` is implicitly set to `undefined`, and in others the cancel `reason` is just a plain String. To address this inconsistency, the patch changes things such that cancelling is done with `AbortException`s everywhere instead.
Currently the indicator may remain visible even after the document has been closed, which seems weird given that no page is either visible nor rendering :-)
Add a work-around, in `glyphlist.js`, for bad PDF generators which use a non-standard `/f_f` string in the `Encoding` dictionary when referring to the ff ligature (issue 11016)
This patch will not incur any (measurable) overhead, since the glyphlist is already quite long and one more entry won't really matter, which is important given that this sort of PDF corruption ought to be very rare.
Furthermore, this patch purposely does *not* add a bunch of similarly modified ligature names on pure speculation. Any similar additions, for other ligatures, should only be made if there's real-world examples of PDF files where that's actually necessary.
For very large and complex PDF files this will help performance slightly, since `EvaluatorPreprocessor.read` is called a lot during parsing in the worker.
This patch was tested using the PDF file from issue 2618, i.e. http://bugzilla-attachments.gnome.org/attachment.cgi?id=226471, using the following manifest file:
```
[
{ "id": "issue2618",
"file": "../web/pdfs/issue2618.pdf",
"md5": "",
"rounds": 200,
"type": "eq"
}
]
```
This gave the following results when comparing this patch against the `master` branch:
```
-- Grouped By browser, stat --
browser | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ------------ | ----- | ------------ | ----------- | --- | ----- | -------------
Firefox | Overall | 200 | 3402 | 3358 | -43 | -1.28 | faster
Firefox | Page Request | 200 | 1 | 1 | 0 | 26.71 |
Firefox | Rendering | 200 | 3401 | 3357 | -44 | -1.28 | faster
```
Ensure that setting the `zoomDisabledTimeout` isn't skipped, regardless of the supported zoom keys, when handling mouse wheel events (PR 7097 follow-up)
*Possible follow-up:* It probably wouldn't hurt to try and shorten the `supportedMouseWheelZoomModifierKeys` name a bit, but I'm not attempting that here since it'd also require updating `PdfStreamConverter.jsm` in mozilla-central in order to be consistent.
For very large and complex PDF files this will help performance slightly, since `Parser.shift` is called *a lot* during parsing.
This patch was tested using the PDF file from issue 2618, i.e. http://bugzilla-attachments.gnome.org/attachment.cgi?id=226471 (with well over *four million* `Parser.shift` calls for just the one page), using the following manifest file:
```
[
{ "id": "issue2618",
"file": "../web/pdfs/issue2618.pdf",
"md5": "",
"rounds": 100,
"type": "eq"
}
]
```
This gave the following results when comparing this patch against the `master` branch:
```
-- Grouped By browser, stat --
browser | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ------------ | ----- | ------------ | ----------- | --- | ----- | -------------
Firefox | Overall | 100 | 3386 | 3322 | -65 | -1.92 | faster
Firefox | Page Request | 100 | 1 | 1 | 0 | -8.08 |
Firefox | Rendering | 100 | 3385 | 3321 | -65 | -1.92 | faster
```
Remove an unnecessary `PDFDocumentProperties.setFileSize` call, relevant for the Firefox built-in viewer, and use the "normal" code-path in `PDFViewerApplication.open` instead
Since calling `getDocument` with a `PDFDataRangeTransport` argument will always unconditionally override a manually provided `length` argument, see a1a667809f/src/display/api.js (L390-L394), this patch should thus be safe.
Attempt to significantly reduce the number of `ChunkedStream.{ensureByte, ensureRange}` calls by inlining the `this.progressiveDataLength` checks at the call-sites
This functionality is very old, and pre-dates e.g. the introduction of the `EventBus` by a number of years. Rather than attaching two callback functions to every single `PDFPageView` instance, it's thus now possible to utilize the `EventBus` such that you only need a grand total of two listeners to achieve the same result.
For the `onAfterDraw` callback the replacement is particularly simple, given that a 'pagerendered' event is already being dispatched in the appropriate spot. An added benefit here is the ability to remove the event listener, since we only ever care about *one* (arbitrary) page being rendered for the `BaseViewer.onePageRendered` promise.
For the `onBeforeDraw` callback, a new 'pagerender' event was thus added to replace the callback.
Given that this special-case only matters for the Firefox PDF viewer, it's probably better to just move it into `firefoxcom.js` instead to reduce unnecessary confusion.
The way that this method handles documents without an `ID` entry in the Trailer dictionary feels overly complicated to me. Hence this patch adds `getByteRange` methods to the various Stream implementations[1], and utilize that rather than manually calling `ensureRange` when computing a fallback `fingerprint`.
---
[1] Note that `PDFDocument` is only ever initialized with either a `Stream` or a `ChunkedStream`, hence why the `DecodeStream.getByteRange` method isn't implemented.
The `finalize` helper function has only a *single* call-site, and furthermore it's just a one-liner too. Furthermore it's only ever called with a `Promise` as its argument, meaning that it's unnecessarily convoluted as well (i.e. the `Promise.resolve()` part shouldn't be necessary).
Hence this code can be both simplified *and* inlined at its only call-site instead.
Currently `wrapReason` is manually called at *every* `resolveOrReject` call-site, despite it being completely unnecessary unless there's an actual error being handled. This is obviously inefficient, and it's easy enough to avoid by having `resolveOrReject` handle this only when actually needed.
Similar to the `zoomReset` method we need to ensure that this code won't run for zoom events originating within the browser UI itself, since checks in e.g. the `keydown` event handler won't help in that case.
Note that, in the old code, there was a code-path which could prevent this from happening thus affecting future cleanup.
Furthermore, ensure that we'll always attempt to cleanup when handling the 'PageError' message, similar to the code in e.g. the `PDFPageProxy._renderPageChunk` method.
The `receivingOperatorList` property is currently tracked *twice* in the rendering code, both directly and inversely through the `intentState.operatorList.lastChunk` boolean. This type of double bookkeeping is never a good idea, since it's just too easy for the properties to accidentally fall out of sync.
In this case there's even a `cleanup`-related bug caused by this, which means that `PDFPageProxy._tryCleanup` will never be able to discard any data if there's an error on the worker-thread (as handled through the 'PageError' message).
Hence the simplest solution seems, at least to me, to update `PDFPageProxy._tryCleanup` to replace the `intentState.receivingOperatorList` check with a `!intentState.operatorList.lastChunk` check and completely remove the former property.
When searching occurs for the first time in a document, the `textContent` of all pages will be fetched from the API. If there's a pending search operation when the document loads that will thus lead to a lot of `getTextContent` calls very early on, which may unnecessarily delay rendering of the first page. Generally, in the viewer, a number of non-essential API calls[1] will be deferred until the first page has been rendered, and there's no good reason as far as I can tell to handle searching differently.
---
[1] Such as e.g. `getOutline` and `getAttachments`.
*Please note:* A a similar change was attempted in PR 5005, but it was subsequently backed out in PR 5069.
Unfortunately I don't think anyone ever tried to debug *exactly* why it didn't work, since it ought to have worked, and having re-tested this now I'm not able to reproduce the problem any more. However, given just how inefficient the current code is, with thousands of strictly unnecessary function calls for each `find` invocation, I'd really like to try fixing this again.
This reduces the total number of function calls, when reading the XRef table respectively when fetching uncompressed XRef entries.
Note in particular the `XRef.readXRefTable` method, where there're *two* back-to-back `isCmd` checks rather than just one.
A lot of the `new Parser()` call-sites look quite unwieldy/ugly as-is, with a bunch of somewhat randomly ordered arguments, which we can avoid by changing the constructor to accept an object instead. As an added bonus, this provides better documentation without having to add inline argument comments in the code.
See https://github.com/mozilla/eslint-plugin-no-unsanitized
Since we've generally never allowed e.g. `innerHTML`, which is enforced during review, there's only one linting failure with this patch. (Which is white-listed, according to the existing comment and the fact that it's test-only code.)
Since all other `IPDFStream` implementations live in their own files, it seems reasonable for these to do so as well.
Furthermore, converts all of the relevant code to ES6 classes and updates the interface definitions to mark a couple of methods `async`.
Unless the `PDFLinkService` instance contains all of the expected methods, a lot of things will break in various places in the default viewer. Hence there's not much value in having this check, and outright falling seems more appropriate.
Finally, this also makes the return value explicit in this case, since that's consistent with the rest of the `PDFFindController._shouldDirtyMatch` method.
As have already been stated multiple times, simply increasing the printing resolution may have undesirable effects on both memory usage *and* general performance. Hence why PR 10854 did *not* add a preference, and only exposed AppOption by default in `GENERIC` builds for now.
However, considering how differently printing works in the built-in Firefox version (with `mozPrintCallback`) compared to the general default viewer, any testing done in the latter case might not be completely relevant to the first (and most important) case of the Firefox PDF Viewer.
Note that considering the implementation of `AppOptions.get`, this patch will be safe and should allow experimenting with `printResolution` in all builds of the default viewer[1]. By not, however, having `printResolution` appear in AppOptions for either the `MOZCENTRAL` or `CHROMIUM` build targets, there should be no indication of official support for now.
Furthermore, it shouldn't be a preference at this point in time (or even at all), since that makes it too easy for users to change it permanently[2] and possible "break" printing.
---
[1] By running `PDFViewerApplicationOptions.get('printResolution', /* value here */);` in the console after the viewer loads.
[2] I've seen Firefox bugs, filed in Bugzilla, where users modified e.g. preferences manually in `about:config` and then some time later (maybe months) wondered why something was suddenly broken. In those cases, trying to work out that a preference change was the culprit can take time/effort.
Given that `cleanupAfterRender` is already set for large images, when handling 'obj' messages, this patch *should* thus be safe in general (since otherwise there ought be existing bugs related to cleanup and printing).
The border `width` will instead fallback to the default value of `1`, rather than ignoring it altoghether, to also ensure that e.g. `LinkAnnotation`s become clickable as intended.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1552113
Usually when the worker is terminated it will also be completely destroyed/removed, which means that any global caches (such as the ones in `src/core/primitive.js`) should be automatically cleared in the process.
However, for certain ways of loading the `pdf.worker.js` file, e.g. passing in a re-usable worker to `getDocument`, using the `workerPort` functionality, or even disabling workers completely (even though this is never a good idea), the worker file may be kept in memory and these caches will not be cleared as expected.
Calling `someArray = []` will create a new Array, which seems completely unnecessary when it's sufficient to just call `someArray.length = 0` to achieve the same effect.
Even though I cannot imagine these particular cases having any noticeable performance impact, similar changes were made in `core/` code years ago since it's apparently more efficient memory wise.
The purpose of these caches is to reduce peak memory usage, by only ever having *a single* instance of a particular object.
However, as-is these caches are never cleared and they will thus remain until the worker is destroyed. This could very well have a negative effect on total memory usage, particularly for large/long documents, hence it seems to make sense to clear out these caches together with various other ones.
This is similar to the existing caching used to reduced the number of `Cmd` and `Name` objects.
With the `tracemonkey.pdf` file, this patch changes the number of `Ref` objects as follows (in the default viewer):
| | Loading the first page | Loading *all* the pages |
|----------|------------------------|-------------------------|
| `master` | 332 | 3265 |
| `patch` | 163 | 996 |
The specification states that `CreationDate` is only available for
markup annotations instead of for all annotation types.
Moreover, popup annotations are not markup annotations according to the
specification, so the creation date inheritance from the parent
annotation is also removed there (note that only the modification date
is used in e.g., the viewer).
This includes the information in the core and display layers. The
date parsing logic from the document properties is rewritten according
to the specification and now includes unit tests.
Moreover, missing unit tests for the color of a popup annotation have
been added.
Finally the styling of the popup is changed slightly to make the text a
bit smaller (it's currently quite large in comparison to other viewers)
and to make the drop shadow a bit more subtle. The former is done to be
able to easily include the modification date in the popup similar to how
other viewers do this.
Currently `handleColorN` will fallback to add a completely unparsed/unvalidated operator when no valid pattern was found. This is unfortunate, since it could very easily lead to a couple of different errors:
- `DataCloneError`s when attempting to send the data to the main-thread, e.g. when `args` is `Dict`/`Stream`.
- Errors in `getShadingPatternFromIR` on the main-thread, unless `args` just happens to have the expected format.
- Errors when actually attempting to render the pattern on the main-thread, since the `args` will most likely not have the expected format.
Hence it probably makes sense to error in `PartialEvaluator.handleColorN`, and having invalid patterns fail gracefully via the existing `ignoreErrors` code-paths instead.
It appears that this has been broken ever since PR 9089, which also introduced this code, since the `QueueOptimizer`/`NullOptimizer` choice was made based on the still undefined `this.intent` property.
Furthermore, fixing this also uncovered the fact that the `NullOptimizer.reset` method was missing.
First of all, while this simple approach appears to work OK in practice I'm not sure if it's the best way of addressing the problem (assuming that you even want to).
Second of all, while the solution implemented here only requires tracking/checking one new boolean in order for this to work, I'm nonetheless not entirely happy about this since it will add additional overhead (albeit *very* small) to the parsing of path operators in PDF documents just for a handful of *corrupt* ones.
This way we can avoid manually building a "document id" in multiple places in `evaluator.js`, and it also let's us avoid passing in an otherwise unnecessary `PDFManager` instance when creating a `PartialEvaluator`.
While PR 10714 did address the `disableRange=true` case, it also managed to "break" the `disableStream=true` case instead since the indeterminate loadingBar is now displayed when it shouldn't; sorry about that!
The solution is simple enough though, don't attempt to fallback to `_fullRequestReader.onProgress` when handling "incomplete" loading information.
Please see the specification, https://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#M11.9.12864.1Heading.71.Viewer.Preferences
Furthermore, note that this patch *only* adds API support and unit-tests but does not attempt to integrate e.g. the `ViewerPreferences -> Direction` property into the viewer (which would be necessary to address issue 10736).
The reason for this is that it's not entirely clear to me exactly if/how that could be implemented; e.g. would it be as simple as setting the `dir` attribute on the `viewerContainer` DOM element, or will it be more complicated?
There's also the question of how the `ViewerPreferences -> Direction` value interacts with the `PageMode`, and this will generally require a fair bit of manual testing. Since the direction of the *entire* viewer depends on the browser locale, there's also a somewhat open question regarding what default value to use for different locales.
Finally, if the viewer supports `ViewerPreferences -> Direction` then I'm assuming that it will be necessary to allow users to override the default value, which will require (most likely) new `SecondaryToolbar` buttons and icons for those etc.
Hence this patch only lays the necessary foundation for eventually addressing issue 10736, but defers the actual implementation until later. (Time permitting, I'll try to look into the viewer part later.)
*Please note:* This patch purposely ignores `src/display/network.js`, since its support for progressive reading depends on the non-standard `moz-chunked-arraybuffer` responseType which is currently in the process of being removed.
The file `test/pdfs/annotation-caret-ink.pdf` is already available in
the repository as a reference test for this since I supplied it for
another patch that implemented ink annotations.
This mirrors the canvas implementation where we ignore these operators.
This avoids console spam regarding unimplemented operators we're not
interested in.
For the Tracemonkey paper, we're now down to one warning about tiling
patterns which is in fact a valid one.
In particular, this should reduce intermediate string creation by using
template strings and reduce variable lookup times by removing unneeded
variables and caching `this.current` in more places.
With PR 10675 having fixed the completely broken `disableRange=true` setting in the Firefox version of PDF.js, I couldn't help but noticing that loading progress is never reported properly in that case.
Currently loading progress is only reported for the `rangeProgress` chrome-event, which obviously isn't dispatched with `disableRange=true` set. However, the `progressiveRead` chrome-event includes loading progress as well, but this information isn't being used in any way.
Furthermore, the `PDFDataRangeTransport.onDataProgress` method wasn't able to handle "complete" loading information, and neither was `PDFDataTransportStream._onProgress` since that method would only ever attempt to report it through a RangeReader (which won't exist when `disableRange=true` is set).
Note how at https://mozilla.github.io/pdf.js/api/ it's being described as API docs, however `src/core/annotation.js` is not part of the public API.
Furthermore, given that the code residing in the `src/core/` folder is run in a worker-thread, it's not even accessible on the main-thread (since `postMessage` is being used to transfer the data).
Hence the different API methods simply returns a "proxy" to the underlying data, but not actually the same objects and data structures as in the worker-thread itself; thus it doesn't make a whole lot of sense to expose this in API docs as far as I'm concerned.
Finally, the patch fixes a small JSDoc related typo in `src/display/api.js` when referring to the `TextStyle` typedef.
The `moz-chunked-arraybuffer` responseType is a non-standard property, which has been subsumed by the Fetch API, and it's in the process of being removed from Firefox; please see https://bugzilla.mozilla.org/show_bug.cgi?id=1120171 and https://bugzilla.mozilla.org/show_bug.cgi?id=1411865
*Please note:* Rather than waiting for both `Fetch` *and* `ReadableStream` to be available in e.g. a Firefox ESR version (which is probably going to be 68 at the earliest), let's just decide that PDF.js release `2.1.266` will be the last one with `moz-chunked-arraybuffer` support and land this patch (since nothing should outright break without it anyway).
Note how `XRef.fetchUncompressed`, which is used *a lot* for most PDF documents, is calling the `makeSubStream` method without providing a `length` argument.
In practice this results in the `makeSubStream` method, on the `ChunkedStream` instance, calling the `ensureRange` method with `NaN` as the end position, thus resulting in no data being requested despite it possibly being necessary.
This may be quite bad, since in this particular case it will lead to a new `ChunkedStream` being created *and* also a new `Parser`/`Lexer` instance. Given that it's quite possible that even the very first `Parser.getObj` call could throw `MissingDataException`, this could thus lead to wasted time/resources (since re-parsing is necessary once the data finally arrives).
You obviously need to be very careful to not have `ChunkedStream.makeSubStream` accidentally requesting the *entire* file, hence its `this.end` property is of no use here, but it should be possible to at least check that the `start` of the data is present before any potentially expensive parsing occurs.
This transform resulted in an incorrectly positioned object when the
bounding box's upper-left corner did not start at (0,0), because
the translation was not reverted. This patch adds the missing transform.
The test file (tiling-pattern-box.pdf) is based on the PDF from #2825.
All but the first cube (including the PDF data) have been removed.
To trigger the bug that is fixed by this commit, I changed the BBox of
the first pattern from "[ 0 0 596 842]" to "[90 0 596 842]". Without
this patch, the dashed vertical line that intersects the corners at A
and E would disappear.
The new test file (tiling-pattern-large-steps.pdf) was manually created,
to have the following characteristics:
- Large xstep and ystep (90000)
- Page width is 4000 (which is larger than MAX_PATTERN_SIZE)
- Visually, the page consists of a red rectangle with a black border,
surrounded by a 50 unit white padding.
- Before patch: blurry; After patch: sharp
Fixes#6496Fixes#5698Fixes#1434Fixes#2825
Without this some fonts may incorrectly end up with matching `hash`es, thus breaking rendering since we'll not actually try to load/parse some of the fonts.
Note that `PartialEvaluator.preEvaluateFont` will return an empty string when no hash was computed. This will complete short-circuit the `fontAlias` comparison in `PartialEvaluator.loadFont`, since fonts which are totally different will then match if their `hash`es are empty.
Currently if trying to set `disableRange=true` in the built-in PDF Viewer in Firefox, either through `about:config` or via the URL hash, the PDF document will never load. It appears that this has been broken for a couple of years, without anyone noticing.
Obviously it's not a good idea to set `disableRange=true`, however it seems that this bug affects the PDF Viewer in Firefox even with default settings:
- In the case where `initialData` already contains the *entire* file, we're forced to dispatch a range request to re-fetch already available data just so that file loading may complete.
- (In the case where the data arrives, via streaming, before being specifically requested through `requestDataRange`, we're also forced to re-fetch data unnecessarily.) *This part was removed, to reduce the scope/risk of the patch somewhat.*
In the cases outlined above, we're having to re-fetch already available data thus potentially delaying loading/rendering of PDF files in Firefox (and wasting resources in the process).
Support for the non-standard `moz-chunked-arraybuffer` response type is in the process of being removed from Firefox; see e.g. https://bugzilla.mozilla.org/show_bug.cgi?id=1411865
For the time being, you probably want to keep support for this in the general PDF.js library given that feature detection is used. However, removing the unit-test immediately seems reasonable, since it will otherwise start failing once the platform support for `moz-chunked-arraybuffer` is gone.
Fixes 8851; please note that if unit-tests for the code in `fetch_stream.js` are wanted, which I'm assuming they are, those should live in their own file rather than being lumped into `network_spec.js` anyway.
This lays the necessary foundation for handling zoom events originating within the browser itself, rather than in the viewer. Please note that this will also require a follow-up patch to `mozilla-central`, such that the viewer is actually notified when zooming occurs.
Moreover, group the lexer unit tests per method. This matches what we
do for other classes and makes it more easily visible which methods
we don't or insufficiently unit test.
The parser itself is not unit tested yet, so this patch provides a start
for doing so. The `inlineStreamSkipEI` method is used in other end
marker detection methods, so it's important that its functionality is
correct for proper parsing.
This function is currently called with the `OperatorList` instance as its argument, hence I cannot think of any good reason for not just moving it into the `OperatorList` properly. (This will also help with other planned changes regarding the `ImageCache` functionality.)
By transfering `ArrayBuffer`s you can avoid having two copies of the same data, i.e. one copy on each of the worker/main-thread, for data that's used only *once* on the worker-thread.
Note how the code in [`PDFImage.createMask`](80135378ca/src/core/image.js (L284-L285)) goes to great lengths to actually enable tranfering of the image data. However in [`PartialEvaluator.buildPaintImageXObject`](80135378ca/src/core/evaluator.js (L336)) the `cached` property is always set to `true`, which disqualifies the image data from being transfered; see [`getTransfers`](80135378ca/src/core/operator_list.js (L552-L554)).
For most ImageMask data this patch won't matter, since images found in the `/Resources -> /XObject` dictionary will always be indexed by name. However for *inline* images which contains ImageMask data, where only "small" images are cached (in both `parser.js` and `evaluator.js`), the current code will result in some unnecessary memory usage.
This will further help reduce the amount of image data that's currently being held alive, by explicitly removing the `src` attribute.
Please note that this is mostly relevant for browsers which do not support `URL.createObjectURL`, or where `disableCreateObjectURL` was manually set by the user, since `blob:` URLs will be revoked (see the previous patch).
However, using `about:memory` (in Firefox) it does seem that this may also be generally helpful, given that calling `URL.revokeObjectURL` won't invalidate the image data itself (as far as I can tell).
Natively supported JPEG images are sent as-is, using a `blob:` or possibly a `data` URL, to the main-thread for loading/decoding.
However there's currently no attempt at releasing these resources, which are held alive by `blob:` URLs, which seems unfortunately given that images can be arbitrarily large.
As mentioned in https://developer.mozilla.org/en-US/docs/Web/API/URL/createObjectURL the lifetime of these URLs are tied to the document, hence they are not being removed when a page is cleaned-up/destroyed (e.g. when being removed from the `PDFPageViewBuffer` in the viewer).
This is easy to test with the help of `about:memory` (in Firefox), which clearly shows the number of `blob:` URLs becomming arbitrarily large *without* this patch. With this patch however the `blob:` URLs are immediately release upon clean-up as expected, and the memory consumption should thus be considerably reduced for long documents with (simple) JPEG images.
Note how `PDFDocumentProxy.destroy` is a nothing more than an alias for `PDFDocumentLoadingTask.destroy`. While removing the latter method would be a breaking API change, there's still room for at least some clean-up here.
The main changes in this patch are:
- Stop providing a `PDFDocumentLoadingTask` instance *separately* when creating a `PDFDocumentProxy`, since the loadingTask is already available through the `WorkerTransport` instance.
- Stop tracking the `PDFDocumentProxy` instance on the `WorkerTransport`, since that property is completely unused.
- Simplify the 'Multiple `getDocument` instances' unit-tests by only destroying *once*, rather than twice, for each document.
For Type3 fonts text-selection is often not that great, and there's a couple of heuristics used to try and improve things. This patch simple extends those heuristics a bit, and fixes a pre-existing "naive" array comparison, but this all feels a bit brittle to say the least.
The existing Type3 test-coverage isn't that great in general, and in particular Type3 `text` tests are few and far between, hence why this patch adds *two* different new `text` tests.
Notable changes:
- Remove the `return this;` from the `MurmurHash3_64.update` method, since it's completely unused and doesn't make a lot of sense.
- Remove the loop(s) from the `MurmurHash3_64.hexdigest` method, since creating a temporary array and then looping over it is wasteful given how simple this can be written with modern JavaScript.
The `DOMCanvasFactory` class is now fully covered. Moreover, missing
cases for the `getFilenameFromUrl` function have been included.
Finally, `var` usage has been removed.
The generated `default_preferences.json` file is necessary when initializing the Firefox preferences, which only supports certain types, hence this patch adds additional validation to help prevent run-time errors in Firefox.
Given that these changes add a code-path to `AppOptions.getAll` which could throw, the `OptionKind.PREFERENCE` branch is also modified to require *exact* matching to prevent (future) errors in the viewer.
Finally the conditionally defined `defaultOptions` will no longer (potentially) be considered during the `gulp default_preferences` task, to make it more difficult for them to be accidentally included.
Given that the function is (purposely) independent of the verbosity level and that its message is worded to only apply on the main-thread, there's no reason to duplicate this across the built `pdf.js`/`pdf.worker.js` files.
Currently for every single parsed/rendered page there's no less than *four* `Date.now()` calls being made on the worker-side. This seems totally unnecessary, since the result of these calls are, by default, not used for anything *unless* the verbosity level is set to `INFO`.
The default size of these canvases seem to be `300 x 150` (two orders of magnitude larger than the ones in PR 10597), which probably is sufficient enough to matter since there's one such canvas for each textLayer that's rendered in the viewer.
Also fixes the incorrect rejection reason, i.e. one using a string rather than an `Error`, in the `TextLayerRenderTask.cancel` method.
While this particular canvas may be small, there can still be an arbitrarily large number of them (one per page rendered), which can/will eventually add up memory wise. This can be easily avoided by using the `cachedCanvases` abstraction instead, which will ensure that the `isFontSubpixelAAEnabled` canvas is removed together with other temporary canvases in `CanvasGraphics.endDrawing`.
The `src/shared/util.js` file is being bundled into both the `pdf.js` and `pdf.worker.js` files, meaning that its code is by definition duplicated.
Some main-thread only utility functions have already been moved to a separate `src/display/display_utils.js` file, and this patch simply extends that concept to utility functions which are used *only* on the worker-thread.
Note in particular the `getInheritableProperty` function, which expects a `Dict` as input and thus *cannot* possibly ever be used on the main-thread.
In order to, quickly, unblock future Webpack updates this patch implements a *temporary* work-around for the regression; please refer to issue 10177 for additional details.
Please note that this patch doesn't require *any* changes to the source code, but only (small) build-time changes to `gulpfile.js` and as such will be easy to revert.
This file (currently) contains not only DOM-specific helper functions/classes, but is used generally for various helper code relevant for main-thread functionality.
*Hopefully this patch makes sense, since I cannot claim to fully understand this function.*
With the changes made in PR 3354 *some* Type3 glyph outlines are no longer rendering correctly, since the final paths were being accidentally ignored.
The fact that Type3 fonts are not very common in PDF documents, and that most Type3 glyphs are unaffected by this regression, probably explains why this has gone unnoticed since 2013.
There doesn't appear to be any particular reason for only running these unit-tests in browsers, since the `PDFDataRangeTransport` functionality itself should be back-end agnostic.
This allows simplification of the 'creates pdf doc from URL and aborts loading after worker initialized' API unit-test.
Note that the `DOMFileReaderFactory` uses the Fetch API, for simplicity, since it should be available in all browsers where we're running tests.
Currently these methods are only used from the respective `reset` methods, and from `{BaseViewer, PDFThumbnailViewer}._cancelRendering` which only runs when the active document is closed.
This patch changes `{PDFPageView, PDFThumbnailView}.cancelRendering` to *only* cancel any pending rendering operations, and doesn't attempt to reset e.g. the `renderingState`, since that causes visual glitches (duplicated canvases in the viewer) when called directly.
Furthermore, unless you "know" what you're doing, the `{PDFPageView, PDFThumbnailView}.reset` methods are what *should* normally be used instead.
This implements the nice suggestion from https://github.com/mozilla/pdf.js/pull/10099#discussion_r219698000, which I at the time didn't think would work.
I'll probably have to plead temporary insanity here, since it *should* have been totally obvious to me how this could be implemented. By simply not registering the event until the `textLayer` is actually rendered, removing the event on `cancel` works just fine.
This patch also removes the `pagecancelled` event, given that it's no longer used anywhere in the code-base and that its implemention was flawed since it wasn't guaranteed to be called in *every* situation where rendering was actually cancelled.
Currently any editing of the preferences require updates in *three* separate files, which isn't a great developer experience to say the least.
This has annoyed me sufficiently to write this patch, which moves the definition of all preferences into `AppOptions` and adds a new `gulp` task to generate the `default_preferences.json` file for the builds where it's needed.
After PR 9340 all glyphs are now re-mapped to a Private Use Area (PUA) which means that if a font fails to load, for whatever reason[1], all glyphs in the font will now render as Unicode glyph outlines.
This obviously doesn't look good, to say the least, and might be seen as a "regression" since previously many glyphs were left in their original positions which provided a slightly better fallback[2].
Hence this patch, which implements a *general* fallback to the PDF.js built-in font renderer for fonts that fail to load (i.e. are rejected by the sanitizer). One caveat here is that this only works for the Font Loading API, since it's easy to handle errors in that case[3].
The solution implemented in this patch does *not* in any way delay the loading of valid fonts, which was the problem with my previous attempt at a solution, and will only require a bit of extra work/waiting for those fonts that actually fail to load.
*Please note:* This patch doesn't fix any of the underlying PDF.js font conversion bugs that's responsible for creating corrupt font files, however it does *improve* rendering in a number of cases; refer to this possibly incomplete list:
[Bug 1524888](https://bugzilla.mozilla.org/show_bug.cgi?id=1524888)
Issue 10175
Issue 10232
---
[1] Usually because the PDF.js font conversion code wasn't able to parse the font file correctly.
[2] Glyphs fell back to some default font, which while not accurate was more useful than the current state.
[3] Furthermore I'm not sure how to implement this generally, assuming that's even possible, and don't really have time/interest to look into it either.
- The only existing call-site, of this method, is never passing more than *one* font at a time anyway.
- As far as I can remember, this functionality has never actually been used (caveat: I didn't check the git history).
- This allows simplification of the method, especially by making use of the fact that it's now asynchronous.
- It should be just as easy to call `BaseFontLoader.bind` from within a loop, rather than having the loop in the method itself.
Currently all fonts are using the `_queueLoadingCallback` method to determine when they have been loaded[1]. However in most cases this is just adding unnecessary overhead, especially with `BaseFontLoader.bind` now being asynchronous, given how fonts are loaded:
- For fonts loaded using the Font Loading API, it's already possible to easily tell when a font has been loaded simply by checking the `loaded` promise on the FontFace object itself.
- For browsers, e.g. Firefox, which support synchronous font loading it's already assumed that fonts are immediately available.
Hence the `_queueLoadingCallback` method is moved into the `GenericFontLoader`, such that it's only utilized for fonts which are loaded using CSS.
---
[1] In the "fonts loaded using CSS" case, this is already a hack anyway as outlined in the comments.
pdf.js had a problem when copying characters on supplementary planes
(0xPPXXXX where PP is nonzero). This is because certain methods of
PartialEvaluator use classic String.fromCharCode instead of ES6's
String.fromCodePoint.
Despite the fact that readToUnicode method *tried* to parse out-of-UCS2
code points by parsing UTF-16BE, it was inadequate because
String.fromCharCode only supports UCS-2 range of Unicode.
Unsurprisingly IE11 doesn't support this, so a polyfill is needed since otherwise the sidebar can no longer be opened.
Also, simplifies the existing `classList.toggle` polyfill.
There's a bunch of code, in the viewer, which for historical reasons use `switch` statements to add and remove CSS classes.
This code can be simplified, and unnecessary duplication avoided, by using `classList.toggle` instead.
This polyfill is currently used in only *one* file, i.e. `src/display/api.js`, and only when trying to build a *fallback* `workerSrc` path.
Given that the global `workerSrc` should *always* be set[1] when using the PDF.js library[2], and that the fallback `workerSrc` should only be regarded as a best-effort solution anyway, there isn't a particularily strong reason to keep the compatibility code in my opinion.
---
[1] Other supported options include setting the global `workerPort`, or passing in a `PDFWorker` instance as part of the `getDocument` call.
[2] Which is clearly mentioned in the JSDocs in `src/display/worker_options.js`.
This avoids having the initialization code "spread out", and will become even simpler once the `TODO` is addressed (which I'm planning on fixing as soon as possible).
This patch ignores the recently added `disableOpenActionDestination` preference, since the latest PDF.js version found on the "Chrome Web Store" doesn't include it.
The intention with preferences such as `sidebarViewOnLoad`/`scrollModeOnLoad`/`spreadModeOnLoad` were always that they should be able to *unconditionally* override their view history counterparts.
Due to the way that these preferences were initially implemented[1], trying to e.g. force the sidebar to remain hidden on load cannot be guaranteed[2]. The reason for this is the use of "enumeration values" containing zero, which in hindsight was an unfortunate choice on my part.
At this point it's also not as simple as just re-numbering the affected structures, since that would wreak havoc on existing (modified) preferences. The only reasonable solution that I was able to come up with was to change the *default* values of the preferences themselves, but not their actual values or the meaning thereof.
As part of the refactoring, the `disablePageMode` preference was combined with the *adjusted* `sidebarViewOnLoad` one, to hopefully reduce confusion by not tracking related state separately.
Additionally, the `showPreviousViewOnLoad` and `disableOpenActionDestination` preferences were combined into a *new* `viewOnLoad` enumeration preference, to further avoid tracking related state separately.
The new "private" method will return a boolean, indicating if the `sidebarviewchanged` event was dispatched, thus allowing some simplification of the `PDFSidebar.setInitialView` method.
When `firstVisibleElementInd === 0`, regardless of the number of views, there's no reason to attempt to backtrack at all since it's never possible to find an element before the *first* one anyway.
This piggybacks of the existing `cancel` functionality, to ensure that any pending operations are closed *and* that any temporary canvases are actually being removed.
Also simplifies `finishPaintTask` in `PDFPageView.draw` slightly, by converting it to an async function.
All objects in the PDF document follow this pattern:
```
0000000001 0 obj
<<
% Some content here...
>>
endobj
0000000002 0 obj
<<
% More content here...
endobj
```
This attempts to provide more "default" methods in the base class, in order to reduce unnecessary duplication and to improve self-documentation of the `BaseViewer` class slightly.
The following changes are made (in no particular order):
- Have `BaseViewer` implement the `_scrollIntoView` method, and *extend* it as necessary in `PDFViewer`/`PDFSinglePageViewer`.
- Simply inline the `BaseViewer._resizeBuffer` method, in `BaseViewer.update`, since there's only one call-site at this point.
- Provide a default implementation of `_isScrollModeHorizontal` in `BaseViewer`, and have `PDFSinglePageViewer` override it.
- Provide a default implementation of `_getVisiblePages`, and have `PDFViewer` extend it and `PDFSinglePageViewer` override it.
Based on the discussion in https://bugzilla.mozilla.org/show_bug.cgi?id=1521413, this patch simply removes the `ReadableStream` polyfill completely from MOZCENTRAL builds.
With this patch, the size of the `gulp mozcentral` build target is thus further reduced (building on PR 10470):
| | `build/mozcentral`
|-------|-------------------
|master | 3 339 666
|patch | 3 209 572
The `FontInspector` was completely broken by PR 10197, and trying to select a particular font (using the checkboxes) or clicking on a piece of text currently does nothing.
With https://bugzilla.mozilla.org/show_bug.cgi?id=1505122 landing in Firefox 65, the native `ReadableStream` implementation is now enabled by default in Firefox.
Obviously it would be nice to simply stop bundling the polyfill in MOZCENTRAL builds altogether, however given that it's still possible to disable[1] `ReadableStream` this is probably not a good idea just yet.
Nonetheless, now that native support is available, it seems unnecessary (and wasteful) to keep bundling the polyfill twice[2] in MOZCENTRAL builds. Hence this patch, which contains a suggest approach for packing the polyfill in a *separate* file which is then *only* loaded if/when needed.
With this patch, the size of the `gulp mozcentral` build target is thus reduced accordingly:
| | `build/mozcentral`
|-------|-------------------
|master | 3 461 089
|patch | 3 340 268
Besides the PDF.js files taking up less space in Firefox this way, the additional benefit is that there's (by default) less code that needs to be loaded and parsed when the PDF Viewer is used which also cannot hurt.
---
[1] In `about:config`, by toggling the `javascript.options.streams` preference.
[2] Once in the `build/pdf.js` file, and once in the `build/pdf.worker.js` file.
In many cases in the code you don't actually care about the index itself, but rather just want to know if something exists in a String/Array or if a String starts in a particular way. With modern JavaScript functionality, it's thus possible to remove a number of existing `indexOf` cases.
Since most of the important rendering code is already (almost) identical between `PDFViewer.update` and `PDFSinglePageViewer.update`, it's possible to further reduce duplication by moving the code into `BaseViewer.update` instead.
Apparently in IE 11 when `history.{pushState, replaceState}` is called, there's actually a difference between not providing the *third* argument vs providing it set implicitly to `undefined`. It appears that in IE 11 it's actually being stringified, rather than ignored, which seems completely wrong (obviously other browsers aren't affected, so no surprises there).
This is yet another reason why I think the feature itself was a really bad idea, since it now requires extra/duplicated code just to prevent weird/incorrect URL behaviour in crappy browsers.
This will allow the Metadata to be successfully extracted from the PDF file in issue 10395.
Furthermore, this patch also fixes a bug in `Metadata.get` which causes the method to return `null` rather than an empty string or zero (since either ought to be allowed).
Previously a couple of different attempts at fixing this problem has been rejected, given how *crucial* this code is for the correct function of the viewer, since no one has thus far provided any evidence that the problem actually affects the default viewer[1] nor an example using the viewer components directly (without another library on top).
The fact that none of the prior patches contained even a *simple* unit-test probably contributed to the unwillingness of a reviewer to sign off on the suggested changes.
However, it turns out that it's possible to create a reduced test-case, using the default viewer, that demonstrates the error[2]. Since this utilizes a hidden `<iframe>`, please note that this error will thus affect Firefox as well.
Note that while errors are thrown when the hidden `<iframe>` loads, the default viewer doesn't break completely since rendering does start working once the `<iframe>` becomes visible (although the errors do break the initial Toolbar state).
Before making any changes here, I carefully read through not just the immediately relevant code but also the rendering code in the viewer (given it's dependence on `getVisibleElements`). After concluding that the changes should be safe in general, the default viewer was tested without any issues found. (The above being much easier with significant prior experience of working with the viewer code.)
Finally the patch also adds new unit-tests, one of which explicitly triggers the relevant code-path and will thus fail with the current `master` branch.
This patch also makes `PDFViewerApplication` slightly more robust against errors during document opening, to ensure that viewer/document initialization always completes as expected.
Please keep in mind that even though this patch prevents an error in `getVisibleElements`, it's still not possible to set the initial position/zoom level/sidebar view etc. when the viewer is hidden since rendering and scrolling is completely dependent[3] on being able to actually access the DOM elements.
---
[1] And hence the PDF Viewer that's built-in to Firefox.
[2] Copy the HTML code below and save it as `iframe.html`, and place the file in the `web/` folder. Then start the server, with `gulp server`, and navigate to http://localhost:8888/web/iframe.html
```html
<!DOCTYPE html>
<html>
<head>
<title>Iframe test</title>
<script>
window.onload = function() {
const button = document.getElementById('button1');
const frame = document.getElementById('frame1');
button.addEventListener('click', function(evt) {
frame.hidden = !frame.hidden;
});
};
</script>
</head>
<body>
<button id="button1">Toggle iframe</button>
<br>
<iframe id="frame1" width="800" height="600" src="http://localhost:8888/web/viewer.html" hidden="true"></iframe>
</body>
</html>
```
[3] This is an old, pre-exisiting, issue that's not relevant to this patch as such (and it's already being tracked elsewhere).
With modern JavaScript supporting block-scoped variables, it's no longer necessary to have large numbers of top-level variable definitions (especially when all variables are, implicitly, set to `undefined`).
Besides moving variable definintions, a number of them are also converted to `const` to help ensure that they cannot be *accidentally* modified in the code.
Finally, the `views.length` calculation is now cached *once* rather than being re-computed multiple times.
Handle more cases of corrupt PDF files with missing 'endobj' operators, where the "obj" string is immediately followed by the dictionary (PR 9288 follow-up)
This is *really* the best that we can do here, since other proposed solutions would interfere with (and break) the painstakingly implemented browsing history that's present in the default viewer.
I'm still not convinced that this is a good idea in general, but this patch implements it in a way where it is possible to toggle[1] for users that wish to have this feature. In particular, there's a couple of reasons why I'm not finding this feature necessary/great:
- It's already possible to easily obtain the current hash, by simply clicking on the `viewBookmark` button at any time.
- Hash changes requires a bit of special handling[2], i.e. extra code, to prevent issues when the browser history is traversed (see `PDFHistory._popState`). Currently this is only necessary when the user has manually changed the hash, with this patch it will always be the case (assuming the feature is active).
- It's not always possible to change the URL when updating the browser history. For example: In the Firefox built-in viewer, the URL cannot be modified for local files (i.e. those using the `file://` protocol).
This leads to inconsistent behaviour, and may in some cases even result in errors being thrown and the history thus not updating, if the browser prevents changes to the URL during `pushState`/`replaceState` calls.
---
[1] Using the `historyUpdateUrl` viewer preference.
[2] This depends, to a great extent, on browsers always firing `popstate` events *before* `hashchange` events, which may or may not actually be guaranteed.
This should hopefully be sufficient to address issue 6847, and given the limited impact of the code changes I'm not completely sure if this would need to be controlled by a preference!?
Initially my intention was to try and provide some (slightly more detailed) implementation suggestions in the issue, but having looked briefly at doing that it would essentially have amounted to actually writing the code anyway. (Especially considering that the recent questions seemed to more-or-less ignore the information already provided in the first post.)
Finally, note that since `performance.navigation.type` is marked as deprecated, a slightly different approach was choosen instead.
The error was triggered for a particular set of metadata, where an end tag was encountered without the corresponding begin tag being present in the data.
(The patch also fixes a minor oversight, from a recent PR, in the `SimpleDOMNode.nextSibling` method.)
Given that the issue, as filed, is incomplete since no PDF file was provided for debugging, this patch is really the best that we can do here. *Please note:* This patch will *not* enable the Metadata to be successfully parsed, but it should at least prevent the errors.
This method creates quite a few intermediate strings on each call and
it's called often, even for smaller documents like the Tracemonkey
document. Scrolling from top to bottom in that document resulted in
14126 strings being created in this method. With this commit applied,
this is reduced to 2018 strings.
This method creates quite a few intermediate strings on each call and
it's called often, even for smaller documents like the Tracemonkey
document. Scrolling from top to bottom in that document resulted in
12936 strings being created in this method. With this commit applied,
this is reduced to 3610 strings.
The `toString` method always creates two string objects (for the 'R'
character and for the `num` concatenation) and in the worst case
creates three string objects (one more for the `gen` concatenation).
For the Tracemonkey paper alone, this resulted in 12000 string
objects when scrolling from the top to the bottom of the document.
Since this is a hot function, it's worth minimizing the number of string
objects, especially for large documents, to reduce peak memory usage.
This commit refactors the `toString` method to always create only one
string object.
For PDF documents with sufficiently broken XRef tables, it's usually quite obvious when you need to fallback to indexing the entire file. However, for certain kinds of corrupted PDF documents the XRef table will, for all intents and purposes, appear to be valid. It's not until you actually try to fetch various objects that things will start to break, which is the case in the referenced issues[1].
Since there's generally a real effort being in made PDF.js to load even corrupt PDF documents, this patch contains a suggested approach to attempt to do a bit more validation of the XRef table during the initial document loading phase.
Here the choice is made to attempt to load the *first* page, as a basic sanity check of the validity of the XRef table. Please note that attempting to load a more-or-less arbitrarily chosen object without any context of what it's supposed to be isn't a very useful, which is why this particular choice was made.
Obviously, just because the first page can be loaded successfully that doesn't guarantee that the *entire* XRef table is valid, however if even the first page fails to load you can be reasonably sure that the document is *not* valid[2].
Even though this patch won't cause any significant increase in the amount of parsing required during initial loading of the document[3], it will require loading of more data upfront which thus delays the initial `getDocument` call.
Whether or not this is a problem depends very much on what you actually measure, please consider the following examples:
```javascript
console.time('first');
getDocument(...).promise.then((pdfDocument) => {
console.timeEnd('first');
});
console.time('second');
getDocument(...).promise.then((pdfDocument) => {
pdfDocument.getPage(1).then((pdfPage) => { // Note: the API uses `pageNumber >= 1`, the Worker uses `pageIndex >= 0`.
console.timeEnd('second');
});
});
```
The first case is pretty much guaranteed to show a small regression, however the second case won't be affected at all since the Worker caches the result of `getPage` calls. Again, please remember that the second case is what matters for the standard PDF.js use-case which is why I'm hoping that this patch is deemed acceptable.
---
[1] In issue 7496, the problem is that the document is edited without the XRef table being correctly updated.
In issue 10326, the generator was sorting the XRef table according to the offsets rather than the objects.
[2] The idea of checking the first page in particular came from the "standard" use-case for the PDF.js library, i.e. the default viewer, where a failure to load the first page basically means that nothing will work; note how `{BaseViewer, PDFThumbnailViewer}.setDocument` depends completely on being able to fetch the *first* page.
[3] The only extra parsing is caused by, potentially, having to traverse *part* of the `Pages` tree to find the first page.
If, as PR 10368 suggests, more parameters should be added to `getViewport` I think that it would be a mistake to not change the signature *first* to avoid needlessly unwieldy call-sites.
To not break any existing code and third-party use-cases, this is obviously implemented with a deprecation warning *and* with a working fallback[1] for the old method signature.
---
[1] This is limited to `GENERIC` builds, which should be sufficient.
Given that it's really not clear to me if this is actually desired functionality in the default viewer, and considering that it doesn't fit in *great* with the way that `PDFHistory` is initialized, this feature is currently off by default[1].
---
[1] It's controlled with the `disableOpenActionDestination` Preference/AppOption.
Note that the OpenAction dictionary may contain other information besides just a destination array, e.g. instructions for auto-printing[1].
Given first of all that an arbitrary `Dict` cannot be sent from the Worker (since cloning would fail), and second of all that the data obviously needs to be validated, this patch purposely only adds support for fetching a destination from the OpenAction entry[2].
---
[1] This information is, currently in PDF.js, being included through the `getJavaScript` API method.
[2] This significantly reduces the complexity of the implementation, which seems fine for now. If there's ever need for other kinds of OpenAction to be fetched, additional API methods could/should be implemented as necessary (could e.g. follow the `getOpenActionWhatever` naming scheme).
The custom entries, provided that they exist *and* that their types are safe to include, are exposed through a new `Custom` infoDict entry to clearly separate them from the standard ones.
Fixes 5970.
Fixes 10344.
This required the following changes in the Gulpfile:
- Defining a series of tasks is no longer done with arrays, but with the
`gulp.series` function. The `web` target is refactored to use a
smaller number of tasks to prevent tasks from running multiple times.
- Getting all tasks must now be done through the task registry.
- Tasks that don't return anything must call `done` upon completion.
Moreover, this upgrade allows us to use the latest Node.js on Travis CI
again.
Given that Signature (Widget) annotations are currently not supported, since they cannot be validated, simply ignoring the `fieldValue` seems OK for now considering that attempting to blindly include unparsed/unvalidated data isn't very useful.
Fixes 10347.
This patch re-factors, and extends, the already existing `zoomDisabledTimeout` used during mouse wheel zooming.
Unfortunately I haven't got the required hardware to actually test this patch, but there's a decent chance that it will fix, or at least reduce, the problems reported in https://bugzilla.mozilla.org/show_bug.cgi?id=1503412.
Given that a larger number of pages may now be visible at once, and importantly that their layout may be non-vertical, one of the conditions should be tweaked to not accidentally miss cases where a page is still visible.
Please note: This patch is based on code-inspection, and the only ill effect occurring without it would be a couple of (near) duplicate history entries in some *rare* edge-cases.
This patch does three things:
- Updates the `gulp unittestcli` command, using `gulp lint` as a guide, such that it can be run locally on Windows without any modifications.
- Updates the `gulp lib` command to support disabling of Babel through the `SKIP_BABEL` environment variable. Note that all other build targets support this mode, and there's no good reason for `lib` to be any different here.
- Updates the `npm test` command, used in Node.js/Travis, to test the code as-is test. Since modern Node.js versions seem to have no problems with ES6 compatible code in general, we should just test the source code as-is instead (similar to the tests running on the regular bots).
With the removal of the global `PDFJS` object, in PDF.js version `2.0`, the viewer options are no longer as easily accessible as they previously were (and issues have been filed about this).
In particular, since the viewer files aren't necessarily loaded *immediately*, this means that `PDFViewerApplication`/`PDFViewerApplicationOptions` aren't necessarily available directly. By dispatching an event once all viewer files are loaded but *before* the viewer initialization has run, setting `AppOptions` during load (in custom implementations of the default viewer) should hopefully become a little bit easier[1].
---
[1] In hindsight, this should probably have been implemented when the global `PDFJS` object was removed...
Rather than having a (somewhat) randomly choosen list of Preferences which `AppOptions` are allowed to override, it makes much more sense to simply add an AppOption to allow custom implementations to ignore Preferences altogether (it's also inline with the AppOption that allows the `ViewHistory` to be bypassed on load).
This should have been part of the previous commit that updated the
Wintersmith dependency. Markdown support is no longer included in Pug
itself and should be done by a transformer instead.
The intent of the code, based on existing comments, is to perform a binary search. However, because of what appears to be a typo in the code responsible for computing the current search index, this code is always checking *every* entry (albeit only at the "final" node) starting from the last one.
Update configuration to work with Babel 7
Explicitly require globals - eslint-plugin-mozilla needs it, but doesn't require it on its own.
Fix Regexp to match Babel 7's inlined _interopRequireDefault
According to the specification, see https://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#G6.2384179, the keys of NameTree/NumberTree should be ordered.
For corrupt PDF files, which violate this assumption, we thus need to fallback to an exhaustive search in order to e.g. find all destinations.
*Please note:* Given that this only implements a fallback for the "final" node of the Tree, there's obviously a risk that the patch isn't sufficient for dealing with all kinds of out-of-order corruption. However, this kind of problem should be rare in practice, and without a real-world test-case it's difficult to implement a completely general solution (and there's obviously a question if you'd even want to).
This changes all occurrences of `var` to `let`/`const` in this code, and updates the signature of the constructor to use object destructuring for better readability (and self documentation).
Also, `useRequestAnimationFrame` is changed to a parameter and the `typeof window` check is now done *once* rather than at every `_scheduleNext` call.
This changes all occurrences of `var` to `let`/`const` in this code, and updates the signatures of a couple of methods to use object destructuring.
Finally, when creating `InternalRenderTask` instances *only* the necessary parameter are now provided, since passing through the `RenderParameters` as-is seems completely unnecessary.
Jasmine recommends to use the `configure` method on the environment
during boot. This commit makes the code correspond to how it's done in
Jasmine's default boot file. The options dropdown in the HTML reporter
now works again after these changes, because this broke in the upgrade
to Jasmine 3, and the unit tests are executed in a random order by
default, which is important to make sure the unit tests are
self-contained and don't depend on the result of another unit test.
Rather than closing [bug 1505824](https://bugzilla.mozilla.org/show_bug.cgi?id=1505824) as WONTFIX (which is my preferred solution), given how *minor* this "problem" is, it's still possible to adjust the error messages a bit.
The main point here, which is relevant even if the changes in `BaseViewer` are ultimately rejected during review, is that we'll no longer attempt to call `BaseViewer.currentPageLabel` with an empty string from `webViewerPageNumberChanged` in `app.js`.
The other changes are:
- Stop printing an error in `BaseViewer._setCurrentPageNumber`, and have it return a boolean indicating if the page is within bounds.
- Have the `BaseViewer.{currentPageNumber, currentPageLabel}` setters print their own errors for invalid pages.
- Have the `BaseViewer.currentPageLabel` setter no longer depend, indirectly, on the `BaseViewer.currentPageNumber` setter.
- Improve a couple of other error messages.
When `highlightAll` is *not* set, there's only going to be a single match per page and unconditionally calling `PDFFindController.scrollMatchIntoView` doesn't really matter.
However, when `highlightAll` is set the current code may result in a large number of unnecessary `PDFFindController.scrollMatchIntoView` calls. Since `TextLayerBuilder._renderMatches` already checks if a particular match is the selected one, for highlighting purposes, it's simple enough to also skip scrolling completely for non-selected matches.
Only scroll search results into view as a result of an actual find operation, and not when the user scrolls/zooms/rotates the document (bug 1237076, issue 6746)
Most of the code in `PDFFindController` assumes that a valid `state` always exits, hence it cannot hurt to add a simple check to avoid errors being thrown.
Currently searching, and particularily highlighting of search results, may interfere with subsequent user-interactions such as scrolling/zooming/rotating which can result in a somewhat jarring UX where the document suddenly "jumps" to a previous position.
This is especially annoying in cases where the highlighted search result isn't even visible when a user initiated scrolling/zooming/rotating happens, and there exists a couple of bugs/issues about this behaviour.
It seems reasonable, as far as I'm concerned, to treat searching as one operation and any subsequent non-search user interactions with the viewer as separate and thus not scroll the current search result into view *unless* the user is actually doing another search.
This also seems consistent with general searching in e.g. Firefox and Adobe Reader:
- Compare with "regular" searching of e.g. HTML files in Firefox, where the user scrolling and/or zooming the document will not force a currently highlighted search result to become re-scrolled into view.
- Compare also with Adobe Reader, where the user scrolling, zooming, and/or rotating the document will not force the currently highlighted search result to become re-scrolled into view.
The question is then why search highlighting was implemented this way in PDF.js to begin with. It might be that this wasn't really intended behaviour, but more a consequence of the asynchronous nature of the API. Considering that most operations, such as fetching the page, rendering it and extracting its text-content are all asynchronous; searching and highlighting of matches thus becomes asynchronous too.
However, it should be possible to track when search results have been scrolled into view and highlighted, and thus prevent these wierd "jumps" when the user interacts with the document.
*Please note:* Unfortunately this required moving the scrolling of matches back into `PDFFindController`, since I simply couldn't see any other (reasonable) way of implementing the functionality without tracking the `_shouldScroll` property in only *one* spot.
However, given that the new `PDFFindController.scrollMatchIntoView` method follows a similar pattern as `BaseViewer.scrollPageIntoView` and `PDFThumbnailViewer.scrollThumbnailIntoView`, this is hopefully deemed OK.
Given that the `_updateScrollMode`/`_updateSpreadMode` methods are "private", there's no particular reason to not just directly call `_setCurrentPageNumber`.
Note that when e.g. presentation mode is active, we fail[1] to ensure that the `pageNumber` parameter is actually an integer before calling `_setCurrentPageNumber` (that method expects the argument be an integer).
Also changes the method signature, of `scrollPageIntoView`, to use object destructuring instead.
---
[1] Most likely, this is actually *my* oversight :-)
These interfaces are already used in different files, in both the `src/core/` and `src/display/` folders, and having them reside in their own file seems a lot clearer and is also similar to the existing viewer interfaces.
As part of moving the `interface` definitions, they're also converted to ES6 classes.
First of all, note how there's currently *two* methods for checking if a certain object exists, which seems completely unwarranted.
Furthermore, the rarely used `getData` method was removed and its only callsite changed to use a combination of `PDFObjects.{has, get}` instead.
Finally, the methods were rearranged slightly, to bring the most important ones (for an API user) to the top of the class.
Currently we'll only attempt to start from the current page when a new search is done, however for 'findagain' operations we'll always continue from the last match position.
This could easily lead to confusing behaviour if the user has scrolled to a completely different part of the document. In an attempt to improve this somewhat, for repeated 'findagain' operations, we'll instead reset the position to the current page when it's *absolutely* certain that the user has scrolled.
Note that this required adding a new `BaseViewer` method, and exposing that through `PDFLinkService`, in order to check if a given page is visible.
In an attempt to avoid issues, in custom implementations of `PDFFindController`, the code checks for the existence of the `PDFLinkService.isPageVisible` method *before* using it.
Unfortunately the `PDFFindController.executeCommand` method has now become a bit more complicated than one would like, but hopefully this small change will improve the structure somewhat (especially for subsequent patches).
Remove the `attachDOMEventsToEventBus` functionality, since `EventBus` instances are able to re-dispatch events to the DOM (PR 10019, bug 1492849 follow-up)
With only *one* event now being dispatched when the scale changes, in combination with there only being two call-sites, it doesn't seem necessary to keep the helper method for dispatching the 'scalechanging' event.
Note how nowhere in the code `canvasInRendering.get()` is ever called, and that this structure is really only used to store references to `<canvas>` DOM elements.
The reason for this being a `WeakMap` is probably because at the time we weren't using `core-js` polyfills yet, and since there already existed a manually implemented `WeakMap` polyfill it was probably simpler to use that.
Please note that, given the lack of a runnable example, I'm not totally sure if this first of all is enough to *completely* address the issue as filed and second of all if we actually want this new behaviour.
Currently `PDFFindController._calculateMatch` is (indirectly) dispatching an `updatetextlayermatches` event for every *single* page of the document. For short documents, such as the `tracemonkey` file, this probably doesn't matter too much, but for documents with a couple of thousand pages it seems unfortunate.
It shouldn't be necessary, in general, to dispatch `updatetextlayermatches` events here, since that's already being taken care of in `PDFFindController._updateMatch` which is always called when a match has been found.
However, when `highlightAll` is set we still need to ensure that pages which finished rendered *before* searching begun are updated correctly.
[Regression] Ensure that "Highlight All" is propagated to all pages for 'findagain' events where the findbar was previously closed (PR 10100 follow-up)
**STR:**
1. Open the default viewer, with the `tracemonkey` file.
2. Open the findbar, and search for "trace".
3. Enable the "Highlight All" option.
4. Close the findbar.
5. Re-open the findbar, and click on the "findNext" button.
6. Scroll down to the *second* page of the document.
**ER:**
Since "Highlight All" is active, all matches on the *second* page should be highlighted.
**AR:**
No matches are highlighted on the *second* page.
This is relevant for e.g. `PDFSinglePageViewer`, and `PDFViewer` with Presentation Mode active.
By moving this code to a helper method in `BaseViewer`, it's thus possible to reduce the amount of duplicate code that currently needed in `PDFViewer` and `PDFSinglePageViewer`.
For a short document, such as e.g. the `tracemonkey` file, this repeated normalization won't matter much, but for documents with a couple of thousand pages it seems completely unnecessary (and wasteful) to keep repeating the normalization whenever for every single page.
Currently the text-content is normalized every time that a new search operation is started, which seems completely useless considering that the "raw" text-content is never used for anything.
For a short document, such as e.g. the `tracemonkey` file, this repeated normalization won't matter much, but for documents with a couple of thousand pages it seems completely unnecessary (and wasteful) to keep repeating the normalization whenever e.g. a new search operation starts.
In the event that multiple instances of `PDFFindController` ever exists simultaneously, they will all be able to share just one `normalize` function in this way. Furthermore, the regular expression is now created lazily rather than at class construction time.
Despite all highlighted matches being removed in response to the 'findbarclose' event, there's a risk that a match could still be scrolled into view *after* the findbar has been closed[1].
Hence we need to ensure that long running searches, particularily those happening in large and/or slow loading documents[2], are ignored as well.
---
[1] The match is hidden, as expected, but the document could still scroll unexpectedly.
[2] Large documents loaded with `disableAutoFetch = true` and `disableStream = true` set are particularily susceptible to this issue.
Given that dispatching the 'updatetextlayermatches' event with `pageIndex = -1` set is now used to target the textLayers of *all* pages, there's no need to send individual events to every single page during `_nextMatch`. Since there can be an arbitrary number of pages in a document, this small/simple optimization seems too easy to ignore.
This patch does four things:
- Change the search related methods in `TextLayerBuilder` to be "private", since there're only called from within the class itself now.
- Use `const` for local variables not intended to change in the search related methods in `TextLayerBuilder`.
- Finally, removes most `this.findController` checks since they are redundant. Note how both `this._convertMatches` and `this._renderMatches` are *only* ever called, from `this._updateMatches`, when `this.findController` is actually defined. Hence there's really no need to repeat those checks all over the place, especially with all the relevant methods now being marked as "private".
- Always initialize the `this._pageMatchesLength` property with an empty array, to simplify the code in `TextLayerBuilder`.
This line in the annotation tests subtracts an array from a number. This is because `splice(-1, 1)` returns a one-element array, while `pop()` returns only the element itself.
`CMapCompressionType` makes a lot of sense to export, for anyone attempting to implement a custom `CMapReaderFactory`; fixes 10148.
`PermissionFlag` likewise needs to be exported, since otherwise the result of the `getPermissions` API method becomes difficult to interpret; follow-up to 10033.
*Please note:* I'm totally fine with this patch being rejected, and the issue closed as WONTFIX; however these changes should address the issue if that's desired.
From a conceptual point of view, reporting loading progress doesn't really make a lot of sense for PDF files opened by passing raw binary data directly to `getDocument` (since obviously *all* data was loaded).
This is compared to PDF files loaded via e.g. `XMLHttpRequest` or the Fetch API, where the entire PDF file isn't available from the start and knowing the loading progress makes total sense.
However I can certainly see why the current API could be considered inconsistent, which isn't great, since a registered `onProgress` callback will never be called for certain `getDocument` calls.
The simplest solution to this inconsistency thus seem to be to ensure that `onProgress` is always called when handling the `DataLoaded` message, since that will *always* be dispatched[1] from the worker-thread.
---
[1] Note that this isn't guaranteed to happen, since setting `disableAutoFetch = true` often prevents the *entire* file from ever loading. However, this isn't relevant for the issue at hand, and is a well-known consequence of using `disableAutoFetch = true`; note how the default viewer even has a specialized code-path for hiding the loadingBar.
*This should have been part of PR 10139.*
In the event that a user has attempted to manually load the worker file on the main-thread, but somehow failed to do that correctly, there's a possibility that `getMainThreadWorkerMessageHandler` could throw. Considering how/where that helper function is being called, an error could still prevent `PDFDocumentLoadingTask` from completing (regardless if it's being resolved/rejected).
With the way that the `getWorkerSrc()` helper function is implemented now, there's no longer a particularly strong reason for keeping the global `pdfjsFilePath` variable around.
With this patch the fallback `workerSrc` will thus, assuming is wasn't already set, be set to the "pdfjsFilePath" which simplifies the `getWorkerSrc()` function and reduces the amount of global state.
Finally, the global `workerSrc` variable was renamed to prevent shadowing.
This should, hopefully, cover all the possible ways[1] in which "fake workers" are loaded. Given the different code-paths, adding unit-tests might not be that simple.
Note that in order to make this work, the various `fakeWorkerFilesLoader` functions were converted to return `Promises`.
---
[1] Unfortunately there's lots of them, for various build targets and configurations.
Rather than having every invocation of `Toolbar._updateUIState` compute a valid `pageScaleValue`, it seems easier to simply ensure that it happens when the value is actually updated.
Looking at the history of this code, this parameter has never been used.
I'm guessing that most likely the code in `web/toolbar.js` began life as a copy of `web/secondary_toolbar.js`, which would probably explain why that parameter exists.
*This patch is based on something that I noticed while working on PR 10126.*
The recent re-factoring of `PDFFindController` brought many improvements, among those the fact that access to `BaseViewer` is no longer required. However, with these changes there's one thing which now strikes me as not particularly user-friendly[1]: The fact that in order for searching to actually work, `PDFFindController.setDocument` must be called *and* a 'pagesinit' event must be dispatched (from somewhere).
For all other viewer components, calling the `setDocument` method[2] is enough in order for the component to actually be usable.
The `PDFFindController` thus stands out quite a bit, and it also becomes difficult to work with in any sort of custom implementation. For example: Imagine someone trying to use `PDFFindController` separately from the viewer[3], which *should* now be relatively simple given the re-factoring, and thus having to (somehow) figure out that they'll also need to manually dispatch a 'pagesinit' event for searching to work.
Note that the above even affects the unit-tests, where an out-of-place 'pagesinit' event is being used.
To attempt to address these problems, I'm thus suggesting that *only* `setDocument` should be used to indicate that searching may start. For the default viewer and/or the viewer components, `BaseViewer.setDocument` will now call `PDFFindController.setDocument` when the document is ready, thus requiring no outside configuration anymore[4]. For custom implementation, and the unit-tests, it's now as simple as just calling `PDFFindController.setDocument` to allow searching to start.
---
[1] I should have caught this during review of PR 10099, but unfortunately it's sometimes not until you actually work with the code in question that things like these become clear.
[2] Assuming, obviously, that the viewer component in question actually implements such a method :-)
[3] There's even a very recent issue, filed by someone trying to do just that.
[4] Short of providing a `PDFFindController` instance when creating a `BaseViewer` instance, of course.
[Regression] Restore the ability to start searching before a document has loaded, and ignore searches for previously opened documents (PR 10099 follow-up)
Since searching itself is an asynchronous operation, removal of highlights needs to be asynchronous too since otherwise there's a risk that the events happen in the wrong order and find highlights thus remain visible.
Also, this patch will now ensure that only 'findbarclose' events for the *current* document is handled since other ones doesn't really matter. Note in particular that when no document is loaded text-layers are, obviously, not present and subsequently it's unnecessary to attempt to hide non-existent find highlights.
For many years it's been possible to enter a search term into the findbar(s) before the document has finised loading, such that searching starts immediately once it has loaded.
PR 10099 accidentally broke that, which I unfortunately missed during reviewing.
Since searching is asynchronous you cannot directly check in `executeCommand` if the document is loaded/current, but need to wait until searching is actually enabled first.
Furthermore this patch also ensures that the `_findTimeout` is always correctly cleared given that it adds further asynchronous behaviour to searching, since you obviously only want to deal with searches relevant to the current document.
This is similar to the format used by a number of other viewer components, and should simplify the `PDFSidebar` initialization slightly.
Furthermore, by using the `eventBus` it's no longer necessary for `PDFSidebar` to have a direct dependency on `PDFOutlineViewer`.
There's still room for improvement here, but this patch is at least a start (since it's not clear to me how best to handle the viewers).
This attempts to reduced the level of indirection, and the amount of code, when dispatching `fileattachmentannotation` events, by removing the `PDFLinkService.onFileAttachmentAnnotation` method and just accessing `PDFLinkService.eventBus` directly in the `FileAttachmentAnnotationElement` constructor.
Given that other properties, such as `externalLinkTarget`/`externalLinkRel`, are already being accessed directly this pattern seems fine here as well.
This commit shows that we can now unit test the find controller and
that executing regular queries works. Note that this is only a first
step and not a complete suite of unit tests for all possible options
of the find controller.
While writing this unit test, I found two smaller issues that I
addressed directly. The first one is that in the previous find
controller refactoring I forgot to rename some occurrences of a now
private member variable. Fortunately this did not cause any bugs since
we did have a public getter and the fetched value may be changed by
reference, but it's nevertheless good to fix. The second issue is that
some entries in the `test/unit/clitests.json` file were not correct,
resulting in these tests not being executed on e.g., Travis CI.
With `PDFFindController` instances no longer (directly) depending on
`BaseViewer` instances, we can pass a single `findController` when
initializing a viewer, similar to other components.
This removes the dependency on a `PDFViewer` instance from the find
controller, which makes it more similar to other components and makes it
easier to unit test with a mock link service.
Finally, we remove the search capabilities from the SVG example since it
doesn't work there because there is no separate text layer.
Now it follows the same pattern as e.g., the document properties
component, which allows us to have one instance of the find controller
and set a new document to search upon switching documents.
Moreover, this allows us to get rid of the dependency on `pdfViewer` in
order to fetch the text content for a page. This is working towards
getting rid of the `pdfViewer` dependency upon initializing the
component entirely in future commits.
Finally, we make the `reset` method private since it's not supposed to
be used from the outside anymore now that `setDocument` takes care of
this, similar to other components.
This way the resetting of `PDFLinkService`/`PDFDocumentProperties` instances, as is done in `PDFViewerApplication.close`, only requires passing in *one* `null` argument instead of two.
The `started` timestamp is completely usused, and the `end` timestamp is currently[1] being used essentially like a boolean value.
Hence this code can be simplified to use an actual boolean value instead, which avoids potentially hundreds (or even thousands) of unnecessary `Date.now()` calls.
---
[1] Looking briefly at the history of this code, I cannot tell if the timestamps themselves were ever used for anything (except for tracking "boolean" state).
The `Font.loading` property is only ever used *once* in the code, whereas `Font.missingFile` is more widely used. Furthermore the name `loading` feels, at least to me, slight less clear than `missingFile`. Finally, note that these two properties are the inverse of each other.
Ensure that all event properties are included, even if no (internal) listeners are registered, when re-dispatching events to the DOM (PR 10019 follow-up)
The built-in PDF Viewer (in Firefox) cannot use the browser findbar when PDF files are embedded in e.g. iframe/object tags, and the PDF.js findbar (i.e. `PDFFindBar`) will thus be used instead in those cases.
This is slightly problematic, since the `MOZCENTRAL` version of the viewer uses a special, slimmed down, version of the `l10n.js` file that doesn't (currently) support plural forms. To prevent the matchesCounter from breaking completely in this edge-case, temporarily hard-code the plural form to use the default `[other]` version of the locale strings.
This prevents the findbar from intermittently displaying `0 of {number} matches`, which *could* theoretically happen for large and/or slow loading documents.
The find controller should only coordinate finding a string in the
document and should not be responsible for presenting the matches to the
user. The text layer builder already contains the logic to render the
matches in the viewer, so it should also take care of scrolling the
selected match into view.
Currently there's only a single spot in the code-base where `JpegImage.getData` is called, however it nonetheless seem like a good idea to ensure during tests that the `isSourcePDF` parameter is correctly set. (Especially considering that the PDF use-cases will break without it.)
Additionally, in `JpegImage._getLinearizedBlockData`, the code can be made a tiny bit more efficient by checking the value of `isSourcePDF` *first* to avoid useless checks (for the default PDF use-cases).
The find controller already has quite a lot of state to maintain. We can
avoid keeping track of this member variable because when the find
controller is reset, so is the extract text promises array. Therefore,
we can just check if that array contains items or not to determine if
text extraction already started.
Moreover, there is no need to reset the `pageContents` array since the
`reset` method already takes care of that.
Implement unit tests for the `isSameOrigin` and `createValidAbsoluteUrl` utility functions and use the `const` keyword for constants in `src/shared/util.js`
This is an unfortunate oversight on my part, which I stumbled upon when (locally) testing the `mozilla-central` follow-up patch necessary to enable the matches counter in the built-in PDF viewer.
The property is intended to contain a reference to a DOM element, which not only is nowhere to be found *now* but appears to never have existed in the first place.
As outlined in https://bugzilla.mozilla.org/show_bug.cgi?id=1282759 the internal Firefox name for the feature is `entireWord`, hence that name is used here as well for consistency (with "Whole words" being limited to the UI).
Given existing limitations of the PDF.js search functionality, e.g. the existing problems of searching across "new lines", there's some edge-cases where "Whole words" searching will ignore (valid) results.
However, considering that this is a pre-existing issue related to the way that the find controller joins text-content together, that shouldn't have to block this new feature in my opionion.
*Please note:* In order to enable this feature in the `MOZCENTRAL` version, a small follow-up patch for [PdfjsChromeUtils.jsm](https://hg.mozilla.org/mozilla-central/file/tip/browser/extensions/pdfjs/content/PdfjsChromeUtils.jsm) will be required once this has landed in `mozilla-central`.
For the `PDFFindBar` implementation, similar to the native Firefox findbar, the matches count displayed is now limited to a (hopefully) reasonable value.
*Please note:* In order to enable this feature in the `MOZCENTRAL` version, a follow-up patch will be required once this has landed in `mozilla-central`.
There have been lots of problems with trying to map glyphs to their unicode
values. It's more reliable to just use the private use areas so the browser's
font renderer doesn't mess with the glyphs.
Using the private use area for all glyphs did highlight other issues that this
patch also had to fix:
* small private use area - Previously, only the BMP private use area was used
which can't map many glyphs. Now, the (much bigger) PUP 16 area can also be
used.
* glyph zero not shown - Browsers will not use the glyph from a font if it is
glyph id = 0. This issue was less prevalent when we mapped to unicode values
since the fallback font would be used. However, when using the private use
area, the glyph would not be drawn at all. This is illustrated in one of the
current test cases (issue #8234) where there's an "ä" glyph at position
zero. The PDF looked like it rendered correctly, but it was actually not
using the glyph from the font. To properly show the first glyph it is always
duplicated and appended to the glyphs and the maps are adjusted.
* supplementary characters - The private use area PUP 16 is 4 bytes, so
String.fromCodePoint must be used where we previously used
String.fromCharCode. This is actually an issue that should have been fixed
regardless of this patch.
* charset - Freetype fails to load fonts when the charset size doesn't match
number of glyphs in the font. We now write out a fake charset with the
correct length. This also brought up the issue that glyphs with seac/endchar
should only ever write a standard charset, but we now write a custom one.
To get around this the seac analysis is permanently enabled so those glyphs
are instead always drawn as two glyphs.
It's no longer necessary since https://bugzilla.mozilla.org/show_bug.cgi?id=1305963 is fixed quite some time ago.
While we're here, mark the `cachedGetSinglePixelWidth` member as being
private and use ES6 syntax in the `getSinglePixelWidth` method.
The purpose of this patch is to provide a better default behaviour when `JpegImage` is used to parse standalone JPEG images with CMYK colour spaces.
Since the issue that the patch concerns is somewhat of a special-case, the implementation utilizes the already existing decode support in an attempt to minimize the impact w.r.t. code size.
*Please note:* It's always possible for the user of `JpegImage` to control image inversion, and thus override the new behaviour, by simply passing a custom `decodeTransform` array upon initialization.
This patch is the first step to be able to eventually get rid of the `attachDOMEventsToEventBus` function, by allowing `EventBus` instances to simply re-dispatch most[1] events to the DOM.
Note that the re-dispatching is purposely implemented to occur *after* all registered `EventBus` listeners have been serviced, to prevent the ordering issues that necessitated the duplicated page/scale-change events.
The DOM events are currently necessary for the `mozilla-central` tests, see https://hg.mozilla.org/mozilla-central/file/tip/browser/extensions/pdfjs/test, and perhaps also for custom deployments of the PDF.js default viewer.
Once this have landed, and been successfully uplifted to `mozilla-central`, I intent to submit a patch to update the test-code to utilize the new preference. This will thus, eventually, make it possible to remove the `attachDOMEventsToEventBus` functionality.
*Please note:* I've successfully ran all `mozilla-central` tests locally, with these patches applied.
---
[1] The exception being events that originated on the `window` or `document`, since those are already globally available anyway.
The new events follow the same naming pattern as the 'pagesinit'/'pagesloaded' events dispatched on `BaseViewer` instances, and the intention is to allow the eventual removal of 'documentload'.
Ensure that the built `PdfJsDefaultPreferences.jsm` file won't be affected/touched during tree-wide ESLint rule changes in `mozilla-central` (PR 9571 follow-up)
Apparently there's some PDF generators, in this case the culprit is "Nooog Pdf Library / Nooog PStoPDF v1.5", that manage to mess up PDF creation enough that endstream[1] commands actually become truncated.
*Please note:* The solution implemented here isn't perfect, since it won't be able to cope with PDF files that contains a *mixture* of correct and truncated endstream commands.
However, considering that this particular mode of corruption *fortunately* doesn't seem very common[2], a slightly less complex solution ought to suffice for now.
Fixes 10004.
---
[1] Scanning through the PDF data to find endstream commands becomes necessary, in order to determine the stream length in cases where the `Length` entry of the (stream) dictionary is missing/incorrect.
[2] I cannot recall having seen any (previous) issues/bugs with "Missing endstream" errors.
Reduces the amount of boilerplate code when defining the the sub-classes.
Please note that a couple of the closures were kept, since it's not (yet) possible to include helper functions inside of `class`es.
This property is not only completely unused now, it never actually appears to have been used. Even though the memory savings, from not initializing these extra typed arrays, won't be significant in the grand scheme of things it still seems completely unnecessary to keep allocating this data.
As far as I can tell, the main reason for the existence of `defaultColor` seem to be for documentation purposes. Hence the code is changed into comments instead, to keep the information around (but without the unnecessary allocations).
For proof-of-concept, this patch converts a couple of `Promise` returning methods to use `async` instead.
Please note that the `generic` build, based on this patch, has been successfully testing in IE11 (i.e. the viewer loads and nothing is obviously broken).
Being able to use modern JavaScript features like `async`/`await` is a huge plus, but there's one (obvious) side-effect: The size of the built files will increase slightly (unless `SKIP_BABEL == true`). That's unavoidable, but seems like a small price to pay in the grand scheme of things.
Finally, note that the `chromium` build target was changed to no longer skip Babel translation, since the Chrome extension still supports version `49` of the browser (where native `async` support isn't available).
Not only is this method completely unused *now*, looking through the history of the code it never appears to have been used for anything either.
Years ago `mainXRefEntriesOffset` was included when creating `XRef` instances, however it wasn't actually used for anything (the parameter was never checked, nor assigned to a property on `XRef`).
If this method ever becomes useful (again) it's easy enough to restore it thanks to version control, but including dead code in the builds just seems wasteful.
Please note that while this *improves* issue 9984 slightly (and likely others too), it's not a complete solution.
The remaining issues are related to the, more general, problems with the existing heuristics related to attempting to combine separate text items.
One of the `QueueOptimizer` cases wasn't updated to use `Uint8ClampedArray`s, which leads to inconsistent image data on the API side (but no actual rendering bugs, as far as I can tell).
To prevent future errors, a non-production/test-only `assert` was added to ensure that the relevant image data only uses `Uint8ClampedArray`s.
This commit is the first step towards implementing parsing for the
appearance streams of annotations.
Co-authored-by: Jonas Jenwald <jonas.jenwald@gmail.com>
Co-authored-by: Tim van der Meij <timvandermeij@gmail.com>
The font tests use Jasmine too, so while they are technically unit
tests, it's a bit confusing to see `Started unit tests` when the font
tests are run on the bots.
The current font type/subtype detection code is quite inconsistent/unwieldy. In some cases it will simply assume that the font dictionary is correct, in others it will somewhat "arbitrarily" check the actual font file (more of these cases have been added over the years to fix specific bugs).
As is evident from e.g. issue 9949, the font type/subtype detection code is continuing to cause issues. In an attempt to get rid of these hacks once and for all, this patch instead re-factors the type/subtype detection to *always* parse the font file.
Please note that, as far as I can tell, we still appear to need to rely on the composite font detection based on the font dictionary. However, even if the composite/non-composite detection would get it wrong, that shouldn't really matter too much given that there's basically only two different code-paths (for "TrueType-like" vs "Type1-like" fonts).
The font in the PDF is marked as a CIDFontType0, but the font file is
actually a true type font. To fully address this issue we should really
peek into the font file and try to determine what it is. However, this
is the first case of this issue, so I think this solution is acceptable for
now.
Fixes a stupid oversight on my part, since /Filter may (obviously) contain an Array, which resulted in unnecessary console warning spam in perfectly valid PDF files.
Note that it still makes sense to check that /Filter is actually a Name, before attempting to access its `name` property, but the warning should definitely be removed.
This should really have been included in PR 9868, since it will help ensure that the `URL` constructor is correctly imported/exported by `src/shared/util.js`.
According to the PDF specification, see https://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#page=45
> When using the JPXDecode filter with image XObjects, the following changes to and constraints on some entries in the image dictionary shall apply (see 8.9.5, "Image Dictionaries" for details on these entries):
>
> - Width and Height shall match the corresponding width and height values in the JPEG2000 data.
>
> - . . .
Hence it seems reasonable to use the Width/Height of the image data *itself*, rather than the image dictionary when there's a mismatch. Given that JPEG 2000 images are already being parsed, in order to obtain basic parameters, the actual Width/Height is readily available in the `PDFImage` constructor.
Given that the code is currently assuming that the /Filter entry is a `Name`, it cannot hurt to actually ensure that's the case.
Also fixes an error message, for JPEG 2000 images with unsupported ColorSpaces, since `this.numComps` hasn't been initialized when it's accessed during the `throw new Error()` invocation.
[api-minor] Add an `IsLinearized` property to the `PDFDocument.documentInfo` getter, to allow accessing the linearization status through the API (via `PDFDocumentProxy.getMetadata`)
Since PDF.js already supports range requests and streaming, not to mention chunked rendering, attempting to use the `Linearization` dictionary in `PDFDocument.getPage` probably isn't going to improve performance in any noticeable way.
Nonetheless, when `Linearization` data is available, it will allow looking up the first Page *directly* without having to descend into the `Pages` tree to find the correct object.
With the `builtInCMapCache` being a simple Object, it unfortunately means that the `Catalog.cleanup` method isn't resetting it as intended.
By just replacing the `builtInCMapCache` with an empty Object, existing references to it will not actually be updated. The result is that e.g. `Page` instances still keeps references to, what should have been removed, CMap data.
To fix these problems, the `builtInCMapCache` is converted into a `Map` instead (since it can be easily reset).
When updating Preferences using the `set` method, the input is carefully validated. However, no validation is (currently) done when a `BasePreferences` instance is created, which probably isn't that great. Hence this patch that simply ignores, to not unnecessarily break loading of the viewer itself, any invalid Preferences.
Given that the various Preferences are currently, and have been for quite some time, only used when initializing `PDFViewerApplication` re-loading them when a new PDF file is opened in the viewer is essentially a no-op.
Furthermore, with the only usage of `BasePreferences.reload` now gone, the value of that method seems questionable at best. In the event that the functionality is actually needed again, similar to the `ViewHistory`, it'd probably make more sense to simply replace `PDFViewerApplication.preferences` with a new `BasePreferences` instance instead (using e.g. `DefaultExternalServices.createPreferences`).
The `git` protocol is unencrypted which means other parties could potentially eavesdrop your traffic. `https` or `ssh` is often encouraged because of this. (For example, the Ruby package manager `bundler` prints a warning when `git` sources are being used.)
There was a (somewhat) recent question on IRC about accessing the linearization status of a PDF document, and this patch contains a simple way to expose that through already existing API methods.
Please note that during setup/parsing in `PDFDocument` the linearization data is already being fetched and parsed, provided of course that it exists. Hence this patch will *not* cause any additional data to be loaded.
With this file now being a proper (ES6) module, it's no longer (technically) necessary for this structure to be lazily initialized. Considering its size, and simplicity, I therefore cannot see the harm in letting `DocumentInfoValidators` just be simple Object instead.
While I'm not aware of any bugs caused by the current code, it cannot hurt to add an `isDict` check in `PDFDocument.documentInfo` (since the current code assumes that `infoDict` being defined implies it also being a Dictionary).
Finally, the patch also converts a couple of `var` to `let`/`const`.
Note first of all that `PDFDocument` will be initialized with either a `Stream` or a `ChunkedStream`, and that both of these have `length` getters. Secondly, the `PDFDocument` constructor will assert that the `stream` has a non-zero (and positive) length. Hence there's no point in checking `stream.length` in the `linearization` getter.
For most other `DecodeStream` based streams, we'll attempt to estimate the minimum `buffer` length based on the raw stream data. The purpose of this is to avoid having to unnecessarily re-size the `buffer`, thus reducing the number of *intermediate* allocations necessary when decoding the stream data.
However, currently no such optimization is attempted for `StreamsSequenceStream`, and given that they can often be quite large that seems unfortunate. To improve this, at least somewhat, this patch utilizes the raw sizes of the `StreamsSequenceStream` sub-streams to estimate the minimum required `buffer` length.
Most likely this patch won't have a huge effect on memory consumption, however for pathological cases it should help reduce peak memory usage slightly.
One example is the PDF file in issue 2813, where currently the `StreamsSequenceStream` instances would grow their `buffer`s as `2 MiB -> 4 MiB -> 8 MiB -> 16 MiB -> 32 MiB`. With this patch, the same stream `buffers`s grow as `8 MiB -> 16 MiB -> 32 MiB`, thus avoiding a total of `12 MiB` of *intermediate* allocations (since there's two `StreamsSequenceStream` used, for rendering/text-extraction).
Add a `BasePreferences.getAll` method and use it to fetch all Preferences at once in `PDFViewerApplication._readPreferences`; ensure that the `externalLinkTarget` option is correctly set when the viewer is embedded
Given that *all* Preferences are already fetched in `PDFViewerApplication._readPreferences`, the amount of boilerplate/duplication can be considerably reduced with the addition of a `BasePreferences.getAll` method.
The only reason that this check ever existed in the first place, is that originally there was a global `PDFJS.openExternalLinkInNewWindow` option which was then subsumed by the (more generic) `PDFJS.externalLinkTarget` option. (The `externalLinkTarget` has since been moved into a `PDFLinkService` option, as part of PDF.js version `2.0`.)
Hence, during the period where both `PDFJS.openExternalLinkInNewWindow` and `PDFJS.externalLinkTarget` existed side-by-side, there was a need to allow the former one to override the latter one (for backward compatibility purposes). However, that's no longer the case, and this extra `externalLinkTarget` check can now be removed.
*This was a stupid error on my part; sorry about breaking this!*
With the current code, the value of the `externalLinkTarget` option is now (potentially) updated *after* the viewer components have been initialized. For the "viewer in iframe/object tag" case, the result is that the value of the `externalLinkTarget` option isn't adjusted as intended any more.
Without providing useful (custom) error messages for the `no-restricted-globals` rule, see https://eslint.org/docs/rules/no-restricted-globals, it's quite likely that the rule will be incorrectly disabled rather than the required globals being imported as intended.
To reduced duplication of the `no-restricted-globals` rule in multiple `.eslintrc` files, it's instead moved to the top-level `.eslintrc` file and disabled as needed on a folder/file basis outside of `/src` and `/web`.
With the new XML parser, see PR 9573, the referenced PDF file now causes `getMetadata` to fail when incomplete XML tags are encountered. This provides a simple, and hopefully generally useful, work-around that may also help prevent future bugs.
(Without being able to reproduce nor even understand the other (non XML) errors mentioned in issue 8884, I'd say that this patch is enough to close that one as fixed.)
*Another small piece of clean-up of code I've previously written; follow-up to PR 8775.*
Importing `createPromiseCapability`, and then using it in just *one* spot, seems unnecessary since the `waitOnEventOrTimeout` function may just as well return a regular `Promise` directly.
Compared to all the other (static) methods in `Util`, the `toRoman` one looks slightly out of place. Even more so considering that `Util` is being exposed through `pdfjsLib`, where access to a Roman numerals conversion method doesn't make much sense.
*I was feeling bored; so this is a very quick, and somewhat naive, attempt at fixing the bug.*
The breaking error, i.e. `Error during font loading: invalid array length`, was thrown when attempting to re-size the `stack` to a *negative* length when parsing the CALL functions.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1473809.
PR 6141 changed `CompiledFont.compileGlyph` to, in the general case, return an Array. However, that PR apparenly forgot to update the no-glyph, empty-glyph, and endchar-glyph code-path and a String was still being (incorrectly) returned.
Given the way that `FontFaceObject.getPathGenerator` (on the API side) is implemented, this shouldn't have caused any bugs despite the Worker possible returning unexpected data.
It gives warnings when running `gulp locale`, in a previous patch it was
explicitly asked to remove it and there is quite a bit of discussion
about the validity of the locale at
https://bugzilla.mozilla.org/show_bug.cgi?id=418485. In short, let's
remove it to prevent these issues.
Given that the non-default Spread modes (currently) doesn't affect the page layout when horizontal scrolling is enabled, having the Spread buttons appear active when clicking them appears to do *nothing* is probably confusing rather than helpful to users.
If the current viewer is a `PDFSinglePageViewer` instance the Scroll/Spread modes are no-ops, hence displaying buttons that do *nothing* when clicked will probably do very little besides confuse users.
The names 'resetscrollmode'/'resetspreadmode' were probably *not* great choices, given that the only thing being reset are toolbar buttons and not the actual Scroll/Spread modes. Furthermore, there's really no need for two separate events here.
The patch also adds a comment that ought to have been included in PR 9040, to prevent future refactoring/removing of what may appear to be an unnecessary `Promise.resolve` call.
This moves/exposes the `URL` polyfill similarily to the existing `ReadableStream` polyfill, rather than exposing it globally, to avoid interfering with any "outside" code.
Both the `URL` and `ReadableStream` polyfills are now exposed on the `pdfjsLib` object, such that they are accessible to the viewer components.
Furthermore, the `no-restricted-globals` ESLint rule is also enabled to prevent accidental usage of the native `URL`/`ReadableStream` implementations directly in the `src/` and `web/` folders; see also https://eslint.org/docs/rules/no-restricted-globals
Addresses the remaining TODO in https://github.com/mozilla/pdf.js/projects/6
Rather than having to manually call a method on `PDFFindController` instances from `BaseViewer.setDocument`, thus essentially having to resolve the private `_firstPagePromise` from the "outside", this can be done easily with the 'pagesinit' event dispatched on the `eventBus` instead.
Please note this particular `PDFFindController` code pre-dates the `eventBus` by almost three years, which should explain why the code looks the way it does.
Since the Scroll/Spread modes are now document specific, as all other properties such as page/scale/rotation, ensure that the toolbar is always correctly reset.
Since other viewer state, such as the current page/scale/rotation[1], are not available as `BaseViewer` constructor options, this makes the Scroll/Spread modes stand out quite a bit. Hence it probably makes sense to remove/deprecate this, to avoid inconsistent and possibly confusing state in this code.
---
[1] These properties are *purposely* not available in the constructor, since attempting to set them before a document is loaded has number of issues; please refer to https://github.com/mozilla/pdf.js/pull/8539#issuecomment-309706629 for additional details.
Since all the other viewer methods use the getter/setter pattern, e.g. for setting page/scale/rotation, the way that the Scroll/Spread modes are set thus stands out. For consistency, this really ought to use the same pattern as the rest of the `BaseViewer`. (To avoid breaking third-party implementations, the old methods are kept around as aliases.)
Note how the other "...OnLoad" preferences will allow a *non-default* value to always override a history entry. To improve overall consistency for the viewer options, and to reduce possibly confusing behaviour, this patch changes the `scrollModeOnLoad`/`spreadModeOnLoad` preferences to behave as all the other ones.
For some reason, these weren't added to `AppOptions` despite actually being set and read from `web/app.js`. Not adding them creates inconsistencies, since all other options *are* present in `web/app_options.js`.
This property isn't accessed anywhere in the `web/app.js` file, and is also not being reset in `PDFViewerApplication.close`. Hence it seems that it can simply be removed, especially since the fingerprint is already synchronously available through `PDFViewerApplication.pdfDocument.fingerprint` (provided that a document is loaded).
Currently if `RenderTask.cancel` is called *immediately* after rendering was started, then by the time that `InternalRenderTask.initializeGraphics` is called rendering will already have been cancelled.
However, we're still inserting the canvas into the `canvasInRendering` map, thus breaking any future attempts at re-rendering using the same canvas. Considering that `InternalRenderTask.cancel` always removes the canvas from the map, I cannot imagine that we'd ever want to re-add it *after* rendering was cancelled (it was likely just a simple oversight in PR 8519).
Fixes 9456.
With the current code, the location in the viewer could change *well* after the user has started to interact with the viewer (for very large and/or slow loading documents). Attempt to reduce the likelyhood of that happening, by adding an upper bound to the time spent waiting before attempting to re-apply the initial position.
Rather than using a "special" property to check if a `ViewHistory` database entry existed on load, it seems that you could just as well check for the existence of one of the actually needed properties instead (here 'page' is used).
This way we can avoid storing what, more of less, amounts to useless state, which will help reduce the size of the `ViewHistory` database. Given that we don't directly, nor need to, validate the `ViewHistory` data but rely on other parts of the code-base to do so, the existance of an 'exists' property doesn't in and of itself really add much utility.
Finally, to simplify the implementation the 'exists' property won't be actively removed from the `ViewHistory` data. Instead we'll simply stop adding 'exists' when writing `ViewHistory` data, and rely on the existing pruning of old entries to eventually remove any remnants of 'exists' from storage.
Note how, in the current code, only *one* old history entry would ever be removed. That would mean that if e.g. the `cacheSize` is reduced, it would potentially require loading of multiple files before the database would be correctly pruned.
Furthermore, in the case where the database was empty on load there's no need to attempt to shrink it, since trying to reduce the size of an *empty* array won't do much :-)
This wasn't included in PR 9245, since all the API options were still global at that time.
Writing the unit-tests also uncovered an issue with `getOperatorList` not starting the "Page Request" timer.
Incomplete path operators, in particular, can result in fairly chaotic rendering artifacts, as can be observed on page four of the referenced PDF file.
The initial (naive) solution that was attempted, was to simply throw a `FormatError` as soon as any invalid (i.e. too short) operator was found and rely on the existing `ignoreErrors` code-paths. However, doing so would have caused regressions in some files; see the existing `issue2391-1` test-case, which was promoted to an `eq` test to help prevent future bugs.
Hence this patch, which adds special handling for invalid path operators since those may cause quite bad rendering artifacts.
You could, in all fairness, argue that the patch is a handwavy solution and I wouldn't object. However, given that this only concerns *corrupt* PDF files, the way that PDF viewers (PDF.js included) try to gracefully deal with those could probably be described as a best-effort solution anyway.
This patch also adjusts the existing `warn`/`info` messages to print the command name according to the PDF specification, rather than an internal PDF.js enumeration value. The former should be much more useful for debugging purposes.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1443140.
Since the current page will be explicitly scrolled into view *directly* afterwards anyway (compare with e.g. the `pagesRotation` code), trying to maintain the current position when re-applying the zoom level during changing of Scroll modes is redundant.
Note how in `BaseViewer.forceRendering` the Scroll mode is used to determine how pre-rendering will work. Currently this is broken in Presentation Mode, if horizontal scrolling was enabled prior to entering fullscreen.
Furthermore, there's a few additional cases where the `this.scrollMode === ScrollMode.HORIZONTAL` check is pointless either in Presentation Mode or when a `PDFSinglePageViewer` instance is used.
Since the Scroll/Spread modes doesn't make (any) sense in `PDFSinglePageViewer` instances, the general structure of these methods can be improved to reflect that.
Given that this method is a no-op in `PDFSinglePageViewer`, similar to `_regroupSpreads`, let's improve the general code structure by simply moving the method.
There's no good reason to iterate through an arbitrary number of DOM elements this way, since a document could contain thousands of pages, when everything can be easily removed at once; compare with e.g. `BaseViewer._resetView` and `PDFThumbnailViewer._resetView`.
Furthermore given that it's a `PDFViewer` instance, the `this.viewer` property can be accessed directly. Besides, `_setDocumentViewerElement` only exists as a helper method for `setDocument` in the base class and none of this code applies for `PDFSinglePageViewer` instances either.
The Secondary Toolbar buttons for, not to mention the actual toggling of, Scroll/Spread modes are currently completely broken in older browsers (such as IE11).
As a follow-up, it'd probably be a good idea to try and find a *feature complete* `classList` polyfill that could be used instead, but this patch at least addresses the immediate regression.
Please refer to the compatibility information in https://developer.mozilla.org/en-US/docs/Web/API/Element/classList#Browser_compatibility
This patch avoids choosing a (possible) 'trailer' dictionary that `XRef.parse` and/or the `Catalog` constructor/methods will reject anyway.
Since `XRef.indexObjects` is already parsing the entire PDF file, the extra dictionary look-ups added here shouldn't matter much. Besides, this is a fallback code-path that only applies to corrupt PDF files anyway.
Note that the `Catalog` constructor, and some of its methods, are already enforcing that the 'Root' dictionary is valid/well-formed. However, by doing additional validation already in `XRef.parse` there's a slightly larger chance that corrupt PDF files could be successfully parsed/rendered.
With the current code line-breaks are accepted not just after an operator, but after a decimal point as well. When looking at this again, the latter case seems prone to cause false positives and might also interfere with subsequent patches.
Hence this is code is adjusted to actually do what the original commit message says, and nothing more.
The `onOpenWithURL` method, in `PDFViewerApplication.initPassiveLoading`, accepts a `originalURL` parameter which is then passed on to `PDFViewerApplication.open` as is. However, the latter method expects the name of the parameter to be `originalUrl` (note the casing), meaning that `getDocument` will fail in this case.
For consistency, and to avoid confusion, the renaming is done in `web/chromecom.js` as well.
Please note that the standalone `pdf.image_decoders.js` file will be including the complete `src/shared/util.js` file, despite only using parts of it.[1] This was done *purposely*, to not negatively impact the readability/maintainability of the core PDF.js code.
Furthermore, to ensure that the compatibility is the same in the regular PDF.js library *and* in the the standalone image decoders, `src/shared/compatibility.js` was included as well.
To (hopefully) prevent future complaints about the size of the built `pdf.image_decoders.js` file, a few existing async-related polyfills are being skipped (since all of the image decoders are completely synchronous).
Obviously this required adding a couple of pre-processor statements, but given that these are all limited to "compatibility" code, I think this might be OK!?
---
[1] However, please note that previous commits moved `PageViewport` and `MessageHandler` out of `src/shared/util.js` which reduced its size.
The purpose of this patch is to hopefully provide *slightly* better user ergonomics, if/when the PDF.js image decoders are used standalone.
This implementation is (basically) reverting the changes in PR 9386, in conjunction with code from the `parse` method found at https://github.com/notmasteryet/jpgjs/blob/master/src/pdfjs.js
*The danger of fixing one bug is that it can, sometimes too easily, cause another one in the process; sorry for not catching this when testing PR 9816 locally.*
When the *entire* viewer becomes narrow enough, as controlled by the `@media all and (max-width: 840px)` media query, the sidebar should (semi-transparently) overlay the `viewerContainer` instead of moving it horizontally. Unfortunately the changes made in PR 9816 caused the relevant CSS rules to be skipped, because of how the inheritance model works in CSS.
I'm well aware that `!important` is usually advised against, since it "breaks" the CSS inheritance model. However in this case it seemed reasonable to use it, to not only fix the bug at hand but to also prevent similar bugs from occurring in the future.
This is a regression from PR 8993; it causes the pages to be offset horizontally in Presentation Mode, if and only if the sidebar is currently open when the user triggers Presentation Mode.
Please note that while this doesn't seem to affect Firefox, both Chrome and IE are however affected.
Interestingly enough, despite the Chrome extension being affected as well, I cannot find any issue filed about this. (Either Presentation Mode isn't used much at all, or users don't open the sidebar before entering Presentation Mode.)
It appears that Microsoft silently fixed the problem that required disabling of fullscreen mode, in e.g. `iframe`s, in IE 11; please see issue 4711 and PR 5525 for historical context.
Unfortunately my Google-fu isn't strong enough to find any *official* information regarding the fixing of the browser bug in IE. However testing of the default viewer in IE 11, with this patch applied, it now appears that Presentation Mode is working correctly even in an `iframe` in IE 11.
Further anecdotal evidence that the bug is in fact fixed, is for example that jQuery previously contained a work-around for the IE bug. However, that's removed over two years ago now; see ff1a0822f7 and the issues referenced there.
Given that the default viewer isn't intended to be used as-is anyway (in custom deployments), it didn't seem necessary to keep the `disableFullscreen` option around since it was *only* ever added for compatibility purposes.
Fixes 9585.
Obviously it's still not possible to render non-embedded fonts as paths, but in this way the rest of the page will at least be allowed to continue rendering.
*Please note:* Including the 14 standard fonts in PDF.js probably wouldn't be *that* difficult to implement. (I'm not a lawyer, but the fonts from PDFium could probably be used given their BSD license.)
However, the main blocker ought to be the total size of the necessary font data, since I cannot imagine people being OK with shipping ~5 MB of (additional) font data with Firefox. (Based on the reactions when the CMap files were added, and those are only ~1 MB in size.)
Since `ColorSpace` now depends on the native clamping of `Uint8ClampedArray`, this patch adds non-production/test-only `assert`s to enforce that the expected TypedArray is used for the output.
These `assert`s are purposely *not* included in PRODUCTION builds since that would break rendering completely, as opposed to "only" displaying some weird colours, when a `Uint8Array` was used. Furthermore, these are mostly added to help catch explicit developer errors when working with the `ColorSpace` and `PDFImage` code.
Since the tests (currently) run with the `pdf.worker.js` file built, i.e. with `PRODUCTION = true` set, there's no simple way to add e.g. `assert` calls for both non-production *and* test-only builds without also affecting PRODUCTION builds.
The built-in image decoders are already using `Uint8ClampedArray` when returning data, and this patch simply extends that to the rest of the image/colorspace code.
As far as I can tell, the only reason for using manual clamping/rounding in the first place was because TypedArrays used to be polyfilled (using regular arrays). And trying to polyfill the native clamping/rounding would probably have been had too much overhead, but given that TypedArray support is required in PDF.js version `2.0` that's no longer a concern.
*Please note:* Because of different rounding behaviour, basically `Math.round` in `Uint8ClampedArray` respectively `Math.floor` in the old code, there will be very slight movement in quite a few existing test-cases. However, the changes should be imperceivable to the naked eye, given that the absolute difference is *at most* `1` for each RGB component when comparing `master` and this patch (see also the updated expectation values in the unit-tests).
The built-in image decoders are already returning data as `Uint8ClampedArray`, and subsequently the JPEG/JBIG2/JPX streams are as well. However, for general streams we obviously don't want to force the use of `Uint8ClampedArray` unless an "Image" is actually being decoded.
Hence this patch, which adds a parameter that allows the caller of the `getBytes`/`peekBytes` methods to force a `Uint8ClampedArray` (rather than a `Uint8Array`) to be returned.
Not only is the `Util.loadScript` helper function unused on the Worker side, even trying to use it there would throw an Error (since `document` isn't defined/available in Workers).
Hence this helper function is moved, and its code modernized slightly by having it return a Promise rather than needing a callback function.
Finally, to reduced code duplication, the "new" loadScript function is exported and used in the viewer.
The various classes have `this._errored` and `this._reason` properties, where the first one is a boolean indicating if an error was encountered and the second one contains the actual `Error` (or `null` initially).
In practice this means that errors are basically tracked *twice*, rather than just once. This kind of double-bookkeeping is generally a bad idea, since it's quite easy for the properties to (accidentally) get into an inconsistent state whenever the relevant code is modified.
Rather than using a separate boolean, we can just as well check the "error" property directly (since `null` is falsy).
---
Somewhat unrelated to this patch, but `src/display/node_stream.js` is currently *not* handling errors in a consistent or even correct way; compared with `src/display/network.js` and `src/display/fetch_stream.js`.
Obviously using the `createResponseStatusError` utility function, from `src/display/network_utils.js`, might not make much sense in a Node.js environment. However at the *very* least it seems that `MissingPDFException`, or `UnknownErrorException` when one cannot tell that the PDF file is "missing", should be manually thrown.
As is, the API (i.e. `getDocument`) is not returning the *expected* errors when loading fails in Node.js environments (as evident from the `pending` API unit-test).
Since the old comment mentions a now unsupported browser, let's update it such that someone won't accidentally conclude that the code in question can be removed.
This special handling was added in PR 8567, but was made redundant in PR 8721 which stopped sending everything but the kitchen sink to the Worker side.
Since `PDFPageProxy` already provide getters for all the data returned by `GetPage` (in the Worker), there isn't any compelling reason for accessing the `pageInfo` directly on `PDFPageProxy`.
The patch also changes the `GetPage` handler, in `src/core/worker.js`, to use modern JavaScript features.
Since `PDFDocumentProxy` already provide getters for all the data returned by `GetDoc` (in the Worker), there isn't any compelling reason for accessing the `pdfInfo` directly on `PDFDocumentProxy`.
After PR 8617 the `PDFManagerReady` message handler function, in `src/display/api.js`, is now a no-op. Hence it seems completely unnecessary to keep sending this message from `src/core/worker.js`.
With native typed array support now being mandatory in PDF.js, since version 2.0, this probably isn't a huge problem even though the current code seems wrong (it was changed in PR 6571).
Note how in the `!(data instanceof Uint8Array)` case we're currently attempting to send `handler.send('test', 'main', false);` to the main-thread, which doesn't really make any sense since the signature of the method reads `send(actionName, data, transfers) {`.
Hence the data that's *actually* being sent here is `'main'`, with `false` as the transferList, which just seems weird. On the main-thread, this means that we're in this case checking `data && data.supportTypedArray`, where `data` contains the string `'main'` rather than being falsy. Since a string doesn't have a `supportTypedArray` property, that check still fails as expected but it doesn't seem great nonetheless.
Since all the built-in PDF.js image decoders now return their data as `Uint8ClampedArray`, for consistency `JpegDecode` on the main-thread should be doing the same thing; follow-up to PR 8778.
The signature of the `PDFWorker.fromPort` method, in addition to the `PDFWorker` constructor, was changed in PR 9480.
Hence it's probably a good idea to add a bit more validation to `PDFWorker.fromPort`, to ensure that it won't fail silently for an API consumer that updates to version 2.0 of the PDF.js library.
With version 2.0, native support for typed arrays is now a requirement for using the PDF.js library; see PR 9094 where the old polyfills were removed.
Hence the `isTypedArraysPresent` check, when setting up fake workers, no longer serves any purpose here and can thus be removed.
There's no good reason, as far as I can tell, to duplicate the functionality of the `LoopbackPort` in the unit-tests. The only difference between the implementations is that `LoopbackPort` mimics the (native) structured cloning, however that shouldn't matter here since the tests are only sending "simple" data (strings respectively arrays with numbers).
Furthermore the patch also changes `LoopbackPort` to default to using "structured cloning" and deferred invocation of the listeners, since native typed array support is now a requirement for using the PDF.js library.
The `MessageHandler` itself, and its assorted helper functions, are currently the single largest[1] piece of code in the `src/shared/util.js` file. By moving this code into its own file, `src/shared/util.js` thus becomes smaller and more manageable.
The `fontScale` property was added in PR 1531, see commit b312719d7e in particular, apparently for the sole purpose of supporting the "acroforms" example.
However, the `fontScale` property was never used anywhere else in the code-base, and after the modernization of the "acroforms" example in PR 8030 it's been completely unused.
Finally, note that there's also a (more suitably named) `scale` property on `PageViewport` instances, which contains the exact same information as the property being removed here.
To avoid having to manually update the "shared files" list in `gulpfile.js`, whenever any changes are made to the file structure in the `src/shared/` folder, let's simply list the files we do *not* want instead.
Given that the excluded files have been consistent ever since the `lib` build target was added, this patch should help avoid unnecessary churn in `gulpfile.js` in the future.
I made some mistakes when trying to make the content_disposition.js
compatible with non-modern browsers (IE/Edge).
Notably, text decoding was usually skipped because of the inverted
logical check at the top of `textdecode`.
I verified that this new version works as expected, as follows:
1. Visit 55c71eb44e/test/
and get test-content-disposition.js
also get test-content-disposition.node.js if using Node.js,
or get test-content-disposition.html if you use a browser.
2. Modify `test-content-disposition.node.js` (or the HTML file) and
change `../extension/content-disposition.js` to `PDFJS-content_disposition.js`
3. Copy the `getFilenameFromContentDispositionHeader` function from
`content_disposition.js` (i.e. the file without the trailing exports)
and save it as `PDFJS-content_disposition.js`.
4. Run the tests (`node test-content-disposition.node.js` or by opening
`test-content-disposition.html` in a browser).
5. Confirm that there are no failures: "Finished all tests (0 failures)"
The code has a best-efforts fallback for Microsoft Edge, which lacks the
TextDecoder API. The fallback only supports the common UTF-8 encoding.
To simulate this in a test, modify `PDFJS-content_disposition.js` and
deliberately throw an error before `new TextDecoder`. There will be two
failures because we don't want to include too much code to support text
decoding for non-UTF-8 encodings in Edge
```
test-content-disposition.js:265 Assertion failed: Input: attachment; filename*=ISO-8859-1''%c3%a4
Expected: "ä"
Actual : "ä"
test-content-disposition.js:268 Assertion failed: Input: attachment; filename*=ISO-8859-1''%e2%82%ac
Expected: "€"
Actual : "€"
```
Compared to running the unit-tests in "regular" browsers, where any console output won't get mixed up with test output, in Node.js/Travis the test output looks quite noisy.
By ignoring `info`/`warn` calls, when running unit-tests in Node.js/Travis, the test output is a lot smaller not to mention that any *actual* failures are more easily spotted.
Please note that while the current code works, both in the viewer and the unit-tests, it can leave the `WorkerTransport._passwordCapability` Promise in a pending state.
In the `PasswordRequest` handler, in src/display/api.js, we're returning the Promise from a `capability` object (rather than just a "plain" Promise). While an error thrown anywhere within this handler was fortunately enough to propagate it to the Worker side, it won't cause the Promise (in `WorkerTransport._passwordCapability`) to actually be rejected.
Finally note that while we're now catching errors in the `PasswordRequest` handler, those errors are still propagated to the Worker side via the (now) rejected Promise and the existing `return this._passwordCapability.promise;` line.
This prevents warnings about uncaught Promises, with messages such as "Error: Worker was destroyed during onPassword callback", when running the unit-tests both in browsers *and* in Node.js/Travis.
This should provide a better out-of-the-box experience when using PDF.js in a Node.js environment, since it's missing native support for both `@font-face` and `Image`.
Please note that this change *only* affects the default values, hence it's still possible for an API consumer to override those values when calling `getDocument`.
Also, prevents "ReferenceError: document is not defined" errors, when running the unit-tests in Node.js/Travis.
This avoids the initialization of, potentially thousands of, unnecessary `Stream` objects, by getting the required number of bytes directly instead.
Given the special behaviour, when `length === 0`, of the `getBytes`/`skip` methods, it's also necessary to handle that particular case to prevent errors when encountering empty CharStrings.
This commit adds `scrollModeOnLoad` and `spreadModeOnLoad` preferences
that control the default viewer state when opening a new document for
the first time.
This commit also contains a minor refactoring of some of the option UI
rendering code in extensions/chromium/options/options.js, as I couldn't
bear creating two more functions nearly identical to the four that
already existed.
This builds on the scrolling mode work to add three buttons for joining
page spreads together: one for the default view, with no page spreads,
and two for spreads starting on odd-numbered or even-numbered pages.
Prior to this commit, if the vertical scroll bar is absent and the horizontal
scroll bar is present, a link to a particular point on the page which should
induce a horizontal scroll did not do so, because the absence of a vertical
scroll bar meant that the viewer was not recognized as the nearest scrolling
ancestor. This commit adds a check for horizontal scroll bars when searching
for the scrolling ancestor.
In `rasterizeAnnotationLayer` we load the source CSS files directly, so
these are not processed by Autoprefixer. Since the prefixed rules have
now been removed from the source CSS files, we must manually provide one
prefixed rule that Chrome needs in the overrides CSS file for checkbox
and radio button rendering to work in the reference tests.
SystemJS version 0.21.0 works, but version 0.21.1 breaks with various
`Unable to load transpiler to transpile` errors. Pin the version to
0.21.0 for now to avoid this issue.
We need to pass `disableFontFace` and `nativeImageDecoderSupport`
because Node.js has no native support for `@font-face` and `Image`.
Doing so makes it possible to render e.g., the Tracemonkey paper, which
failed before. I made this PDF file the default because it's also the
default in other examples/demos and because it showcases the
possibilities better than the very simple hello world PDF file.
Building the library with `gulp dist-install` is easier and is already
recommended in the other examples.
*This is a final piece of clean-up of code that I recently wrote, after which I'm done :-)*
When the `getMainThreadWorkerMessageHandler` function was added, in PR 9385, it did so by basically introducing a `web/app.js` dependency in `src/display/api.js` through the `window.pdfjsNonProductionPdfWorker` property[1]. Even though this is limited to non-`PRODUCTION` mode, i.e. `gulp server`, it still seems unfortunate to have that sort of viewer dependency in the API code itself.
With the new, much nicer and shorter, names introduced in PR 9565 we can remove this non-`PRODUCTION` hack and just use `window.pdfjsWorker` in both the viewer and the API regardless of the build mode.
---
[1] It didn't seem correct to piggy-back on the `window.pdfjsDistBuildPdfWorker` property in non-`PRODUCTION` mode.
Please note that this patch *purposely* doesn't add every standard (or semi-standard) page name in existence, but rather only a few common ones. This is done to lessen the burden on localizers, since it's quite possible that all of the page names could need translation (depending on locale).
It's easy to add more standard page sizes in the future, but we should take care to *only* add those that are very commonly used in actual PDF files.
This uses a whitelist, based on the locale, to determine where non-metric units should be used.
Note that the behaviour implemented here seem consistent with desktop PDF viewers (e.g. Adobe Reader), where the pageSizes are *always* displayed with locale dependent units rather than pageSize dependent ones (since the latter would probably be quite confusing).
A couple of basic unit-tests are added, and a manual `isLandscape` check (in `web/base_viewer.js`) is also converted to use the helper function instead.
Jasmine had a major version bump and required a few minor changes in our
booting code. Most notably, using `pending` in a `describe` block is no
longer supported, so we can only return early there. On the positive
side, the unit tests now run in a random order by default, which
eliminates any dependencies between unit tests.
Note that upgrading to Webpack 4 is out of scope for this patch since
the bots cannot work well with the newly generated bundles (both
browsers on both bots do not react within 120 seconds). Webpack 4 is not
faster for us than Webpack 3, so for now there is no need to upgrade.
Test case to exercise the different encodings:
1. Create a file "some file#@%M<br>%25 .pdf"
2. Build the extension with `gulp chromium` and load it in Chrome.
3. Go to `chrome://extensions/` and ensure that the
"Allow access to file URLs" is disabled.
4. Try to open the file from step 1 in Chrome (maybe reload once).
5. PDF.js should be showing a file chooser button.
6. Click on that button and select a different file.
Test: Check that a confirmation dialog pops up that warns about
a different file name. Cancel the dialog.
7. Click on the button again and select the original file.
Test: Check that the file opens as expected.
The current PageLabel dictionary validation code won't catch some (unlikely) forms of corruption. For example: a `Type`/`S` entry being `null`/`0`/empty string, a `P`/`St` entry being `null`/`0`.
Please note: I'm not aware of any bugs caused by the old code, but I've had this patch sitting locally for some time and figured it couldn't hurt to submit it.
The units are currently repeated after each dimension, which seems unnecessary and is also not done in other PDF viewers (such as e.g. Adobe Reader).
Furthermore, the name of the l10n arguments can be simplified slightly, since the name of the strings themselves should be enough information.
Finally, the `width`/`height` should be formatted according to the current locale, as is already done for other strings in the document properties dialog.
The `getPageSizeInches` method was implemented on `PDFDocumentProxy`, which seems conceptually wrong since the size property isn't global to the document but rather specific to each page. Hence the method is moved into `PDFPageProxy`, as `get pageSizeInches` instead to address this.
Despite the fact that new API functionality was implemented, no unit-tests were added. To prevent issues later on, we should *always* ensure that new functionality has at least some test-coverage; something that this patch also takes care of.
The new `PDFDocumentProperties._parsePageSize` method seemed unnecessary convoluted. Furthermore, in the "no data provided"-case it even returned incorrect data (an array, rather than the expected object).
Finally, the fallback strings didn't actually agree with the `en-US` locale. This inconsistency doesn't look too great, and it's thus addressed here as well.
This required changing the import script in two ways:
- we should use the `default` branch and not the `tip` tag since the
latter may refer to another branch than `default` (this is the case for
the `vi` locale, which caused in the files to be overwritten with
incorrect contents since `tip` referred to the
`THUNDERBIRD600b1_2018031614_RELBRANCH` branch);
- we should check if the response code is indeed 200 because recently a
script removed all empty localization files upstream (refer to
https://bugzilla.mozilla.org/show_bug.cgi?id=1443175).
PR #9493 moved from `appConfig.defaultUrl` to `AppOptions.get('defaultUrl')`.
However, it forgot to replace `appConfig.defaultUrl` in chromecom.js,
and as a result the extension is not able to open any PDF file.
This patch fixes that issue.
Chrome 60 and earlier does not include credentials (cookies) in requests
made with fetch, regardless of extension permissions. This was fixed in
61.0.3138.0 by
2e231cf052
This patch disables the fetch backend in all affected Chrome versions.
The browser detection is done by checking for a change that coincides
with the release of Chrome 61.
Test case:
1. Copy the `isChromeWithFetchCredentials` function from the patch.
2. Run it in the JS console of Chrome and verify the return value.
Verified results:
- 49.0.2623.75 - false (earliest supported version by us)
- 60.0.3112.90 - false (last major version affected by bug)
- 61.0.3163.100 - true (first major version without bug)
- 65.0.3325.146 - true (current stable)
Test case 2:
1. Build the extension (`gulp chromium`) and load it in Chrome.
2. Open the developer tools, and open any PDF file.
3. In the "Network tab" of the developer tools, look at "request type".
In Chrome 60-: Should be "xhr"
In Chrome 61+: Should be "fetch"
This function combines the logic of two separate methods into one.
The loop limit is also a good thing to have for the calls in
`src/core/annotation.js`.
Moreover, since this is important functionality, a set of unit tests and
documentation is added.
It's only used in two places in the class and those callsites can
directly get the information from the dictionary, which is more readable
and avoids an additional method call.
With options being moved from the global `PDFJS` object and into `getDocument`, a side-effect is that we're now passing in a fair number of useless parameters to the various transport/network streams.
Even though this doesn't *currently* cause any problems, it nonetheless seem like a good idea to explicitly provide the parameters that are actually necessary.
One additional complication with removing this option from the global `PDFJS` object, is that the viewer currently needs to check `disableAutoFetch` in a couple of places. To address this I'm thus proposing adding a getter in `PDFDocumentProxy`, to allow checking the *actually* used values for a particular `getDocument` invocation.
The `appConfig` contains (mostly) references to various DOM elements, used when initializing the viewer components.
Hence `defaultUrl` seem like a slightly better fit for the new `AppOptions` abstraction, not to mention that it should thus be easier to set/modify it for custom deployments of the default viewer.
The way that various options are handled in the default viewer is currently a bit of a mess (to say the least). Some viewer options reside in the global `PDFJS` object, while others reside in `Preferences`. To make matters worse, some options even exist in both of the two.
Since the goal, with PDF.js version `2.0`, is to reduce our usage of the global `PDFJS` object, we'll instead want pass in the options when initializing the viewer components and when calling API methods (such as `getDocument`).
However given the current state of things in the default viewer, this wouldn't be exactly easy to implement. Hence this patch, which attempts to consolidate the way that viewer (and later API) options are handled by introducing a `AppOptions` singleton that provides *one* centralized way of interacting with the various options in the default viewer.
Unfortunately, as far as I can tell, we still need the ability to adjust certain API options depending on the browser environment in PDF.js version `2.0`. However, we should be able to separate this from the general compatibility code in the `src/shared/compatibility.js` file.
I don't understand why the previous way importing the polyfills didn't work, and I don't have time to try and figure it out, however this patch seems to fix things.
Fixes 9514.
Fixes 9516.
In a1cfa5f4d7c8fcf55e9f3b51a23885dca8782915, the textLayerMode
preference was introduced, to replace the disableTextLayer and
enhanceTextSelection preferences.
As a result, the text selection preference was no longer visible
in Chrome (because preferences are only rendered by default for
boolean preferences, not for enumerations).
This commit adds the necessary bits to
extensions/chromium/options/options.{html,js}
so that the textLayerMode preference can be changed again.
Also, migration logic has been added to move over preferences
from the old to the new names:
- In web/chromecom.js, the logic is added to translate
preferences that were set by an administrator (it is read-only,
so this layer is unavoidable).
- In extensions/chromium/options/migration.js, similar logic is
added, except in this case the preference storage is writable,
so this migration logic happens only once.
The "enhanced text selection" mode is still experimental, so it
has been marked as experimental to signal that there may be bugs.
The list of tasks that block promotion to stable is at #7584.
This partially reverts df0836b9b8bdfe7c8924a7409455abf7ba54b647.
The entry in preferences_schema.json is restored because that is
required to make managed preferences visible to the extension code.
The default key is still removed from default_preferences.json,
because this change only concerns the Chrome extension, not the
other parts of PDF.js. To account for the missing key, the
deprecated key was added back in chromecom.js
The key needs to be restored in preferences_schema.json too,
because that's the only way to make managed preferences visible.
I'm using `Object.assign`, which was introduced in Chrome 45,
so the preference module will break in Chrome 45 and earlier.
This is fine, because we do not support Chrome before 49.
Compared to most other options currently/previously residing on the global `PDFJS` object, some of the Worker specific ones (e.g. `workerPort`/`workerSrc`) probably cannot be moved into options provided directly when initializing e.g. `PDFWorker`.
The reason is that in some cases, e.g. the Webpack examples, we try to provide Worker auto-configuration and I cannot see a good solution for that use-case if we completely remove the globally available Worker configuration.
However inline with previous patches for PDF.js version `2.0`, it does seem like a worthwhile goal to move away from storing options directly on the global `PDFJS` object, since that is a pattern we should avoid going forward. Especially since one of the (eventual) goals is to attempt to *completely* remove the global `PDFJS` object, and rely solely on exporting/importing the needed functionality.
By introducing the `GlobalWorkerOptions` we thus have larger flexibility in the future, if/when the global `PDFJS` object will be removed.
With PDF.js version `2.0` we'll only support browsers with built-in `TypedArray` functionality, hence there doesn't seem to be any good reason not to implement this now.
Fixes 4888.
In order to simplify things, the undocumented `enableStats` option was removed and `pdfBug` is now instead used to enabled general debugging *and* page request/rendering stats.
Considering that in the default viewer the `stats` was only used when debugging was also enabled, this simplification (code wise) definitely seem worthwhile to me.
Rather than having two different (but connected) options for the textLayer, I think that it makes sense to try and unify this. For example: currently if `disableTextLayer === true`, then the value of `enhanceTextSelection` is simply ignored.
Since PDF.js version `2.0` already won't be backwards compatible in lots of ways, I don't think that we need to worry about migrating existing preferences here.
This removes the `PDFJS.externalLinkTarget`/`PDFJS.externalLinkRel` dependency from the viewer components, but please note that as a *temporary* solution the default viewer still uses it.
This removes the `PDFJS.imageResourcesPath` dependency from the viewer components and the test-suite, but please note that as a *temporary* solution the default viewer still uses it.
This removes the `PDFJS.maxCanvasPixels` dependency from the viewer components, but please note that as a *temporary* solution the default viewer still uses it.
This removes the `PDFJS.useOnlyCssZoom` dependency from the viewer components, but please note that as a *temporary* solution the default viewer still uses it.
Unfortunately, as far as I can tell, we still need the ability to adjust certain viewer options depending on the browser environment in PDF.js version `2.0`. However, we should be able to separate this from the general compatibility code in the `src/shared/compatibility.js` file.
The reason for the bug is that we're only computing a checksum of the image data itself, but completely ignore the inline dictionary. The latter is important, since in practice it's not uncommon for inline images to be identical but use e.g. different ColourSpaces.
There's obviously a couple of different ways that we could compute a hash/checksum of the dictionary.
Initially I tried using `MurmurHash3_64` to compute a hash of the keys/values in the dictionary. Unfortunately this approach turned out to be *way* too slow in practice, especially for PDF files with a huge number of inline images; in particular issue 2618 would regresses quite badly with this solution.
The solution that is instead implemented in this patch, is to compute a checksum of the dictionary contents. While this is a much simpler, not to mention a lot more efficient, solution there's one drawback associated with it:
If the contents of inline image dictionaries are ordered differently, they will not be considered equal with this approach which could thus lead to failures to cache repeated inline images. In practice this doesn't seem to be a problem in any of the PDF files I've tested, and generally I'd rather err on the side of *not* caching given that too aggressive caching can easily lead to rendering bugs.
One small, but somewhat annoying, complication is that by the time `Parser.makeInlineImage` is called, we no longer know the *exact* stream position where the inline image dictionary starts. Having access to that information is crucial here, and the easiest solution I could come up with is to track this in the current `Lexer` instance.[1]
With the patch, we're thus able to fix the referenced issues without incurring large regressions in problematic cases such as issue 2618.
Fixes 9398; also improves/fixes the `issue8823` reference test.
---
[1] Obviously I'd have preferred if this patch could be limited to `Parser.makeInlineImage`, without the need for this "hack", but I'm not sure what that'd look like here.
This rule is available from https://www.npmjs.com/package/eslint-plugin-mozilla, and is enforced in mozilla-central. Note that we have the necessary `Array`/`String` polyfills and that most cases have already been fixed, see PRs 9032 and 9434.
This rule is available from https://www.npmjs.com/package/eslint-plugin-mozilla, and is enforced in mozilla-central. Note that we have a polyfill for `ChildNode.remove()` and that most cases have already been fixed, see PRs 8056 and 8138.
Fallback to the built-in JPEG decoder when browser decoding fails, and attempt to handle JPEG images with DNL (Define Number of Lines) markers (issue 8614)
The code responsible for highlighting/scrolling thumbnails is quite old, and parts of it hasn't really changed much over time.
In particular, the `scrollThumbnailIntoView` method is currently using `document.querySelector` in order to find both the new/current thumbnail element. This seems totally unnessary, since we can easily keep track of the current thumbnail (similar to the "regular" viewer) and thus avoid unnecessary DOM look-ups.
Furthermore, note how the `PDFThumbnailView` constructor is currently highlighting the *first* thumbnail. This is yet another leftover from older viewer code, which we can now remove and instead take care of in `PDFThumbnailViewer.setDocument`.
Given that `PDFThumbnailView` does not, nor should it, have any concept of which thumbnail is currently selected, this change also improves the general structure of this code a tiny bit.
First of all, note how in both `fetch_stream.js` and `node_stream.js` we always overwrite the `this._contentLength` property even when the response headers doesn't actually contain any (valid) length information. This could thus result in the `length` parameter, as passed to the network stream, being completely ignored despite having no better information available.
Secondly, in `node_stream.js` the `this._isRangeSupported` property wasn't always updated correctly based on the response headers.
Please refer to the specification, found at https://www.w3.org/Graphics/JPEG/itu-t81.pdf#page=49
Given how the JPEG decoder is currently implemented, we need to know the value of the scanLines parameter (among others) *before* parsing of the SOS (Start of Scan) data begins.
Hence the best solution I could come up with here, is to re-parse the image in the *hopefully* rare case of JPEG images that include a DNL (Define Number of Lines) marker.
Fixes 8614.
This works by making `PartialEvaluator.buildPaintImageXObject` wait for the success/failure of `loadJpegStream` on the API side *before* parsing continues.
Please note that in practice, it should be quite rare for the browser to fail loading/decoding of a JPEG image. In the general case, it should thus not be completely surprising if even `src/core/jpg.js` will fail to decode the image.
Since `loadJpegStream` is only used at a *single* spot in the code-base, and given that it's very heavily tailored to the calling code (since it relies on the data structure of `PDFObjects`), this patch simply inlines the code in `src/display/api.js` instead.
This method currently requires a fair number of parameters, which creates quite unwieldy call-sites. When invoking `buildPaintImageXObject`, you have to remember not only which arguments to supply, but also the correct order, to prevent run-time errors.
After the latest update of the minimum supported Firefox version, the development addon is no longer being signed. Hence this patch attempts to address that, by pinning the `minVersion` to a specific version number instead.
*Please note:* The version numbers were taken from https://addons.mozilla.org/en-US/firefox/pages/appversions/, so it ought to have worked.
At this point in time, trying to keep the development addon compatible with prior Firefox versions is already quite difficult and will become even harder very soon.
Please keep in mind that since Firefox 57, only WebExtensions are allowed/possible to install. The only exceptions are Firefox Nightly, with the `xpinstall.signatures.required` preference[1] set to `false`, and the (as of this writing) current ESR release.[2]
With the current compatibility situation, we thus need to effectively support both Nightly *and* ESR in the addon, while trying to keep up with current/upcoming changes in `mozilla-central`. With old addons no longer being officially supported, the amount of old code being removed/refactored is now increasing quite quickly.
*Please note:* The changes proposed here was to a large extent prompted by bugs such as:
- https://bugzilla.mozilla.org/show_bug.cgi?id=1431533 (already landed)
- https://bugzilla.mozilla.org/show_bug.cgi?id=767640 (already landed)
- https://bugzilla.mozilla.org/show_bug.cgi?id=1432992 (still open)
Upstreaming all of those changes for the `MOZCENTRAL` version of PDF.js, while also keeping the Firefox addon running in older versions of the browser, would quickly become non-trivial.
Since we're using the ruleset from `eslint-plugin-mozilla` for the addon files, which is enforced in `mozilla-central`, we must ensure that the neccessary changes are upstreamed to the PDF.js repo such that the `mozilla-central` version of PDF.js can still be updated without failures.
Trying to feature detect, for the `FIREFOX` build target, some of the changes in the referenced bugs would probably become both quite messy and (not to mention) difficult. E.g. with the upcoming and automatically defined `Cc, Ci, Cu, Cr` variables, any sort of feature detection might be tricky since those need to be defined in the global scope of the files in question.
Finally, given the amount of effort that we'd now need to spend to even attempt to keep the Firefox addon compatible, I just don't think it's worth the effort any more. Especially since the number of people that have, thus far, been doing this work is *very* low and those resources would be better spend elsewhere.
Unfortunately, this probably means that the development addon will no longer be compatible with release versions of Seamonkey. However the README has already mentioned, for quite some time, that support isn't guaranteed.
*In closing:* For all of the reasons mentioned above, I thus propose that we reduce the maintenance burden of the Firefox addon by only supporting the current Firefox Nightly.
---
[1] While the preference exists, and can be toggled, its value is ignored in non-Nightly versions of Firefox.
[2] There's unbranded builds of e.g. the release/beta versions of Firefox, where old and non-WebExtensions addons can be installed. However those builds can probably be assumed to be officially unsupported, and thus not recommended for users.
This commit is the first step for extracting a base class for the
`AES128Cipher` and the `AES256Cipher` classes. The objective here is to
make code changes (not altering the logic) to make the implementations
as similar as possible as found by creating a diff of both classes.
In particular, we extract the key size and cycles of repetitions
constants since they are different for AES-128 and AES-256. Moreover, we
rename functions to be similar.
In the `AES256Cipher` class, there was an additional assignment to
`this` in the decryption function. However, this was unnecessary because
the assignment would also be done when the loop was exited.
In the JPEG images in the referenced PDF file, the DHT (Define Huffman Tables) segments contain more data than expected based on the length parameter.
Fixes 9425.
Despite this patch removing the `disableWorker` option itself, please note that we'll still fallback to loading the worker file(s) on the main-thread when running in environments without proper Web Worker support.
Furthermore it's still possible, even with this patch, to force the use of fake workers by manually loading the necessary file using a `<script>` tag on the main-thread.[1]
That way, the functionality of the now removed `SINGLE_FILE` build target and the resulting `build/pdf.combined.js` file can still be achieved simply by adding e.g. `<script src="build/pdf.worker.js"></script>` to the HTML (obviously with the path adjusted as needed).
Finally note that the `disableWorker` option is a performance footgun, and unfortunately many existing third-party examples actually use it without providing any sort of warning/justification.
---
[1] This approach is used in the default viewer, since certain kind of debugging may be easier if the code is running directly on the main-thread.
This method returns the currently used `workerSrc`, which thus allows obtaining the fallback `workerSrc` value (e.g. when the option wasn't set by the user).
Please note that this build target, and the resulting `build/pdf.combined.js` file, is equivalent to setting the `PDFJS.disableWorker` option to `true` which is a performance footgun.
https://crbug.com/362061 was fixed in Chrome 36, and the lowest
supported Chrome version in the extension is Chrome 49, so the
work-around for a filesystem:-bug in chromecom can be removed.
Merge ftp and file handler now their implementations are identical.
Remove redundant comment (the referenced Chrome bug has been fixed
a long time ago - https://crbug.com/302548 ).
This was introduced in #3582 to work around https://crbug.com/274024 .
The bug in Chrome has been fixed a long time ago (at least 33), so let's
simplify the code.
Add comments with supported browser versions where missing.
Method:
- Use MDN compat tables if available.
- Otherwise test in Chrome (31+) otherwise.
(the Chrome Web Store does not update older versions of
Chrome, so probably nobody is interested in even older
versions, even though there is an existing comment for
Chrome<29 at `document.currentScript`).
It turns out that PR 9245 unfortunately broke benchmarking completely, sorry about that!
The bug is that we were attempting to reset the current instance of `StatTimer`, instead of creating a new one as was previously done. By resetting the current instance, the `StatTimer` data fetched in `test/driver.js` is now wiped out since it points to the *same* underlying object.
This re-use of a `StatTimer` instance was asked for during review, and unfortunately I didn't test this thoroughly enough before submitting the final version of the PR.[1]
---
[1] Note that while I did test the benchmarking scripts with that PR *before* initially submitting it, I did however forget to do that after addressing the review comments which might explain why this problem went unnoticed.
Remove "returns null when content disposition is form-data".
The name of the test is already misleading: It suggests that
the return value is null if the Content-Disposition starts with
"form-data". This is not the case, anything with the "filename"
parameter is accepted.
So, to correct this, one would have to rephrase the test description to
"returns null when content disposition has no filename".
But this is already tested by the test called
"gets the filename from the response header".
So, remove the test.
This patch updates the `IPDFStreamReader` interface and ensures that the interface/implementation of `network.js`, `fetch_stream.js`, `node_stream.js`, and `transport_stream.js` all match properly.
The unit-tests are also adjusted, to more closely replicate the actual behaviour of the various actual `IPDFStreamReader` implementations.
Finally, this patch adjusts the use of the Content-Disposition filename when setting the title in the viewer, and adds `PDFDocumentProperties` support as well.
The `fetch` API is only supported for http(s), even in Chrome extensions.
Because of this limitation, we should use the XMLHttpRequest API when the
requested URL is not a http(s) URL.
Fixes#9361
These were removed in PR 9170, since they were unused in the browsers that we'll support in PDF.js version `2.0`.
However looking at the output of Travis, where a subset of the unit-tests are run using Node.js, there's warnings about `btoa` being undefined. This doesn't appear to cause any errors, which probably explains why we didn't notice this before (despite PR 9201).
When navigating away from the viewer, there's no good reason for disallowing replacement of a *temporary* history entry (in addition to empty ones).
Given that the current position is temporarily added to the browser history using a (short) timeout, the history entry will most likely already be correct when the 'pagehide' event fires. However, if the user is quick enough that might not always be the case, in which case the adjusted logic may help.
Please refer to the PDF specification, in particular http://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#G7.3801570
> A colour space shall be specified in one of two ways:
> - Within a content stream, the CS or cs operator establishes the current colour space parameter in the graphics state. The operand shall always be name object, which either identifies one of the colour spaces that need no additional parameters (DeviceGray, DeviceRGB, DeviceCMYK, or some cases of Pattern) or shall be used as a key in the ColorSpace subdictionary of the current resource dictionary (see 7.8.3, "Resource Dictionaries"). In the latter case, the value of the dictionary entry in turn shall be a colour space array or name. A colour space array shall never be inline within a content stream.
>
> - Outside a content stream, certain objects, such as image XObjects, shall specify a colour space as an explicit parameter, often associated with the key ColorSpace. In this case, the colour space array or name shall always be defined directly as a PDF object, not by an entry in the ColorSpace resource subdictionary. This convention also applies when colour spaces are defined in terms of other colour spaces.
This is something that I noticed while attempting to debug https://bugzilla.mozilla.org/show_bug.cgi?id=1374945.
Just looking at the code, the `YRsiz` parameter seemed immediately wrong and the fact that every component used the *same* data also looked strange.
Comparing with the specification, see https://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-REC-T.800-200208-S!!PDF-E&type=items#page=37, confirmed that this is indeed incorrect.
Note that I haven't got any example of a PDF file that is fixed by this patch, but that might be more luck than anything else. Manually checking a couple of files with included JPEG 2000 images, the `Csiz`/`XRsiz`/`YRsiz` parameters were `1` which could explain why this hasn't been an issue before.
Obviously we shouldn't generally make changes to `core` code without adding tests, but in this case I'm simply not sure how to obtain/create one. However, since the existing code doesn't make sense this patch could hopefully be deemed acceptable anyway.
Since multiple empty lines is virtually unused in the code-base, and the few cases that do exist look like "typos", let's enforce greater consistency here; please see https://eslint.org/docs/rules/no-multiple-empty-lines.
It is only used in a few places to handle prefixing style properties if
necessary. However, we used it only for `transform`, `transformOrigin`
and `borderRadius`, which according to Can I Use are supported natively
(unprefixed) in the browsers that PDF.js 2.0 supports. Therefore, we can
remove this class, which should help performance too since this avoids
extra function calls in parts of the code that are called often.
The `gulp-util` module is now deprecated and authors are asked to stop
using it (refer to https://medium.com/gulpjs/gulp-util-ca3b1f9f9ac5 for
more information).
PDF.js does not rely on it that much, fortunately, so it's relatively
easy for us to remove the dependency. This patch does that by making the
following changes:
- Require `gulp-zip` version 4.1.0 or higher since they already removed
their `gulp-util` dependency in that version.
- Replace `gulp-util.log` with the `fancylog` module as recommended in
the article above.
- Replace `gulp-util.File` with the `Vinyl` module as recommended in the
article above.
The only change I had to make for Vinyl is removing the `base` and `cwd`
lines since they may not be empty strings anymore. This way we fall back
to the defaults Vinyl provides, which for us doesn't matter since we
move the file afterwards anyway. Moreover, I used `vfs` for `vinyl-fs`
in the `Gulpfile` to avoid confusion with `vinyl` (which is also how the
documentation names the variable).
This is all we can do on our side; the other modules that still use
`gulp-util` must be updated upstream.
I've been looking into the remaining point in 8637 about blurry images, to see if we could perhaps improve the rendering quality slightly there. After quite a bit of debugging, it seems that the issue is limited to certain progressive JPEG images.
As mentioned previously, I've got no detailed knowledge of the JPEG format, but this patch does seem to improve things quite a bit for the images in question.
Squinting at https://searchfox.org/mozilla-central/rev/6c33dde6ca02b389c52e8db3d22494df8b916f33/media/libjpeg/jdphuff.c#492-639, it seems reasonable that we should take the sign of the data into account. Furthermore, looking at the specification in https://www.w3.org/Graphics/JPEG/itu-t81.pdf#page=118, the "F.2.4.3 Decoding the binary decision sequence for non-zero DC differences and AC coefficients" section even contains a description of this (even though I cannot claim to really understand the details).
The bug that this patch fixes is limited to the built-in JPEG decoder, and was unearthed by PR 9260. The underlying issue has existed since PR 6984, where the contents of this patch ought to have been included (if it weren't for the fact that we had no *easy* way to test `src/core/jpg.js` back then).
*Please note:* The slight movement in the reference test is a result of using the `src/core/jpg.js` decoder, rather than the native browser one.
This was an oversight in PR 9095, which unfortunately breaks rendering in some PDF files (e.g. the one from issue 6737).
It thus appears that we don't have any test-coverage for this code-path, and given the relative complexity of the PDF files affected by this bug I wasn't able to easily create a reduced test-case.
*Please note:* The linked test-case included in this patch is currently *not* rendered correctly (that'd be the PR 6606), but it at least gives us some test-coverage here.
Initially I just implemented the unit tests, but quickly found that they
were failing my expectation of having a size of 256 items. Some of them
did contain 256 items and some did not. I looked up various resources
and figured that they indeed all need to have 256 items. One of the good
resources is https://github.com/davidben/poppler/blob/master/poppler/FontEncodingTables.cc
Aside from some missing `notdef` (empty string) entries at the end of
the arrays, which I assume causes issues since it may cause
out-of-bounds array access which in JavaScript gives `undefined`, there
was a `notdef` entry missing in the `MacExpertEncoding`, causing the
entries after that to be shifted. This fix for this is similar to the
one in #8589.
The unit tests verify that, for known encoding names, the return value
is not only an array, but that it is also of the right length and
contains only strings.
The PDF file in the issue uses a number of *embedded* versions of Lucida fonts, but for some reason does *not* embed the LucidaSans-Demi font. According to https://en.wikipedia.org/wiki/Lucida#Usages that one should be bold, so we can at least improve rendering here (even though it won't look perfect).
Fixes 9291.
I recall being confused as to the purpose of the `encrypted` property all the way back when working on PR 4750.
Looking at the history, this property was added in PR 1698 when password support was added to the API/viewer. However, its only purpose seem to have been to facilitate the addition of a `isEncrypted` function in the API. That function never, as far as I can tell, saw any use and was unceremoniously removed in PR 4144.
Since we want to avoid sending all non-essential data early during initial document loading (e.g. PR 4750), it seems correct to get rid of the `encrypted` property. Especially since it hasn't even been exposed in the API for over three years, with no complaints that I'm aware of.
Finally note that the `encrypt` property on the `XRef` instance isn't tied to the code that's being removed here. Given that we're calling `PDFDocument.parse` during `createDocumentHandler` in the worker which, via `PDFDocument.setup`, calls `XRef.parse` where the `Encrypt` data (if it exists) is always parsed.
This patch refactors the searching for 'endobj', to try and find the next occurance of "obj" and then check if it was in fact an 'endobj' and continue searching otherwise.
This approach is used to avoid having to first find 'endobj', and then re-check the entire contents of the object and having to run (potentially expensive) regular expressions on arbitrary long strings.
Fixes 9105.
Please note that while this could be considered a regression in user-facing behaviour, I'm not convinced that it's really a regression as such since prior to PR 8912 the Metadata would fail to parse (with an XML error) and thus be ignored when setting the viewer title.
With the refactored Metadata parsing we're now able to parse this, which uncovered issues with a subset of broken Ghostscript Metadata that uses HTML character names.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1424938
It is quite confusing that the custom function is called `log2` while it
actually returns the ceiling value and handles zero and negative values
differently than the native function.
To resolve this, we add a comment that explains these differences and
make the function use the native `Math` functions internally instead of
using our own custom logic. To verify that the function does what we
expect, we add unit tests.
All browsers except for IE support `Math.log2` for quite a long time
already (see https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Math/log2).
For IE, we use the core-js polyfill.
According to the microbenchmark at https://jsperf.com/log2-pdfjs/1,
using the native functions should also be faster, in my testing almost
six times as fast.
Note that no other image stream implements a special `getBytes` method, which makes `JpegStream` look somewhat odd.
I'm actually not sure what purpose this methods serves, since I successfully ran all tests locally with it commented out. Furhermore, I also ran tests with an added `if (length && length !== this.bufferLength) { throw new Error('length mismatch'); }` check, and didn't get a single test failure in that case either.
Looking at the history, it seems that this code originated back in PR 4528, but as far as I can tell there's no mention in either commit messages nor PR comments of why it was necessary to add a "special" `getBytes` function for the `JpegStream`.
My assumption is that there's a good reason why this method was added, e.g. to address a *specific* regression in one of the reference tests. However, I did check out commit 58f697f977 locally and ran tests with this method commented out, and there didn't seem to be any image-related failures in that case either!?
Hence I'm suggesting that we attempt to simplify this code slightly be removing this special `getBytes` method. However, please note that there's perhaps a *small* risk of regressions in an edge-case where we currently have insufficient test-coverage.
Unless the debugging tools (i.e. `PDFBug`) are enabled, or the `browsertest` is running, the `PDFPageProxy.stats` aren't actually used for anything.
Rather than initializing unnecessary `StatTimer` instances, we can simply re-use *one* dummy class (with static methods) for every page. Note that by using a dummy `StatTimer` in this way, rather than letting `PDFPageProxy.stats` be undefined, we don't need to guard *every* single stats collection callsite.
Since it wouldn't make much sense to attempt to use `PDFPageProxy.stats` when stat collection is disabled, it was instead changed to a "private" property (i.e. `PDFPageProxy._stats`) and a getter was added for accessing `PDFPageProxy.stats`. This getter will now return `null` when stat collection is disabled, making that case easy to handle.
For benchmarking purposes, the test-suite used to re-create the `StatTimer` after loading/rendering each page. However, modifying properties on various API code from the outside in this way seems very error-prone, and is an anti-pattern that we really should avoid at all cost. Hence the `PDFPageProxy.cleanup` method was modified to accept an optional parameter, which will take care of resetting `this.stats` when necessary, and `test/driver.js` was updated accordingly.
Finally, a tiny bit more validation was added on the viewer side, to ensure that all the code we're attempting to access is defined when handling `PDFPageProxy` stats.
While this doesn't actually fix the underlying issue, it should prevent the ESLint errors and thus make future PDF.js updates easier.
Compared to updating (and testing) the preprocessor, this seems like a reasonable workaround given its simplicity.
There's a number of issues with the fonts in the referenced PDF file. First of all, they contain broken `ToUnicode` data (`NUL` bytes all over the place). However even if you skip those, the `ToUnicode` data appears to contain nothing but a `IdentityH` CMap which won't help provide a proper glyph mapping.
The real issue actually turns out to be that the PDF file uses the "Calibri" font[1], but doesn't include any font files. Since that one isn't a standard font, and uses a fairly different CID to GID map compared to the standard fonts, we're not able to render the file even remotely correct.
To work around this, I'm thus proposing that we include a (incomplete) glyph map for Calibri, and fallback to the standard Helvetica font. Obviously this isn't going to look perfect, but it's really the best that we can hope to achieve given that the PDF file is missing the necessary font data.
Finally, please note that none of the PDF readers I've tried (Adobe Reader, PDFium in Chrome) were able to extract the text (which isn't very surprising, given the broken `ToUnicode` data).
Fixes 9195.
---
[1] According to Wikipedia, see https://en.wikipedia.org/wiki/Calibri, Calibri is (primarily) a Windows font.
For sufficiently large page sizes, always limiting the minimum zoom level to 25% seem a bit too high. One example is pages, e.g. the first one, in:
Hence I think that it makes sense to lower `MIN_SCALE` slightly, since other PDF viewers (e.g. Adobe Reader) isn't limiting the minimum zoom level as aggressively. Obviously this will allow a greater number of pages to be visible at the same time in the viewer, but given that they will be small that shouldn't be an issue.
Note also that e.g. the `page-fit`/`page-width` zoom levels already allow `< MIN_SCALE` values, so I don't see why we shouldn't allow users the same functionality directly.
In some fonts, the included `ToUnicode` data is incomplete causing text-selection to not work properly. For simple fonts that contain encoding data, we can manually build a `ToUnicode` map to attempt to improve things.
Please note that since we're currently using the `ToUnicode` data during glyph mapping, in an attempt to avoid rendering regressions, I purposely didn't want to amend to original `ToUnicode` data for this text-selection edge-case.
Instead, I opted for the current solution, which will (hopefully) give slightly better text-extraction results in PDF file with incomplete `ToUnicode` data.
According to the PDF specification, see [section 9.10.2](http://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#G8.1873172):
> A conforming reader can use these methods, in the priority given, to map a character code to a Unicode value.
> ...
Reading that paragraph literally, it doesn't seem too unreasonable to use *different* methods for different charcodes.
Fixes 8229.
This was added, during the refactoring in PR 8556, to avoid outright breaking third-party users of the default viewer.
With PDF.js version `2.0`, where we're making API changes that aren't backwards compatible, we ought to be able to remove this piece of viewer code as well.
Split the existing `WebGLUtils` in two classes, a private `WebGLUtils` and a public `WebGLContext`, and utilize the latter in the API to allow various code to access the methods of `WebGLUtils`
We've never attempted to limit the errors displayed in the default viewer to the current PDF file, but that's not really been a problem before. However after PR 7926, it's now possible to get password related error messages for *previously* opened PDF files in the default viewer.
**STR:**
1. Open a password protected PDF file, e.g. `issue6010_1.pdf` from the test-suite.
2. Cancel the password prompt.
3. Open any new PDF file in the viewer.
**AR:**
The error UI is displayed, with a `No password given` message.
**ER:**
No error displayed, since it's only relevent for a now closed PDF file.
This is obviously a minor issue, caused by us now rejecting the still pending `pdfLoadingTask` during the `PDFViewerApplication.close` call, but I don't think that it (generally) makes sense to show errors if they're not relevant to the *currently* displayed PDF file.
adding commas each in two sentences in order to make readers fully understand when reading index.md and to organize the context properly based on grammatical rules.
In order to move viewer related options from the global `PDFJS` object and into the initialization of the relevant components, we'll need to parse the hash parameters *before* calling `PDFViewerApplication._initializeViewerComponents`.
We're currently disabling range requests and streaming for a number of configurations. A couple of those will no longer be supported (with PDF.js version 2.0), hence we ought to be able to clean up the compatibility code slightly.
This compatibility code is only relevant for browsers that will no longer be supported (with PDF.js version 2.0), hence we ought to be able to remove it.
The only reason for adding this parameter in the first place, all the way back in PR 4074, was that the "maintain document position on zooming" feature was landed and backed out a couple of times before it finally stuck.
Hence it seemed, at the time, like a good idea to have a simple way to disable that behaviour. However, that was almost four years ago, and it's just not likely that we'd want/need to ever disable it now.
Furthermore I really cannot imagine why anyone would actually *want* to reset the position whenever zooming occurs, since it results in a quite annoying UX.
*So, to summarize:* Based on the above, I think that we should try to remove this parameter now. On the off chance that anyone complains, re-adding it shouldn't be difficult.
If the sidebar is resized such that the thumbnails are displayed in multiple columns, then scrolling the currently active thumbnail into view doesn't work correctly in some cases.
The reason is that the code in `PDFThumbnailViewer.scrollThumbnailIntoView` implicitly assumes that the thumbnails will be present in just *one* column. Since that may no longer be the case, it's not sufficient to simply check if the thumbnail is visible. Instead we must explicitly check that *all*, i.e. 100 percent, of the thumbnail is already visible, and otherwise scroll it into view.
*This patch is the result of me starting to look into moving parameters from `PDFJS` into `getDocument` and other API methods.*
When familiarizing myself with the code, the signatures of the various network streams seemed to be unnecessarily cumbersome since `disableRange` is currently handled separately from other parameters.
I'm assuming that the explanation for this is probably "for historical reasons", as is often the case. Hence I'd like to clean this up *before* we start the larger, and more invasive, `PDFJS` parameter re-factoring.
The interface of all of the "image" streams look kind of weird, and I'm actually a bit surprised that there hasn't been any errors because of it.
For example: None of them actually implement `readBlock` methods, and it seems more luck that anything else that we're not calling `getBytes()` (without providing a length) for those streams, since that would trigger a code-path in `getBytes` that assumes `readBlock` to exist.
To address this long-standing issue, the `ensureBuffer` methods are thus renamed to `readBlock`. Furthermore, the new `ensureBuffer` methods are now no-ops.
Finally, this patch also replaces `var` with `let` in a number of places.
By making use of modern CSS features, in this case [CSS variables](https://developer.mozilla.org/en-US/docs/Web/CSS/Using_CSS_variables), implementing sidebar resizing is actually quite simple. Not only will the amount of added code be fairly small, but it should also be easy to maintain since there's no need for complicated JavaScript hacks in order to update the CSS. Another benefit is that the JavaScript code doesn't need to make detailed assumptions about the exact structure of the HTML/CSS code.
Obviously this will not work in older browsers, such as IE, that lack support for CSS variables. In those cases sidebar resizing is simply disabled (via feature detection), and the resizing DOM element hidden, and the behaviour is thus *identical* to the current (fixed-width) sidebar.
However, considering the simplicity of the implementation, I really don't see why limiting this feature to "modern" browsers is a problem.
Finally, note that a few edge-cases meant that the patch is a bit larger than what the basic functionality would dictate. Among those is first of all proper RTL support, and secondly (automatic) resizing of the sidebar when the width of the *entire* viewer changes. Another, pre-existing, issue fixed here is the incomplete interface of `NullL10n`.
*Please note:* This patch has been successfully tested in both LTR and RTL viewer locales, in recent versions of Firefox and Chrome.
Fixes 2072.
https://eslint.org/docs/rules/no-var
Please note that two files were excluded:
1. `web/debugger.js`, since there's code in other files that currently depend on the global availability of code in `web/debugger.js`. Furthermore, since that file isn't used in production, doing a ES6 conversion probably isn't a priority.
2. `web/grab_to_pan.js`, since that file could be considered to be "external" code. We have made smaller changes to that file over the years, however doing a full ES6 `class` conversion might be a step too far!?
In the PDF file, the `ToUnicode` data first maps the hyphen correctly, and then *overwrites* it to point to the softhyphen instead. That one cannot be rendered in browsers, and an empty space thus appear instead.
Fixes 9084.
Based on a number of opened issues, it seems that the "Is an extension" field might be causing some confusion as to its meaning. Without providing too much detail, I'm still thinking that we could attempt to clarify that it's referring to either of the *browser* extensions.
Nothing uses this option anymore, so setting it is a no-op now. We can
safely remove it.
Use `SKIP_BABEL` (instead of `PDFJS_NEXT`) now if you want to skip Babel
translation for a build.
Since we're already using core-js elsewhere in `compatibility.js`, we can reduce the amount of code we need to maintain ourselves.
https://github.com/zloirock/core-js#weakmap
If we want to (eventually) make it possible to resize the sidebar, then having its width indirectly affect the toolbar is going to wreck havoc on the media queries used to show/hide buttons in the main toolbar (since many of them depend on the toolbar state, and thus its width).
Updating all of the media queries dynamically with JavaScript seems like a non-starter, given that it'd cause *very* messy code. It thus seem to me that we'd need to fix the position of the sidebar, to have any hope of (in the short term) addressing issue 2072.
Hence, I'm suggesting that the we always layout the sidebar in a consistent vertical position, and only animate the `viewerContainer` rather than the entire `mainContainer`.
Fixes 4052.
Fixes bug 850591.
Some PDF files contain JavaScript actions that consist of nothing more that one, or possibly several, empty string(s). At least to me, printing a warning/showing the fallback seems completely unnecessary in that case.
Furthermore, this patch also makes use of an early `return`, so that we no longer will attempt to check for printing instructions when no JavaScript is present in the PDF file.
*Note:* It would perhaps make sense to change the API/core code, such that we ignore empty entries there instead. However, that would probably be considered a breaking changing with respect to backwards compatibility, hence this simple viewer only solution.
Fixes 5767.
I don't know if this is a regression, but I noticed earlier today that depending on the initial scale *and* sidebar state, the `annotationLayer` of the first rendered page may end up duplicated; please see screen-shot below.
[screen-shot]
I can reproduce this reliable with e.g. https://arxiv.org/pdf/1112.0542v1.pdf#zoom=page-width&pagemode=bookmarks.
When the document loads, rendering of the first page begins immediately. When the sidebar is then opened, that forces re-rendering which thus aborts rendering of the first page.
Note that calling `PDFPageView.draw()` will always, provided an `AnnotationLayerFactory` instance exists, call `AnnotationLayerBuilder.render()`. Hence the events described above will result in *two* such calls, where the actual annotation rendering/updating happens asynchronously.
For reasons that I don't (at all) understand, when multiple `pdfPage.getAnnotations()` promises are handled back-to-back (in `AnnotationLayerBuilder.render()`), the `this.div` property seems to not update in time for the subsequent calls.
This thus, at least in Firefox, result in double rendering of all annotations on the first page.
Obviously it'd be good to find out why it breaks, since it *really* shouldn't, but this patch at least provides a (hopefully) acceptable work-around by ignoring `getAnnotations()` calls for `AnnotationLayerBuilder` instances that we're destroying (in `PDFPageView.reset()`).
*It appears that this accidentally broken in PR 8775.*
Note that `PDFHistory.forward` is only used with certain named actions, and these aren't that commonly used, which ought to explain why this error managed to sneak in.
Steps to reproduce the issue (and verify the fix):
1. Navigate to e.g. http://mirrors.ctan.org/info/lshort/english/lshort.pdf
2. Click on a couple of links, or outline items, such that the history is populated with a few entries.
3. In the console, execute `PDFViewerApplication.pdfHistory.back()` one or more times, thus navigating back to a previous viewer position.
4. In the console, execute `PDFViewerApplication.pdfHistory.forward() one or more times.
At the last step above, no (forward) navigation happens with the current `master`; now compare with this patch.
Moreover, adjust one linked test case that did not conform to the
standard Web Archive URL format and adjust one linked test case because
the link was dead.
To support this, the following changes have to be made as well:
- Ignore `package-lock.json` since NPM creates it, but we should not
have it in the repository.
- Switch from `babel-preset-es2015` to `babel-preset-env` to resolve
the deprecation warning in the test logs. The latter is more recent
and flexible, but should be the same functionality-wise.
- `transform` now needs to have the `utf-8` encoding option provided.
If not given, it will call the callback with a `Buffer` object,
which results in an unhandled promise rejection since what is
returned from the callback is a string, not a `Buffer`.
This patch makes use of the existing `ignoreErrors` property in `src/core/evaluator.js`, see PRs 8240 and 8441, thus allowing us to attempt to recovery as much as possible of a page even when it contains broken XObjects.
Fixes 8702.
Fixes 8704.
Split the existing `PDFFunction` in two classes, a private `PDFFunction` and a public `PDFFunctionFactory``, and utilize the latter in `PDFDocument` to allow various code to access the methods of `PDFFunction`
*Follow-up to PR 8909.*
This requires us to pass around `pdfFunctionFactory` to quite a lot of existing code, however I don't see another way of handling this while still guaranteeing that we can access `PDFFunction` as freely as in the old code.
Please note that the patch passes all tests locally (unit, font, reference), and I *very* much hope that we have sufficient test-coverage for the code in question to catch any typos/mistakes in the re-factoring.
The `inline` parameter is passed to a number of methods/functions in `PDFImage`, despite not actually being used. Its value is never checked, nor is it ever assigned to the current `PDFImage` instance (i.e. no `this.inline = inline` exists).
Looking briefly at the history of this code, I was also unable to find a point in time where `inline` was being used.
As far as I'm concerned, `inline` does nothing more than add clutter to already very unwieldy method/function signatures, hence why I'm proposing that we just remove it.
To further simplify call-sites using `PDFImage`/`NativeImageDecoder`, a number of methods/functions are changed to take Objects rather than a bunch of (somewhat) randomly ordered parameters.
I don't have a good example at hand right know, but I recall seeing custom deployments of PDF.js that bundle a *specific* version of the `build/pdf.js` file and then set `PDFJS.workerSrc` to point to https://mozilla.github.io/pdf.js/build/pdf.worker.js.
That practice seems really bad since, besides (obviously) causing unnecessary server load, it will very quickly result in a version mismatch between the `pdf.js` and `pdf.worker.js` files in those PDF.js deployments.
Such a version mismatch could easily lead to either breaking errors, or even worse slightly inconsistent behaviour for an API call (if the API -> Worker interface changes, which does happen from time to time).
To avoid the problems described above, I'm thus proposing that we enforce that the versions of the `pdf.js` and `pdf.worker.js` files must always match.
The new `PDFSinglePageViewer` class extends the previously created abstract `BaseViewer` class.
There's *a lot* of existing functionality in `PDFViewer` that depends on all the pages being loaded and synchronously available, once the `setDocument` method has been called.
Given that initializing `PDFPageView` instances requires passing a DOM element to which the page is attached, the simplest solution I could come up with is to append all pages to a (hidden) document fragment and just swap them (one at a time) into the viewer when page switching occurs.
This patch introduces an abstract `BaseViewer` class, that the existing `PDFViewer` then extends. *Please note:* This lays the necessary foundation for the next patch.
Looking at `ColorSpace.parseToIR`, it will do one of the following things when called:
1. Return a String.
2. Return an Array.
3. Throw a `FormatError`.
4. In one case, return the result of *another* `ColorSpace.parseToIR` call.
However, under no circumstances will it ever return an `AlternateCS` instance.
Since it's often useful to understand why code, which has become unused, existed in the first place, let's grab a hard hat and a shovel and start digging through the history of this code :-)
The current condition was introduced in commit c198ec4323, in PR 794, but it was actually already obsolete by that time.
The preceeding `instanceof SeparationCS` condition predates commit a7278b7fbc, in PR 700.
That condition was originally introduced all the way back in commit 4e3f87b60c, in PR 692. However, it was made obsolete by commit 9dcefe1efc, which is included in the very same PR!
Hence we're left with the conclusion that not only has this code be unused for *almost* six years, it was basically never used at all save for a few refactoring commits that're part of PR 692.
Bug 1392647 has a PDF where the default width of the font
is 0. It draws some charcodes that don't have glyphs, but
we were wrongly using the 1000 default width for these
charcodes causing some text to be overlapping.
The `DOMParser` is most likely overkill and may be less secure.
Moreover, it is not supported in Node.js environments.
This patch replaces the `DOMParser` with a simple XML parser. This
should be faster and gives us Node.js support for free. The simple XML
parser is a port of the one that existed in the examples folder with a
small regex fix to make the parsing work correctly.
The unit tests are extended for increased test coverage of the metadata
code. The new method `getAll` is provided so the example does not have
to access internal properties of the object anymore.
(for issue #6289)
This does the same for 16 bit as the existing 8 bit tiff predictor code, an addition of the last word to this word.
The last two "& 0xFF" may or may not be needed, I see this isn't done in the 8 bit code, but I'm not a JS developer.
Rather that registering a 'change' event listener on the `window`, which will thus (unnecessarily) fire in *a number* of other situations such as e.g. when the user changes the pageNumber or the current search term, we could/should just register it directly on the dynamically created `fileInput` DOM element instead.
I can see no really compelling reason why we actually need to listen for `file` changes on the `window` itself, and this way we're also able to keep the `fileInput` related code confined to one part of the code which should aid readability.
Furthermore, in custom deployments, there's less risk that we're going to interfere with "outside" code this way.
Finally, preprocessor guards were added to the `webViewerOpenFile` function, since that code doesn't make sense in e.g. the extension builds.
Currently `PDFFunction` is implemented (basically) like a class with only `static` methods. Since it's used directly in a number of different `src/core/` files, attempting to pass in `isEvalSupported` would result in code that's *very* messy, not to mention difficult to maintain (since *every* single `PDFFunction` method call would need to include a `isEvalSupported` argument).
Rather than having to wait for a possible re-factoring of `PDFFunction` that would avoid the above problems by design, it probably makes sense to at least set `isEvalSupported` globally for `PDFFunction`.
Please note that there's one caveat with this solution: If `PDFJS.getDocument` is used to open multiple files simultaneously, with *different* `PDFJS.isEvalSupported` values set before each call, then the last one will always win.
However, that seems like enough of an edge-case that we shouldn't have to worry about it. Besides, since we'll also test that `eval` is actually supported, it should be fine.
Fixes 5573.
Rather than (basically) duplicating the `SimpleLinkService` in `test/driver.js`, with potential test failuires if you forget to update the test mock, it seems much nicer to just re-use the viewer component.
Note that `SimpleLinkService` is already bundled into the `build/components/pdf_viewer.js` file. Hence we only need to expose it similar to the other viewer components in that file, and make sure that the `gulp components` command runs as part of the test-setup.
reference tests
This patch allows us to use the common styles as used by the viewer as a
baseline for the annotation layer reference tests. They are extended
with a small set of overrides to ensure that all elements are visible
during the test.
The overrides file now only contains the absolutely necessary rules to
make all elements visible and is therefore no longer an almost verbatim
copy of the common styles.
This patch provides a new unit tested factory for creating SVG
containers and elements. This code is duplicated twice in the
codebase, but with upcoming changes this would need to be duplicated
even more. Moreover, consolidating this code in one factory allows
us to replace it easily for e.g., supporting Node.js. Therefore, move
this to a central place and update/ES6-ify the related code.
Finally, we replace `setAttributeNS` with `setAttribute` because no
namespace is provided.
This changes both `PDFViewer` and `PDFThumbnailViewer` to return early in the `pagesRotation` setters if the rotation doesn't change.
It also fixes an existing issue, in `PDFViewer`, that would cause errors if the rotation changes *before* the scale has been set to a non-default value.
Finally, in preparation for subsequent patches, it also refactors the rotation code in `web/app.js` to update the thumbnails and trigger rendering with the new `rotationchanging` event.
When testing the new `PDFHistory` implementation in practice, I felt that the current value of `UPDATE_VIEWAREA_TIMEOUT` is too large to be truly useful.
The purpose of the timeout is to attempt to address (the PDF.js part of) https://bugzilla.mozilla.org/show_bug.cgi?id=1153393, and it's currently fairly easy for the user e.g. close the browser before the timeout had a change to finish.
Obviously, the timeout is a best-effort solution, but with the current value of `UPDATE_VIEWAREA_TIMEOUT` it's not as useful as one would want.
Please note that lowering it shouldn't be a problem, since it still prevents the browser history from updating at *every* 'updateviewarea' event or during (quick) scrolling, which is all that's really needed to not impact the UX negatively.
---
Furthermore, with this lower timeout, we can also simplify the part of the 'popstate' event handler that attempted to update the browser history with the current position before moving back. In most cases, the current position will now already exist in the history, and this *greatly* decreases the complexity of this code path.
The main impetus for this change though, is that I unfortunately found that given the asynchronous nature of updating the browser history, there is some *edge* cases where that code could cause history corruption.
In practice, the user could thus get "stuck" at a particular history entry and not be able to move back. I haven't got any reliable STR for this, since it's so difficult to trigger, but it involved navigating around in a document such that a number of destinations are added to the browser history and then changing the rotation before going back/forward in the history.
Rather that attempting to patch this code, and making it even more difficult to understand than it already is or adding more asynchronous behaviour, by far the easiest solution is to remove it and simply rely on the (lowered) `UPDATE_VIEWAREA_TIMEOUT` instead.
Since e.g. zooming can occur when navigating to a new destionation, ensure that a resulting 'updateviewarea' event doesn't trigger adding of a *temporary* position to the browser history at a bad time.
Rather than displaying links that does *nothing* when clicked, it probably makes more sense to simply not render them instead. Especially since it turns out that, at least at this point in time, this is *very* easy to both implement and test.
Fixes 3897.
It seems that the status check, for non-HTTP loads, causes the default viewer to *refuse* to open local PDF files.
***STR:***
1. Make sure that fetch support is enabled in the browser. In Firefox Nightly, set `dom.streams.enabled = true` and `javascript.options.streams = true` in `about:config`.
2. Open https://mozilla.github.io/pdf.js/web/viewer.html.
3. Click on the "Open file" button, and open a new PDF file.
***ER:***
A new PDF file should open in the viewer.
***AR:***
The PDF file fails to open, with an error message of the following format:
`Message: Unexpected server response (200) while retrieving PDF "blob:https://mozilla.github.io/a4fc455f-bc05-45b5-b6aa-2ecff3cb45ce".`
When looking briefly at using `Number.isInteger`/`Number.isNan` rather than `isInt`/`isNaN`, I noticed that there's a couple of not entirely straightforward cases to consider.
At first I really couldn't understand why `parseInt` is being used like it is in `XRef.fetchUncompressed`, since the `num` and `gen` properties of an object reference should *always* be integers.
However, doing a bit of code archaeology pointed to PR 4348, and it thus seem that this was a very deliberate change. Since I didn't want to inadvertently introduce any regressions, I've kept the `parseInt` calls intact but moved them to occur *only* when actually necessary.[1]
Secondly, I noticed that there's a redundant `isCmd` check for an edge-case of broken operators. Since we're throwing a `FormatError` if `obj3` isn't a command, we don't need to repeat that check.
In practice, this patch could perhaps be considered as a micro-optimization, but considering that `XRef.fetchUncompressed` can be called *many* thousand times when loading larger PDF documents these changes at least cannot hurt.
---
[1] I even ran all tests locally, with an added `assert(Number.isInteger(obj1) && Number.isInteger(obj2));` check, and everything passed with flying colours.
However, since it appears that this was in fact necessary at one point, one possible explanation is that the failing test-case(s) have now been replaced by reduced ones.
By using the (heuristic) `POSITION_UPDATED_THRESHOLD` constant, we can ensure that the current document position will be added to the browser history when a sufficiently "large" number of `updateviewarea` events have been dispatched.
This patch attempts to address an issue in the old `PDFHistory` implementation, where the current position wouldn't be correctly saved when the browser was closed.
In theory this *should* already be working, however as the discussion in https://bugzilla.mozilla.org/show_bug.cgi?id=1153393 showed, it seems that both `pagehide` and `beforeunload` arrive to late to successfully update the history during closing.
Hence a timeout is used to *temporarily* add the current position to the browser history when the viewer is idle.
Note that we need to take care not to update the browser history too often, since that would render the viewer more or unusable. Furthermore, if the timeout is *too* long it may end up effectively disable this whole functionality.
The `UPDATE_VIEWAREA_TIMEOUT` constant is thus a heuristic value, which we may need to tweak taking the above into account.
This patch completely re-implements `PDFHistory` to get rid of various bugs currently present, and to hopefully make maintenance slightly easier. Most of the interface is similar to the existing one, but it should be somewhat simplified.
The new implementation should be more robust against failure, compared to the old one. Previously, it was too easy to end up in a state which basically caused the browser history to lock-up, preventing the user from navigating back/forward. (In the new implementation, the browser history should not be updated rather than breaking if things go wrong.)
Given that the code has to deal with various edge-cases, it's still not as simple as I would have liked, but it should now be somewhat easier to deal with.
The main source of complication in the code is actually that we allow the user to change the hash of a already loaded document (we'll no longer try to navigate back-and-forth in this case, since the next commit contains a workaround).
In the new code, there's also *a lot* more comments (perhaps too many?) to attempt to explain the logic. This is something that the old implementation was serverly lacking, which is a one of the reasons why it was so difficult to maintain.
One particular thing to note is that the new code uses the `pagehide` event rather than `beforeunload`, since the latter seems to be a bad idea based on https://bugzilla.mozilla.org/show_bug.cgi?id=1336763.
The current implementation of `PDFHistory` contains a number of smaller bugs, which are *very* difficult to address without breaking other parts of its code.
Possibly the main issue with the current implementation, is that I wrote it quite some time ago, and at the time my understanding of the various edge-cases the code has to deal with was quite limited.
Currently `PDFHistory` may, despite most of those cases being fixed, in certain edge-cases lock-up the browser history, essentially preventing the user from navigating back/forward.
Hence rather than trying to iterate on `PDFHistory` to make it better, the only viable approach is unfortunately rip it out in its entirety and re-write it from scratch.
*It appears that this accidentally broke with PR 8394.*
Currently, the following will be printed in the console:
```
An error occurred while loading the PDF.
[object Promise],[object Promise]
```
With this patch we'll again get proper output, e.g. something with this format:
```
An error occurred while loading the PDF.
PDF.js v? (build: ?)
Message: unknown encryption method
```
By updating to uglify-es, rather than uglify-js, the minifier *itself* now supports ES6 code. This means that it's now possible to minify code built with `PDFJS_NEXT = true` set, i.e. with Babel transpilation disabled, which wasn't the case previously.
Note that uglify-es is based on the API of uglify-js v3, which differs from the one that we previously used.
Of particular importance is the fact that it's no longer possible to provide a path to a file for minification, but one must instead directly provide the source of the file.
For more information, please see https://github.com/mishoo/UglifyJS2/tree/harmony
Since this patch will now treat (some) `NUL` bytes as "ASCII", the number of `followingBytes` checked are thus increased to (hopefully) reduce the risk of introducing new false positives.
Fixes 8823.
According to the specification, see http://www.adobe.com/content/dam/Adobe/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#page=377, a `Dest` entry in an outline item should *not* contain a dictionary.
Unsurprisingly there's PDF generators that completely ignore this, treating is an `A` entry instead.
The patch also adds a little bit more validation code in `Catalog.parseDestDictionary`.
See https://eslint.org/docs/rules/for-direction; helps avoid typos that would cause infinite `for` loops.
Also, updates `eslint` and `eslint-plugin-mozilla` to the latest available versions.
Since the very early days of the viewer, it's been possible to pass in a `scale` when opening a PDF file. However, most of the time it was/is actually being ignored, which limits its usefulness considerably.
In older versions of the viewer, if a document hash was present (i.e. `PDFViewerApplication.initialBookmark` being set) or if the document existed in the `ViewHistory`, the `scale` passed to `PDFViewerApplication.open` would thus always be ignored.
In addition to the above, in the current viewer there's even more cases where the `scale` parameter will be ignored: if a (valid) browser history entry exists on document load, or if the `defaultZoomValue` preference is set to a non-default value.
Hence the result is that in most situation, a `scale` passed to `PDFViewerApplication.open` will be completely ignored.
A much better, not to mention supported, way of setting the initial scale is by using the `defaultZoomLevel` preference. In comparision, this also has the advantage of being used in situations where the `scale` would be ignored.
All in all this leads to the current situation where we have code which is essentially dead, since no part of the viewer (by default) relies on it.
To clean up this code, and to avoid having to pass (basically) unused parameters around, I'd thus like to remove the ability to pass a `scale` to `PDFViewerApplication.open`.
The property and the setter for text rise were already present, but they
were never used or called. This patch completes the implementation by
calling the setter when the operator is encountered and by using the
text rise value when rendering text.
See Bug 1389443 for more information. This allows us to register
PdfJs without waiting for file IO in nsHandlerService to finish.
Once that file IO is finished, we can set everything up properly
and double-check the registration.
The method signature was improved in PR 7440, which has now been present in a number of releases (starting with `v1.6.210`).
Hence we should be able to remove this now, and just print an error message if the old format is used.
This was added in PR 7793, which has now been present in a number of PDF.js releases (from version `v1.7.225`).
Hence we should be able to remove it now, considering that the migration code was only intended as a best effort solution to avoid wiping out all existing user data at once. Also, keep in mind that `ViewHistory` is already limited with regards to the number of documents it will simultaneous store data for.
I've gotten tired of having to manually re-install the `ttest` package on numerous occasions when benchmarking changes, so let's just list it in package.json to simplify things.
As discussed in PR 8673, we cannot solve the general issue (since that would require parsing every single page). However, we can mitigate the effect somewhat, by waiting for the FileAttachment annotations of the initially rendered page.
Rather than duplicating initialization code, we can just call `this.reset()` instead (which is also similar to other existing code, e.g. `PDFViewer`). This will also help ensure that the DOM is completely reset, before any outline items or attachments are displayed.
This bug is similar to the canvas bug of #6721.
I found this bug when I tried to run pdf2svg on a SVG file, and the generated
SVG could not be viewed in Chrome due to a SVG/XML parsing error:
"PCDATA invalid Char value 3"
Reduced test case:
- https://github.com/mozilla/pdf.js/files/1229507/pcdatainvalidchar.pdf
- expected: "hardware performance"
- Actual SVG source: "hardware\x03performance"
(where "\x03" is a non-printable character, and invalid XML).
In terms of rendering, this bug is similar to #6721, where an unexpected glyph
appeared in the canvas renderer. This was fixed by #7023, which skips over
missing glyphs. This commit follows a similar logic.
The test case from #6721 can be used here too:
- https://github.com/mozilla/pdf.js/files/52205/issue6721_reduced.pdf
expected: "Issue 6721"
actual (before this patch): "Issue ààà6721"
Implement a serialization "generator" for `DOMElement` in domutils.js
that yields the serialization of the SVG element. This method is used by
a newly added `ReadableSVGStream` class, which can be used like any
other readable stream in Node.js.
This reduces the memory requirements. Now, it is not needed to require
the serialization to fully fit in memory.
Note: The implementation of the serializer is a state machine in ES5
since the rest of the file is also in ES5. Its functionality is
equivalent to:
```
function* serializeSVGElement(elem) {
yield '<' + elem.nodeName;
if (elem.nodeName === 'svg:svg') {
yield ' xmlns:xlink="http://www.w3.org/1999/xlink"' +
' xmlns:svg="http://www.w3.org/2000/svg"';
}
for (let i in elem.attributes) {
yield ' ' + i + '="' + xmlEncode(elem.attributes[i]) + '"';
}
yield '>';
if (elem.nodeName === 'svg:tspan' || elem.nodeName === 'svg:style') {
yield xmlEncode(elem.textContent);
} else {
for (let childNode of elem.childNodes) {
yield* serializeSVGElement(childNode);
}
}
yield '</' + elem.nodeName + '>';
}
```
Since we're now using `Uint8ClampedArray`, rather than `Uint8Array`, doing manual clamping shouldn't be necessary given that that is now handled natively.
This shouldn't have any measurable performance impact, but just to sanity check that I've done some quick benchmarking with the following manifest file:
```json
[
{ "id": "S2-eq",
"file": "pdfs/S2.pdf",
"md5": "d0b6137846df6e0fe058f234a87fb588",
"rounds": 100,
"type": "eq"
}
]
```
which gave the following results against the current `master` (repeated benchmark runs didn't result in any meaningful differences):
```
-- Grouped By browser, stat --
browser | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ------------ | ----- | ------------ | ----------- | --- | ----- | -------------
firefox | Overall | 100 | 592 | 592 | 1 | 0.12 |
firefox | Page Request | 100 | 3 | 3 | 0 | -9.88 |
firefox | Rendering | 100 | 588 | 589 | 1 | 0.18 |
```
When running browser tests, e.g. via `gulp unittest`, the test files are not
processed by babel, and neither by the "unittestcli" gulp target.
This commit copies the babelPluginReplaceNonWebPackRequire plugin from the
unittestcli target to the SystemJS config so that `__non_webpack_require__` is
replaced with `require` for all build targets, and adds a unit test to ensure
that this indeed works as expected.
This patch removes the `clamp0to255` helper function, as well as manual clamping code in `src/core/jpg.js`.
The adjusted constants in `_convertCmykToRgb` were taken from CMYK to RGB conversion code found in `src/core/colorspace.js`.
*Please note:* There will be some very slight movement in a number of existing test-cases, since `Uint8ClampedArray` appears to use `Math.round` (or equivalent) and the old code used (basically) `Math.floor`.
At this point, the default viewer is already not usable in older browsers in `gulp server` mode, since we only run the code through Babel as part of the build step.
Hence there shouldn't be much point in manually loading `compatibility.js` in `viewer.html` the way that we've been doing, especially considering that it's already being loaded by `src/shared/util.js`.
This appears to simply have been forgotten in the re-factoring in PR 4815, where the `coded` property was renamed to the *much* more descriptive `isType3Font` property.
From looking at blame, it seems that these checks became obsolete with PR 692 (which landed close to six years ago). Note how, after that PR, there's no longer anything being assigned to the `code` property of an Object.
In issue #8707, there's a char code mapped to a non-
existing glyph which shouldn't be drawn. However, we
saw it was missing and tried to then use the post table and
end up mapping it incorrectly.
This illuminated a problem with issue #5704 and bug
893730 where glyphs disappeared after above fix. This was
from the cmap returning the wrong glyph id. Which in turn was
caused because the font had multiple of the same type of cmap
table and we were choosing the last one. Now, we instead
default to the first one. I'm unsure if we should instead be
merging the multiple cmaps, but using only the first one works.
Added unit-tests for DeviceGray, DeviceRGB and DeviceCMYK
Added unit-tests for CalGray
Added unit-tests for CalRGB
Removed redundant code
Added unit-tests for LabCS
Added unit-tests for IndexedCS
Update comment
Change lookup to Uint8Array as mentioned in pdf specs(these tests will pass after PR #8666 is merged).
Added unit-tests for AlternateCS
Resolved code-style issues
Fixed code-style issues
Addressed issues pointed out in https://github.com/mozilla/pdf.js/pull/8611#pullrequestreview-52865469
The initial issue with #8255 was I added a missing glyphs
check to adjustMapping, but this caused us to skip re-mapping
a glyph if the fontCharCode was a missingGlyph which in turn
caused us to overwrite a valid glyph id with an invalid one. While
fixing this, I also added a warning if the private use area is full since
this also accidentally happened when I made a different mistake.
This brought to light a number of issues where we map
missing glyphs to notdef, but often the notdef is actually defined
and then ends up being drawn. Now the glyphs don't get
mapped in toFontChar and so they are not drawn by the canvas.
Fixing the above brought up another issue though in bug1050040.pdf.
In this PDF, the font fails to load by the browser and before we were still
drawing the glyphs because it looked like the font had them, but with the fixes
above the glyphs showed up as missing so we didn't attempt draw them. To
fix this, I now throw an error when the loca table is in really bad shape and
we fall back to trying to use a system font. We now also use this fall back if
there are any format errors during converting fonts.
The PDF file uses a non-embedded SegoeUISymbol font, which is *not* a standard font (and is mainly used by Microsoft, see https://en.wikipedia.org/wiki/Segoe).
Fixes 8697.
Issue 7696 was one of the issues fixed by PR 8580. The other ones were all cases of missing glyphs, however in this particular one glyphs did render but every single one was incorrect.
Hence it probably cannot hurt to have a small, reduced, reference test for that PDF file as well.
This replaces `assert` calls with `throw new FormatError()`/`throw new Error()`.
In a few places, throwing an `Error` (which is what `assert` meant) isn't correct since the enclosing function is supposed to return a `Promise`, hence some cases were changed to `Promise.reject(...)` and similarily for `createPromiseCapability` instances.
Fix TypeError that occurs in colorspace.js on accidentally passing an 'Array' instead of 'TypedArray'
Changed getRgbItem(...) to getRgbBuffer(...) since this.lookup has values in range[0, 255] whereas getRgbItem(...) expects those to be in range [0, 1]
Revert changes for IE9 compatibility
Looking at the blame, it seems that this typo was present even before PR 700 (almost six years ago).
The result of using `'num'`, rather than the *correct* `'numPages'` string, is that the `Catalog.numPages` getter isn't actually being shadowed.
Note that the PageMode, as specified in the API, will only be honoured when either: the user hasn't set the `sidebarViewOnLoad` preference to a non-default value, or a non-default `sidebarView` entry doesn't exist in the view history, or the "pagemode" hash parameter is included in the URL.
Since this is new functionality, the patch also includes a preference (`disablePageMode`), to make it easy to opt-out of this functionality if the user/implementor so wishes.
- Mark the test as async, and don't swallow exceptions.
- Fix the DOMElement polyfill to behave closer to the actual getAttributeNS
method, which excludes the namespace prefix.
Add UI for the cursorToolOnLoad pref in the UI of the Chrome extension.
Add logic to migrate the enableHandToolOnLoad pref to cursorToolOnLoad.
For past values in the mutable extension storage area:
1. If enableHandToolOnLoad=true, save cursorToolOnLoad=1.
2. Remove enableHandToolOnLoad.
For the managed extension storage, which is immutable since it is based
on administrative policies, use the following logic:
1. If enableHandToolOnLoad=true and cursorToolOnLoad=0 (default).
set cursorToolOnLoad=0 and assume enableHandToolOnLoad=false.
2. As usual, managed preferences can (and will) be overridden by the user.
The first migration logic is in extensions/chromium/options/migration.js
and can be removed after a few months / less than many years.
The second migration logic is in web/chromecom.js, and should be kept
around for a long while (many years).
The need for this migration logic arises from the change by:
https://github.com/mozilla/pdf.js/pull/7635
* Check for undefined
new URL(file, window.location.href) throws the following error in IE11 + iPad Safari:
Unable to get property 'replace' of undefined or null reference
* Adapting previous change to pdf.js code standards
Added curly braces
* Moved check for undefined above try/catch
According to the PDF specification, please see http://www.adobe.com/content/dam/Adobe/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#G6.2394361, if an Adobe JPEG marker is present it should always take precedence. This even seem to be consistent with the existing comment that is present in the code.
Hence it seems reasonable to interpret `transformCode === 0` as no color conversion being necessary.
Fixes the rendering of page 1 in `issue-4926` (from the test-suite), when the built-in `src/core/jpg.js` image decoder is used.
Use the environment's zlib implementation if available to get
reasonably-sized SVG files when an XObject image is converted to PNG.
The generated PNG is not optimal because we do not use a PNG predictor.
Futher, when our SVG backend is run in a browser, the generated PNG
images will still be unnecessarily large (though the use of blob:-URLs
when available should reduce the impact on memory usage). If we want to
optimize PNG images in browsers too, we can either try to use a DEFLATE
library such as pako, or re-use our XObject image painting logic in
src/display/canvas.js. This potential improvement is not implemented by
this commit
Tested with:
- Node.js 8.1.3 (uses zlib)
- Node.js 0.11.12 (uses zlib)
- Node.js 0.10.48 (falls back to inferior existing implementation).
- Chrome 59.0.3071.86
- Firefox 54.0
Tests:
Unit test on Node.js:
```
$ gulp lib
$ JASMINE_CONFIG_PATH=test/unit/clitests.json node ./node_modules/.bin/jasmine --filter=SVG
```
Unit test in browser: Run `gulp server` and open
http://localhost:8888/test/unit/unit_test.html?spec=SVGGraphics
To verify that the patch works as desired,
```
$ node examples/node/pdf2svg.js test/pdfs/xobject-image.pdf
$ du -b svgdump/xobject-image-1.svg
# ^ Calculates the file size. Confirm that the size is small
# (784 instead of 80664 bytes).
```
Do not directly export to global. Instead, export all stubs in domstubs.js and
add a method setStubs to assign all exported stubs to a namespace. Then replace
the import domstubs with an explicit call to this setStubs method. Also added
unsetStubs for undoing the changes. This is done to allow unit testing of the
SVG backend without namespace pollution.
Move the DEFLATE logic in convertImgDataToPng to a separate function.
A later commit will introduce a more efficient deflate algorithm,
and fall back to the existing, naive algorithm if needed.
`__pdfjsdev_webpack__` was used to skip evaluating part of an AST,
in order to not mangle some `require` symbols.
This commit removes `__pdfjsdev_webpack__`, and:
- Uses `__non_webpack_require__` when one wants the output to
contain `require` instead of `__webpack_require__`.
- Adds options to the webpack config to prevent "polyfills" for
some Node.js-specific APIs to be added.
- Use `// eslint-disable-next-line no-undef` instead of `/* globals ... */`
for variables that are not meant to be used globally.
Since we no longer, after PR 8555, allow changing the scale until the document is loaded, that hack is no longer necessary. Furthermore, no part of that event handling function needs to run unless a document is loaded.
The reason that this hack was initially added, is that previously the `ViewHistory` might be updated *before* `PDFViewerApplication.setInitialView` had run (in some cases leading to incorrect inital document scale). Since that is no longer possible, this is now dead code.
Don't allow setting various properties, such as `currentPageNumber`/`currentScale`/`currentScaleValue`/`pagesRotation`, before `{PDFViewer, PDFThumbnailViewer}.setDocument` has been called
This is a trivial follow-up to PR 5383, and it's a bit strange that this has been wrong since late 2014 without anyone noticing (maybe because inline images aren't too common).
So, apparently code works better if you actually spell correctly, who knew ;-)
Fixes 8613.
After PR 8394, where the l10n service was converted to be asynchronous, we're no longer calling `_adjustWidth` after updating the `findMsg` label. Hence it's currently possible that the width of the findbar won't be correct. The solution is simple though, just call `_adjustWidth` after the `findMsg` label has been (asynchronously) updated.
Another existing issue, which was an oversight in PR 8132, is that `PDFFindBar.updateResultsCount` may be called directly from `PDFFindController`. In that case, we're not calling `_adjustWidth` at all, which means that the findbar may also not have the correct width.
The simple solution here is to always call `_adjustWidth` at the end of `updateResultsCount` (which is why we no longer need the `_adjustWidth` call at the end of `updateUIState`).
Currently we're *only* hiding the label, but not actually resetting it until a new match is found.
Obviously it's being hidden, but it seems that it really ought to be completely reset as well (since e.g. `PDFFindBar.reset` won't technically reset *all* state otherwise).
Note that these files were among the first to be converted to ES6 classes, so it probably makes sense to do another pass to bring them inline with the most recent ES6 conversions.
All other code-paths already checks that the `MessageHandler` isn't terminated, but apparently `onFailure` was missing that check (compare e.g. with the `onSuccess` function).
From what I can tell, this is only an issue if workers are *disabled*, hence why I didn't bother adding a unit-test.
Fixes 8584.
These changes consists mainly of replacing `var` with `let`/`const`, adding a couple of default parameters to function signatures, and finally converting `EventBus`/`ProgressBar` to proper classes.
After PR 8510, we now always lookup the localized `page_scale_percent` string to prevent any possible ordering issues. Since the scaleSelect dropdown is updated asynchronous, there's really no point in having a helper function any more, hence this code can rather be placed inline in `Toolbar._updateUIState`.
In general, we may not know the stroke properties when path construction
happens. Since we must know the properties when we apply the stroke, we
should set the properties at that point. Note that we already do that
for the color and opacity, but not yet for the other properties.
The https://github.com/mozilla/pdf.js/wiki/Contributing article has been updated to explicitly mention that an ES6 browser is now a minimum requirement for development.
Since we recently have seen a couple of issues filed which seemed to indicate that people tried to use PDF.js in browsers without full ES6 support, it's probably a good idea to mention this more prominently in the README as well.
*This is an existing issue that I noticed while testing PR 8552.*
When zooming or rotation occurs, we'll try to use the current canvas as a (CSS transformed) preview until the page has been completely re-drawn.
If you manage to change the scale (or rotation) *very* quickly, it's possible that `PDFPageView.update` can be called *before* a previous `render` operation has progressed far enough to remove the `hidden` property from the canvas.
The result is thus that a page may be *entirely* black during zooming or rotation, which doesn't look very good. This effect can be a bit difficult to spot, but it does manifest even in the default viewer.
In the PDF from issue 8527, the clip operator (W) shows up before a path
is defined. The current SVG backend however expects a path to exist
before generating a `<svg:clipPath>` element.
In the example, the path was defined after the clip, followed by a
endPath operator (n).
So this commit fixes the bug by moving the path generation logic from
clip to endPath.
Our canvas backend appears to use similar logic:
`CanvasGraphics_endPath` calls `consumePath`, which in turn draws the
clip and resets the `pendingClip` state. The canvas backend calls
`consumePath` from multiple other places, so we probably need to check
whether doing so is also necessary for the SVG backend.
I scanned our corpus of PDF files in test/pdfs, and found that in every
instance (except for one), the "W" PDF operator (clip) is immediately
followed by "n" (endPath). The new test from this commit (clippath.pdf)
starts with "W", followed by a path definition and then "n".
# Commands used to find some of the clipping commands:
grep -ra '^W$' -C7 | less -S
grep -ra '^W ' -C7 | less -S
grep -ra ' W$' -C7 | less -S
test/pdfs/issue6413.pdf is the only file where "W" (a tline 55) is not
followed by "n". In fact, the "W" is the last operation of a series of
XObject painting operations, and removing it does not have any effect
on the rendered PDF (confirmed by looking at the output of PDF.js's
canvas backend, and ImageMagick's convert command).
Part of the rotation handling code, in what's now `web/app.js`, hasn't really changed since before the viewer was split into multiple files/components.
Similar to other properties, such as current page/scale, we should probably avoid tracking state in multiple places. Hence I'm suggesting that we don't store the rotation in `PDFViewerApplication`, and access the value in `PDFViewer` instead.
Since `PDFViewerApplication.pageRotation` has existed for a very long time, a getter was added to avoid outright breaking third-party code that may depend on it.
Currently a number of these properties do not work correctly if set *before* calling `setDocument`; please refer to the discussion starting in https://github.com/mozilla/pdf.js/pull/8539#issuecomment-309706629.
Rather than trying to have *some* of these methods working, but not others, it seems much more consistent to simply always require that `setDocument` has been called.
Wait for the completion of writing the generated SVG file before
processing the next page. This is to enable the garbage collector to
garbage-collect the (potentially large) SVG string before trying to
allocate memory again for the next page.
Note that since the PDF-to-SVG conversion is now sequential instead of
parallel, the time to generate all pages increases.
Test case:
node --max_old_space_size=200 examples/node/pdf2svg.js /tmp/FatalProcessOutOfMemory.pdf
Before this patch:
- Node.js crashes due to OOM after processing 20 pages.
After this patch:
- Node.js is able to convert all 203 PDFs to SVG without crashing.
Test case:
Using the PDF file from https://github.com/mozilla/pdf.js/issues/8534
node --max_old_space_size=200 examples/node/pdf2svg.js /tmp/FatalProcessOutOfMemory.pdf
Before this patch:
Node.js crashes due to OOM after processing 10 pages.
After this patch:
Node.js crashes due to OOM after processing 19 pages.
Since this call occurs *before* the `PDFViewer.setDocument` call, it won't actually cause any scale change.
Furthermore, moving it should not be necessary, since the `scale` is already used as the fallback case in `PDFViewerApplication.setInitialView` (provided it's non-zero, which isn't even the case in the default viewer).
Hence this patch should cause no functional changes at all, since it simply removes a piece of unnecessary code.
This patch adds Streams API support in getTextContent
so that we can stream data in chunks instead of fetching
whole data from worker thread to main thread. This patch
supports Streams API without changing the core functionality
of getTextContent.
Enqueue textContent directly at getTextContent in partialEvaluator.
Adds desiredSize and ready property in streamSink.
Currently, these properties are reset in what appears to be somewhat arbitrary locations (within the `load` and `open` methods respectively). The explanation is probably that both of these properties predates the existence of any centralized clean-up code in the viewer.
Hence I think that it makes sense to move the resetting of these properties to the `close` method, since that improves the overview of what's actually cleaned-up/reset when changing documents in the viewer.
The `ObjectLoader` currently takes an Object as input, despite actually working with `Dict`s internally. This means that at the (two) existing call-sites, we're passing in the "private" `Dict.map` property directly.
Doing this seems like an anti-pattern, and we could (and even should) simply provide the actual `Dict` when creating an `ObjectLoader` instance.
Accessing properties stored in the `Dict` is now done using the intended methods instead, in particular `getRaw` which (as the name suggests) doesn't do any de-referencing, thus maintaining the current functionality of the code.
The only functional change in this patch is that `ObjectLoader.load` will now ignore empty nodes, such that `ObjectLoader._walk` only needs to deal with nodes that are known to contain data. (This lets us skip, among other checks, meaningless `addChildren` function calls.)
This method is currently called from `PDFViewer._scrollUpdate` on *every* scroll event in the viewer.
However, I cannot see why this code is now necessary (assuming that it once was), since text-selection and searching still works *exactly* the same way with this patch as with the current `master`.
When `PDFPageView.updatePosition` is called, the page can be in either of these states:
1. The page hasn't been rendered, in which case the `textLayer` property doesn't exist yet.
2. The page is currently rendering, meaning that the `textLayer` property exists. Given that the `textContent` won't be fetched until the page has been successfully rendered, `TextLayerBuilder.render` will return immediately and effectively be a no-op (since there's nothing to render yet).
3. The has been been rendered, and the `textLayer` is currently rendering.
4. The page, and its `textLayer`, has been completely rendered. In this case, `TextLayerBuilder.render` will return immediately and effectively be a no-op.
Here, only the *third* case seem to require any further analysis:
When scrolling occurs while the `textLayer` is rendering, `TextLayerBuilder.render` will via a helper method call `TextLayerRenderTask.cancel` (in src/display/text_layer.js) to stop processing.
However, due to the run-to-completion nature of JavaScript, once `TextLayerRenderTask._render` has been invoked `appendText` will always run.[1]
So even though we cancel rendering of pending `textLayer`s during scrolling, via the repeated `TextLayerBuilder.render` calls from within the `PDFPageView.updatePosition` method, that does *not* prevent us from running the code inside of `TextLayerRenderTask._render` over and over for the *same* page; which all seems *very* inefficient to me.[2]
All this will thus have the effect of delaying the *actual* rendering of a `textLayer` ever so slightly while scrolling in the viewer. However, it does so at the expense of potentially hundreds of unnecessary `appendText` calls.[3]
Hence it seems to me that it's less resource intensive overall to simply let rendering of the `textLayer` complete, once it has started. Obviously, we still abort all rendering of a page, and its `textLayer`, when it's being destroyed (e.g. by being evicted from the page cache).
In case that there's any worry that the patch could affect e.g. highlighting of search results, please note that the existing code in `TextLayerBuilder.render` already calls `updateMatches` when the `TextLayerTask` resolves successfully.
*I'm sorry that this became quite long, but to try and summarize:*
`PDFPageView.updatePosition` doesn't actually do anything in *most* cases. In the one case where it matters, it seems that it's actually doing more harm than good; which is why I'm proposing that we just remove it.
---
[1] Although we may be able to skip the `render` call, provided that it happens *after* a `timeout` (as is the case in the default viewer).
[2] With current work being done to support streaming of `TextContent`, we'd also need to add just as many duplicate API calls to `PDFPageView.updatePosition`.
[3] The number of duplicate `appendText` calls is directly proportional not only to the scroll speed, but also to the number of pages in the document.
Since the localization service is now asynchronous, depending on the load the browser is under, there's a small risk that the lookup of the 'page_scale_percent' string could be delayed slightly.
If the scale would change a couple of times in *very* quick succession, there's perhaps a *theoretical* possibility that the Zoom dropdown would display an incorrect value.
Consider the following, somewhat contrived, theoretical example of two zoom commands being executed *right* after one another:
```javascript
PDFViewerApplication.pdfViewer.currentScale = 1.23;
PDFViewerApplication.pdfViewer.currentScaleValue = 'page-width';
```
Only the `currentScale` call will currently trigger a l10n lookup in `selectScaleOption`. However, as far as I understand, there's no *guarantee* that the l10n string is resolved *before* `selectScaleOption` is called again as a result of the `currentScaleValue` call.
This thus has the possibility of putting the Zoom dropdown into an inconsistent state, since it's currently updated synchronously for one code-path and asynchronously for another.
To avoid these issues, I'm proposing that we *always* update the Zoom dropdown asynchronously, such that we can guarantee that the ordering is correct.
After PR 8394, at least in Firefox, the Zoom dropdown now frequently displays an old custom scale instead of the correct one. To see this behaviour, the following STR works for me:
1. Open https://mozilla.github.io/pdf.js/web/viewer.html.
2. Zoom in, by clicking on the corresponding button in the toolbar.
3. Run `PDFViewerApplication.pdfViewer.currentScaleValue` in the console.
4. Compare what's displayed in the Zoom dropdown with what's printed in the console. (If no difference can be observed, try repeating steps 2 through 4 a couple of times.)
I really don't understand why this happens, but it seems that waiting until a custom scale has been set *before* selecting it fixes things in Firefox (and works fine in e.g. Chrome as well).
Note that this patch thus makes this particular piece of the code consistent with the state prior to PR 8394.
*As mentioned the last time that I touched this particular part of the font code, I'm sincerely hope that this doesn't cause any regressions!*
However, the patch passes all tests added in PRs 5770, 6270, and 7904 (and obviously all other tests as well). Furthermore, I've manually checked all the issues/bugs referenced in those PRs without finding any issues.
Fixes 8480.
Adds functionality to accept Queueing Strategy in
sendWithStream method. Using Queueing Strategy we
can control the data that is enqueued into the sink,
and hence regulated the flow of chunks from worker
to main thread.
Adds capability in pull and cancel methods.
Adds ready and desiredSize property in streamSink.
Adds unit test for ReadableStream and sendWithStream.
The click handler used in Presentation Mode didn't check if the first/last page was already reached, which after PR 7529 now causes an unnecessary console error.
Hence we should simply use the already existing `_goToPreviousPage`/`_goToNextPage` methods instead, since they do the necessary bounds checking.
Fixes 8498.
Moreover, rename `FindStates` to `FindState` since enumeration names are
usually not in plural, for readability and consistency with the ones in
`src/shared/util.js`.
http://eslint.org/docs/rules/comma-danglehttp://eslint.org/docs/rules/object-curly-spacing
*Please note:* This patch was created automatically, using the ESLint `--fix` command line option. In a couple of places this caused lines to become too long, and I've fixed those manually; please refer to the interdiff below for the only hand-edits in this patch.
```diff
diff --git a/gulpfile.js b/gulpfile.js
index d18b9c58..7d47fd8d 100644
--- a/gulpfile.js
+++ b/gulpfile.js
@@ -1247,7 +1247,8 @@ gulp.task('gh-pages-git', ['gh-pages-prepare', 'wintersmith'], function () {
var reason = process.env['PDFJS_UPDATE_REASON'];
safeSpawnSync('git', ['init'], { cwd: GH_PAGES_DIR, });
- safeSpawnSync('git', ['remote', 'add', 'origin', REPO], { cwd: GH_PAGES_DIR, });
+ safeSpawnSync('git', ['remote', 'add', 'origin', REPO],
+ { cwd: GH_PAGES_DIR, });
safeSpawnSync('git', ['add', '-A'], { cwd: GH_PAGES_DIR, });
safeSpawnSync('git', [
'commit', '-am', 'gh-pages site created via gulpfile.js script',
```
Fix inconsistent spacing and trailing commas in objects in `src/core/` files, so we can enable the `comma-dangle` and `object-curly-spacing` ESLint rules later on
Fix inconsistent spacing and trailing commas in objects in remaining `src/` files, so we can enable the `comma-dangle` and `object-curly-spacing` ESLint rules later on
Fix inconsistent spacing and trailing commas in objects in `test/` files, so we can enable the `comma-dangle` and `object-curly-spacing` ESLint rules later on
http://eslint.org/docs/rules/comma-danglehttp://eslint.org/docs/rules/object-curly-spacing
Given that we currently have quite inconsistent object formatting, fixing this in *one* big patch probably wouldn't be feasible (since I cannot imagine anyone wanting to review that); hence I've opted to try and do this piecewise instead.
Please note: This patch was created automatically, using the ESLint `--fix` command line option. In a couple of places this caused lines to become too long, and I've fixed those manually; please refer to the interdiff below for the only hand-edits in this patch.
```diff
diff --git a/src/display/canvas.js b/src/display/canvas.js
index 5739f6f2..4216b2d2 100644
--- a/src/display/canvas.js
+++ b/src/display/canvas.js
@@ -2071,7 +2071,7 @@ var CanvasGraphics = (function CanvasGraphicsClosure() {
var map = [];
for (var i = 0, ii = positions.length; i < ii; i += 2) {
map.push({ transform: [scaleX, 0, 0, scaleY, positions[i],
- positions[i + 1]], x: 0, y: 0, w: width, h: height, });
+ positions[i + 1]], x: 0, y: 0, w: width, h: height, });
}
this.paintInlineImageXObjectGroup(imgData, map);
},
diff --git a/src/display/svg.js b/src/display/svg.js
index 9eb05dfa..2aa21482 100644
--- a/src/display/svg.js
+++ b/src/display/svg.js
@@ -458,7 +458,11 @@ SVGGraphics = (function SVGGraphicsClosure() {
for (var x = 0; x < fnArrayLen; x++) {
var fnId = fnArray[x];
- opList.push({ 'fnId': fnId, 'fn': REVOPS[fnId], 'args': argsArray[x], });
+ opList.push({
+ 'fnId': fnId,
+ 'fn': REVOPS[fnId],
+ 'args': argsArray[x],
+ });
}
return opListToTree(opList);
},
```
Fix inconsistent spacing and trailing commas in objects in `web/` files, so we can enable the `comma-dangle` and `object-curly-spacing` ESLint rules later on
Fix inconsistent spacing and trailing commas in objects in `extensions/` files, so we can enable the `comma-dangle` and `object-curly-spacing` ESLint rules later on
http://eslint.org/docs/rules/comma-danglehttp://eslint.org/docs/rules/object-curly-spacing
Given that we currently have quite inconsistent object formatting, fixing this in in one big patch probably wouldn't be feasible (since I cannot imagine anyone wanting to review that); hence I've opted to try and do this piecewise instead.
Please note: This patch was created automatically, using the ESLint `--fix` command line option. In a couple of places this caused lines to become too long, and I've fixed those manually; please refer to the interdiff below for the only hand-edits in this patch.
```diff
diff --git a/extensions/firefox/content/PdfStreamConverter.jsm b/extensions/firefox/content/PdfStreamConverter.jsm
index ea91a71a..0d59dad1 100644
--- a/extensions/firefox/content/PdfStreamConverter.jsm
+++ b/extensions/firefox/content/PdfStreamConverter.jsm
@@ -773,7 +773,8 @@ class RequestListener {
response = function sendResponse(aResponse) {
try {
var listener = doc.createEvent("CustomEvent");
- let detail = Cu.cloneInto({ response: aResponse, }, doc.defaultView);
+ let detail = Cu.cloneInto({ response: aResponse, },
+ doc.defaultView);
listener.initCustomEvent("pdf.js.response", true, false, detail);
return message.dispatchEvent(listener);
} catch (e) {
```
After PR 8459, the run-time of the various `gulp` test commands has regressed quite badly on Windows. For me, `gulp test` now takes approximately *twice* as long when run locally on Windows.
The problem seems to be the Babel transpilation step, which takes well over five minutes to run.[1]
For someone like me, who runs tests a lot locally, this slowdown is really hurting the overall development experience.
To get around this I tested setting `PDFJS_NEXT = true` in `gulpfile.js`, since the transpilation step isn't necessary when testing in a modern browser.
However, having to edit `gulpfile.js` every time that I need to run tests isn't very practical. Hence this patch, which adds an environment variable that allows you to disable the transpilation simply by using e.g. `PDFJS_NEXT=true gulp test`.
I hope that this can be considered an acceptable solution, such that I don't need to maintain this patch locally (or worse, edit `gulpfile.js` locally before testing).
---
[1] This can also be observed on the Windows bot, but it seems fine on Linux.
This patch intends to simplify future ES6 refactoring of the thumbnail code, since the current code isn't going to work well together with proper `Class`es.
Please note that the `glyphlist.js` and `unicode.js` files are converted to CommonJS modules instead, since Babel cannot handle files that large and they are thus excluded from transpilation.
Add different code-paths to `{CMap, ToUnicodeMap}.charCodeOf` depending on length, since `Array.prototype.indexOf` can be extremely inefficient for very large arrays (issue 8372)
With the current way that the `HandTool` is implemented, if someone would try to also add a Zoom tool (as issue 1260 asks for) that probably wouldn't work very well given that you'd then have two cursor tools which may not play nice together.
Hence this patch, which attempts to refactor things so that it should be simpler to add e.g. a Zoom tool as well (given that that issue is marked as "good-beginner-bug", and I'm not sure if that really applies considering the current state of the code).
Note that I personally have no interest in implementing a Zoom tool (similar to Adobe Reader) since I wouldn't use it, but I figured that it can't hurt to make this code a bit more future proof.
PR 7341 added special handling for `nameddest`s that look like pageNumbers, to prevent issues since we previously *incorrectly* supported specifying a pageNumber directly in the hash; i.e. `#10` versus the correct `#page=10` format.
Since this behaviour wasn't correct, PR 7757 fixed and deprecated the old format, which means that we no longer need to maintain the `nameddest` hack in multiple files.
Also updates a `console.log` statement in the `gulp importl10n` command (since I forgot it in a previous patch, and it didn't seem necessary with a separate patch for it).
Added test for ReadableStream.
Adds ref-implementation license-header in streams-lib
and change gulp task to copy external/streams/ in build/
external/streams/ and build/dist/external/streams folder.
Adds README.md and LICENSE.md
This patch replaces a `var self = this;` statement with arrow functions, and `var` with `let` in `PDFLinkService.navigateTo`.
Furthermore, when I started looking at this method, it quickly became clear that its code is somewhat of a mess. Since I'm one of the persons that have touched this code over the years, I figured that it'd be a good idea to try and clean it up a bit.
Also replaces `var` with `let` in code that's touched in the patch. Please note that this should be completely safe, for two separate reasons, since trying to access let in a scope where it's not defined is first of all a runtime error and second of all an ESLint error (thanks to the `no-undef` rule).
Refactors `Driver._cleanup` to return a `Promise` which is resolved once all opened documents have been destroyed.
This is then used in `Driver._nextTask` to ensure that we wait for everything to be cleaned up, such that the tests run sequentially.
Currently this method first uses a loop to build a temporary array to hold Promises, which are then resolved from a recursive helper function once the textContent is fetched for each page.
To me, this is unncessarily complicated, since we can do everything within one loop by simply chaining the asynchronous calls to retrieve the textContent. (Note that this guarantees that the textContent of the pages is still fetched sequentially.)
Also replaces `var` with `let` in the functions/methods that are touched in the patch. Please note that this should be completely safe, for two separate reasons, since trying to access `let` in a scope where it's not defined is first of all a runtime error and second of all an ESLint error (thanks to the `no-undef` rule).
This is a downstream change introduced in [1]. That mozilla bug is adding a new property to channel's loadinfo object (nsILoadInfo) that protocol handlers has to set on channels when originalURI on the result channel is set to a different URI than the channel has been created for.
Existence of the new property on nsILoadInfo depends on landing [1].
[1] https://bugzilla.mozilla.org/show_bug.cgi?id=1319111
- the viewer was not loading in development mode on Safari, due to Safari
having crypto.webkitSubtle instead of crypto.subtle.
- the isCachingPossible check was amended to check for crypro.subtle
which is currently not in Safari but in Firefox and Chrome. This
essentially works around the issue by disabling caching for Safari
in development mode.
- maintainer sentiment: people who develop on Safari can get this speedup
once Safari drops prefix for SubtleCrypto.
- note: at time of writing Safari Version 10.1 (12603.1.30.0.34) has an
issue where caching can be enabled for PDF.js but must to be disabled for
worker, otherwise the two sides do not communicate.
- https://github.com/mozilla/pdf.js/pull/8387#issuecomment-299709961
- https://developer.mozilla.org/en-US/docs/Web/API/SubtleCrypto/digest
For some reason, we're putting all kind of images *except* JPEG into the `imageCache` in `evaluator.js`.[1]
This means that in the PDF file in issue 8380, we'll keep sending the *same* two small images[2] to the main-thread and decoding them over and over. This is obviously hugely inefficient!
As can be seen from the discussion in the issue, the performance becomes *extremely* bad if the user has the addon "Adblock Plus" installed. However, even in a clean Firefox profile, the performance isn't that great.
This patch not only addresses the performance implications of the "Adblock Plus" addon together with that particular PDF file, but it *also* improves the rendering times considerably for *all* users.
Locally, with a clean profile, the rendering times are reduced from `~2000 ms` to `~500 ms` for my setup!
Obviously, the general structure of the PDF file and its operator sequence is still hugely inefficient, however I'd say that the performance with this patch is good enough to consider the issue (as it stands) resolved.[3]
Fixes 8380.
---
[1] Not technically true, since inline images are cached from `parser.js`, but whatever :-)
[2] The two JPEG images have dimensions 1x2, respectively 4x2.
[3] To make this even more efficient, a new state would have to be added to the `QueueOptimizer`. Given that PDF files this stupid fortunately aren't too common, I'm not convinced that it's worth doing.
This patch contains the following improvements:
- Only fetch the various document properties *once* per PDF file opened, and cache the result (in a frozen object).
- Always update the *entire* dialog at once, to prevent inconsistent UI state (issue 8371).
- Ensure that the dialog, and all its internal properties, are reset when `PDFViewerApplication.close` is called.
- Inline, and re-factor, the `getProperties` method in `open`, since that's the only call-site.
- Always overwrite the fileSize with the value obtained from `pdfDocument.getDownloadInfo`, to ensure that it's correct.
- ES6-ify the code that's touched in this patch.
Fixes 8371.
Currently these methods accept a large number of parameters, which creates quite unwieldy call-sites. When invoking them, you have to remember not only what arguments to supply, but also the correct order, to avoid runtime errors.
Furthermore, since some of the parameters are optional, you also have to remember to pass e.g. `null` or `undefined` for those ones.
Also, adding new parameters to these methods (which happens occasionally), often becomes unnecessarily tedious (based on personal experience).
Please note that I do *not* think that we need/should convert *every* single method in `evaluator.js` (or elsewhere in `/core` files) to take parameter objects. However, in my opinion, once a method starts relying on approximately five parameter (or even more), passing them in individually becomes quite cumbersome.
With these changes, I obviously needed to update the `evaluator_spec.js` unit-tests. The main change there, except the new method signatures[1], is that it's now re-using *one* `PartialEvalutor` instance, since I couldn't see any compelling reason for creating a new one in every single test.
*Note:* If this patch is accepted, my intention is to (time permitting) see if it makes sense to convert additional methods in `evaluator.js` (and other `/core` files) in a similar fashion, but I figured that it'd be a good idea to limit the initial scope somewhat.
---
[1] A fun fact here, note how the `PartialEvaluator` signature used in `evaluator_spec.js` wasn't even correct in the current `master`.
Note that we have to use `fs.writeFileSync` since `.to()` is not
available anymore. Moreover, introduce `safeSpawnSync` to make sure that
we check the return codes of the spawned processes properly.
To run the regression tests, developers use `gulp browsertest` and the
bot uses `gulp bottest`. We're not passing the `noreftest` option
anywhere in the code (probably because the `bottest` command takes care
of this already), so we should remove this.
Note that by using `let` instead of `var` in `PartialEvaluator.setGState` and `TranslatedFont.loadType3Data`, we can get rid of further `bind` usages since `let` is block-scoped.
Also, the fact that `bind` wasn't used in the `Font` case inside of `setGState` is actually a bug which has been present ever since PR 5205, where a closure was replaced by a standard loop.[1]
---
[1] I'm not aware of any bugs caused by this, but that is probably more a happy accident than anything else, since e.g. just removing the `bind` from the `SMask` case without using block-scoped variables causes test failures.
Please note that I used the addon debugger to set a breakpoint in the `unload` function, in order to ensure that `this` still correctly refers to the `FindEventManager` scope.
Please see http://eslint.org/docs/rules/object-shorthand.
Unfortunately, based on commit 9276d1dcd9, it seems that we still need to maintain compatibility with old Node.js versions, hence certain files/directories that are executed in Node.js are currently exempt from this rule.
Furthermore, since the files specific to the Chromium extension are not run through Babel, the `/extensions/chromium/` directory is also exempt from this rule.
Please see http://eslint.org/docs/rules/object-shorthand.
For the most part, these changes are of the search-and-replace kind, and the previously enabled `no-undef` rule should complement the tests in helping ensure that no stupid errors crept into to the patch.
This is a regression from commit 3888a993b1.
It turns out the even though we have a `URL` polyfill, it's still dependent on the existence of native `URL.{createObjectURL, revokeObjectURL}` functions.
Since no such thing exists in Node.js, our `createObjectURL` utility function breaks there.
The test runner is automated, so if the default browser test is
performed, the browser hangs waiting for user input it never gets.
Disable the test to fix that.
Moreover, enable E10s now that it is mature. This may help with the
performance of the test runner as well.
Please see http://eslint.org/docs/rules/object-shorthand.
For the most part, these changes are of the search-and-replace kind, and the previously enabled `no-undef` rule should complement the tests in helping ensure that no stupid errors crept into to the patch.
Rather than having to manually use `initializedPromise`, which really ought to be a private property, to ensure that setting/getting values in the `ViewHistory` works as intended, this re-factoring simply changes all of its methods to be asynchronous.
Furthermore, a `getMultiple` method (mirroring the existing `setMultiple` one) is also added to `ViewHistory`.
Finally, this patch also addresses an existing issue, where certain preferences (e.g. the default zoom level) would be ignored when calling `setInitialView` if reading from the `ViewHistory` fails for some reason.
Please refer to the JBIG2 standard, see https://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-REC-T.88-200002-I!!PDF-E&type=items.
In particular, section "6.3.5.3 Fixed templates and adaptive templates" mentions that the offsets should be *subtracted*; where the offsets are defined according to "Table 6" under section "6.3.2 Input parameters".
Fixes 7145.
Fixes 7308.
Fixes 7401.
Fixes 7850.
Fixes 8270.
Please see http://eslint.org/docs/rules/object-shorthand.
For the most part, these changes are of the search-and-replace kind, and the previously enabled `no-undef` rule should complement the tests in helping ensure that no stupid errors crept into to the patch.
Please see http://eslint.org/docs/rules/object-shorthand.
For the most part, these changes are of the search-and-replace kind, and the previously enabled `no-undef` rule should complement the tests in helping ensure that no stupid errors crept into to the patch.
With the exception of just one test-case, all the current `ui_utils` unit-tests can run successfully on Node.js (since most of them doesn't rely on the DOM).
To get this working, I had to first of all add a new `LIB` build flag such that `gulp lib` produces a `web/pdfjs.js` file that is able to load `pdf.js` successfully.
Second of all, since neither `document` nor `navigator` is available in Node.js, `web/ui_utils.js` was adjusted slightly to avoid errors.
[Firefox addon] Remove the `window.FirefoxCom` hack from `web/viewer.js`, since it was made redunant by the `setExternalLocalizerServices` refactoring (PR 7202 follow-up)
This prevents issues with the filename detection being skipped, when trying to download the opened PDF attachment, since `getPDFFileNameFromURL` ignores `data:` URLs for performance reasons.
The patch also changes the `defaultFilename` to use the ES6 default parameter notation, and fixes the formatting of the JSDoc comment.
Finally, since `getPDFFileNameFromURL` currently has no unit-tests, a few basic ones are added to avoid regressions.
[Firefox addon] Upstream changes from: Bug 1346616 - Migrate callsites that are retrieving requested locale from pref, to use LocaleService::GetRequestedLocales
In https://bugzilla.mozilla.org/show_bug.cgi?id=1355216, the *third* parameter of `Services.obs.addObserver` was made optional.
However, omitting it in Firefox versions *without* that patch causes failures that completely prevents the addon from working (it won't even load).
As far as I can tell, there isn't *any* way to detect ahead of time if the third parameter can be safely omitted, hence we're forced to fallback to manually checking the version number :-(
*Note:* Since the `PdfJs.jsm` file is only used in the `MOZCENTRAL` build, we at least don't need to add any compatibility hacks there.
In the first commit in PR 8203, I changed how the `DownloadManager` was included/initialized in `GENERIC`/`CHROME` builds.
The change was prompted by the fact that you cannot have conditional `import`s with ES6 modules, and I wanted to avoid bundling the general `DownloadManager` into the various Firefox specific build targets.
What I completely missed though, is that the new code meant that `download_manager.js` will now be pulling in the *entire* viewer (through `app.js`).
This is a *really* stupid mistake on my part, since it causes the `dist/build/pdf_viewer.js` used with the viewer components to now include basically the entire default viewer.
The simplest solution that I could come up with, is to add a `genericcom.js` file (similar to the `firefoxcom.js`/`chromecom.js` files) which will be responsible for importing/initializing the `DownloadManager`.
The browsers have become smarter and made this hack no longer
functional. Since this is now enforced by practically all browsers,
there is nothing more we can do about this. It is up to the user to
serve the viewer over HTTPS or deal with the warning.
Note that this is in no way specific for PDF.js. Any site with password
inputs served over HTTP has this problem right now. This hack was
provided as a convenience for the users, but nothing more than that.
Since calling `PDFPageView.setPdfPage` will in turn call `PDFPageView.reset`, which cancels all rendering and completely resets the page, it's thus possible that we currently cause some unnecessary re-rendering during the initial loading phase of the viewer.
Depending on the order in which data arrives, it's possible (and in practice always seem to happen) that the `pdfPage` property of the *second* page has already been set during `PDFViewer.setDocument`, by the time that the request for the `pdfPage` is resolved in `PDFViewer._ensurePdfPageLoaded`.
Also, note how the `setPdfPage` call in `PDFViewer.setDocument` is already guarded by this kind of check.
In various viewer files, there's a number of cases where we basically duplicate the functionality of `createPromiseCapability` manually.
As far as I can tell, a couple of these cases have existed for a very long time, and notable even before the `createPromiseCapability` utility function existed.
Also, since we can write ES6 code now, the patch also replaces a couple of `bind` usages with arrow functions in code that's touched in the patch.
[api-minor] Always allow e.g. rendering to continue even if there are errors, and add a `stopAtErrors` parameter to `getDocument` to opt-out of this behaviour (issue 6342, issue 3795, bug 1130815)
This patch implements support for line annotations. Other viewers only
show the popup annotation when hovering over the line, which may have
any orientation. To make this possible, we render an invisible line (SVG
element) over the line on the canvas that acts as the trigger for the
popup annotation. This invisible line has the same starting coordinates,
ending coordinates and width of the line on the canvas.
Other PDF readers, e.g. Adobe Reader and PDFium (in Chrome), will attempt to render as much of a page as possible even if there are errors present.
Currently we just bail as soon the first error is hit, which means that we'll usually not render anything in these cases and just display a blank page instead.
NOTE: This patch changes the default behaviour of the PDF.js API to always attempt to recover as much data as possible, even when encountering errors during e.g. `getOperatorList`/`getTextContent`, which thus improve our handling of corrupt PDF files and allow the default viewer to handle errors slightly more gracefully.
In the event that an API consumer wishes to use the old behaviour, where we stop parsing as soon as an error is encountered, the `stopAtErrors` parameter can be set at `getDocument`.
Fixes, inasmuch it's possible since the PDF files are corrupt, e.g. issue 6342, issue 3795, and [bug 1130815](https://bugzilla.mozilla.org/show_bug.cgi?id=1130815) (and probably others too).
- When the minified version is used the resolver of the worker can not find it properly and throws 404 error.
- The problem was that:
- It was getting the current name of the file.
- It was replacing **.js** by **.worker.js**
- When it was loading the unminified version it was working fine because:
- *pdf.js - .js + .worker.js* = **pdf.worker.js**
- When it was loading the minified version it didtn't work because:
- *pdf.min.js - .js + .worker.js* = **pdf.min.worker.js**
- **pdf.min.worker.js** doesn't exist the real file name is **pdf.worker.min.js**
The display layer is responsible for creating the HTML elements for the
annotations from the core layer. If we need to ignore border styling for
the containers of certain elements, the display layer should do so and
not the core layer. I noticed this during the implementation of line
annotations, for which we actually need the original border width in the
display layer, even though we ignore it for the container. If we set the
border style to zero in the core layer, this becomes impossible.
To prevent this, this patch moves the container border removal code from
the core layer to the display layer. This makes the core layer output
the unchanged annotation data and lets the display layer remove any
border styling if necessary.
Note that as discussed on IRC, this makes the viewer slightly slower to load *only* in `gulp server` mode, however the difference seem slight enough that I think it will be fine.
Considering how extremely simple this patch turned out to be, I'm almost worried that I completely misunderstood why the current code looks like it does...
It doesn't really make sense to attempt to utilize the `NativeImageDecoder` in Node, since there's no native image support available, hence building on PR 8035 we can easily disable it in the example.
Fixes 7901.
The existing implementation of fakeRequestAnimationFrame
did not return a timer ID, so the frame could not be cancelled
if you wanted to cancel it. But if you do want to cancel it,
it needs to be cancelled through clearTimeout instead of
cancelAnimationFrame, because the timer IDs are different.
Signed-off-by: Jonathan Barnes <jbarnes@pivotal.io>
If we only invoke the bootstrap-enabled script when PdfJs.enabled is
true, then we don't need to check it again in the script.
This avoids a sync IPC call to the parent process.
It also keeps PdfJs.jsm from importing PdfjsContentUtils.jsm in the
child process until it is actually needed, which is one steps towards
not loading it until it is really needed.
pdfjschildbootstrap.js will always be run, but
pdfjschildbootstrap-enabled.js will only be run if PdfJs.enabled is
true. This will let us avoid some work in the child process in the
next patch.
This will need to be landed in the mozilla-central repository at the
same time as a change to nsBrowserGlue.js. See bug 1352218.
Refactor removing of the `zoomLayer` into a helper method, and use that in `PDFPageView.reset` to ensure that the entire `zoomLayer` is actually removed (issue 8209)
I happened to notice that the error handling wasn't that great, which I missed previously since there were no unit-tests for failure to load built-in CMap files.
Hence this patch, which improves the error handling *and* adds tests.
I found that PR 8105 unfortunately causes a *very serious* performance regression in long PDF documents where the `Pages` tree only has one level; my apologies for this!
Obviously we cannot revert that PR, since that would cause more issues than it solves. Hence it seems to me that the only viable solution here, is to add a simple `RefSetCache` to reduce the amount of redundant lookups.
Previously in PR 8105 caching was thought to be unnecessary, but as it turns out I don't think that we really have a choice in the matter any more.
See http://eslint.org/docs/rules/#ecmascript-6.
To try and enforce consistent rules and to help avoid some possible errors in ES6 code from the start, this patch adds a few basic ESLint rules.
Note that a two of the rules, `no-shadow` and `object-shorthand`, are currently disabled. While it'd certainly be nice to enable both of them, it's currently impossible since that would result in close to one thousand lint errors.
Remove unnecessary `xref` parameters from various method signatures in `PartialEvaluator`, since `this.xref` is already available in the relevant scope
For reasons I don't pretend to understand, we're passing around `xref` arguments to a bunch of methods despite `this.xref` being available in `PartialEvaluator`.
This patch is a small first small step towards cleaning up the, often unwieldy, signatures of methods in `PartialEvaluator`.
I really cannot understand why this change is necessary, since modern browsers such as Firefox and Chrome work just fine with the old code.
Hence this is patch is yet another "hack" that's needed just because IE apparently cannot just work like you'd expect.
For consistency, the Node factory used in the CMap unit-tests is changed as well.
Fixes 8193.
*My apologies for inadvertently breaking this in PR 8064; apparently we don't have any tests that cover this use-case :(*
Without this patch `getTextContent` will fail if called before `getOperatorList`, since loading of fonts during text-extraction may require fetching of built-in CMap files.
*Please note:* The `text` test added here, which uses an already existing PDF file, fails without this patch.
In core/document.js: `PDFDocument.prototype.parse` accesses a dictionary
property, which could throw if the underlying data is not yet available.
In core/obj.js: `get Catalog.prototype.metadata` calls
`stream.getBytes`, which can throw MissingDataException too when the
stream is a ChunkedStream.
Similar to other `try-catch` statements in `/core` code, we must re-throw `MissingDataException` to prevent issues with missing data during document loading.
Note that I'm not sure if/how we can test this, which is why the patch doesn't include any test(s).
Fixes 8180.
*After browsing through (a version of) the JPEG specification, see https://www.w3.org/Graphics/JPEG/itu-t81.pdf, I hope that this patch makes sense.*
Note that while issue 7828 became a problem after PR 7661, it isn't really a regression from than PR. The explanation is rather that we're now relying on `core/jpg.js` instead of the Native Image decoder in more situations than before, which thus exposed an *existing* issue in our JPEG decoder.
Another factor also seems to be that in many JPEG images, the DRI (Define Restart Interval) marker isn't present, in which case this bug won't manifest either.
According to https://www.w3.org/Graphics/JPEG/itu-t81.pdf#page=89 (at the bottom of the page):
"NOTE – The final restart interval may be smaller than the size specified by the DRI marker segment, as it includes only the number of MCUs remaining in the scan."
Furthermore, according to https://www.w3.org/Graphics/JPEG/itu-t81.pdf#page=39 (in the middle of the page):
"[...] If restart is enabled and the restart interval is defined to be Ri, each entropy-coded segment except the last one shall contain Ri MCUs. The last one shall contain whatever number of MCUs completes the scan."
Based on the above, it thus seem to me that we should simply ensure that we're not attempting to continue to parse Scan data once we've found all MCUs (Minimum Coded Unit) of the image.
Fixes 7828.
I happened to notice that some inequalities had the wrong order, and was surprised since I thought that the `yoda` rule should have caught that.
However, reading http://eslint.org/docs/rules/yoda#options a bit more closely than previously, it's quite obvious that the `onlyEquality` option does *exactly* what its name suggests. Hence I think that it makes sense to adjust the options such that only ranges are allowed instead.
There's still some work necessary if we want to be able to run (even a subset of) the API unit-tests on Travis.
However, this patch could be considered a small first step, since the relevant unit-tests will now rely on a `CanvasFactory` rather than using `document.createElement('canvas')` directly.
This patch gets rid of the only case in the code-base where we're throwing a plain `string`, rather than an `Error`, which besides better/more consistent error handling also allows us to enable the [`no-throw-literal`](http://eslint.org/docs/rules/no-throw-literal) ESLint rule.
The point of this patch is to fix a couple of inconsistencies in `viewer.html`, compared to the locale files, such that the viewer would work correctly even without the `l10n/` files present.
*Note:* Since this isn't changing any of the locale files, we should *not* need to update any of the string names.
Looking through the history of the findbar code, it seems that the `findPrevious`/`findNext` buttons have never had a `title` set (note PR 2168, which was the initial findbar implementation).
Furthermore, the `placeholder` of the `findInput` didn't agree 100% with locale file either, so this is also adjusted.
I noticed that we have a `.toolbarButton.group` CSS class, which is currently applied to some buttons in the viewer. Since it attempts to adjust the `margin-right` property, I was initially a bit puzzled as to why there wasn't different rules for LTR/RTL locales.
However, checking the viewer with the DevTools inspector, in both LTR and RTL locales, I quickly found that the rule in question is *always* being overridden by other CSS rules.
It thus seem to me that while this rule was probably useful at some point, it has been dead for years and could now be removed.
The find functionality is currently not available for small devices
because the find dialog is not responsive. This patch fixes that.
To achieve this goal, the HTML is changed to always show the find
button. To prevent issues because of the addition of an extra button for
small views, the previous/next page buttons are hidden if the view
becomes small. These buttons are not useful anyway because on small
devices navigation is usually done via scrolling. The find functionality
is much more useful to have in this case. Moreover, we wrap the existing
elements into separate `div`s so that the browser can position the
elements itself when the view becomes smaller and logically connected
elements stay together when this happens.
In the CSS, extra rules for the find bar have been added to ensure that
the dialog's doorhanger is always below the find button. All findbar
`div`s are forced to be 32 pixels high to prevent the find message text
being aligned under the checkboxes. Finally, the find message is only
visible when there is actually text to display. This prevents wrapping
issues because, by default, the label has padding and margin even if
there is no text.
Instead of just upstreaming the changes from [bug 1345253](https://bugzilla.mozilla.org/show_bug.cgi?id=1345253) as-is, it seemed better to simply get rid of the loops altogether and use the same approach as in `PDFViewer`/`PDFThumbnailViewer`.
This patch is another step towards enabling Babel. Since we're moving
towards ES6 modules, we will not be using UMD headers anymore, so we can
remove the validation.
First of all, `compatibility.js` is already available in `lib/shared/compatibility.js`. Second of all, as can be seen in 9142301f35 (diff-9432ebaa58e10ab02874fcb86f689caa), the `lib/web/compatibility.js` file cannot work since the `require` statement isn't compatible with the output of `gulp lib`.
*This is something that I noticed while working on PR 8126, which is (more) fallout from PR 6065.*
In general, it's actually *not* correct to return `Dict.empty` as the default value for non-existent properties. Please note that a prior PR, see https://github.com/mozilla/pdf.js/pull/5957#issuecomment-103112698, asked for that behaviour but I don't think that's right.
Obviously for properties that are (or should) be `Dict`s it makes sense, however certain properties can be e.g. Strings or Arrays instead. In the latter case, returning `Dict.empty` is just plain wrong, and it's quite fascinating that this hasn't caused any errors in practice. (The existing validation in the various getters has actually saved us here.)
Also, when looking at this code again, it seemed unnecessary to duplicate the `MAX_LOOP_COUNT` check since we could just return immediately instead.
When using content security headers to restrict connections to the same origin,
you may not make connections to `example.com`. This feature detection also
works with a request to the current location.
It appears that I accidentally broke this in PR 6065, sorry about that!
The issue in this particular PDF file is that there's `/Rotate` entries on different levels of the `/Pages` tree. We're supposed to use the `/Rotate` entry in the `/Page` dict (which is `0`), but because of an incorrect condition we instead ended up with the one from the `/Pages` dict (which is `180`).
Fixes 8125.
Given that the mailing list has now been closed, see [bug 1340296](https://bugzilla.mozilla.org/show_bug.cgi?id=1340296), and that there's no weekly meetings any more, we probably shouldn't mention either of those in CONTRIBUTING.md.
Instead, let's just suggest the IRC channel as a means of communication here.
Ideally, the `Annotation` class should not have anything to do with the
page's operator list. How annotations are added to the page's operator
list is logic that belongs in `src/core/document.js` instead where the
operator list is constructed.
Moreover, some comments have been added to clarify the intent of the
code.
Even though the PDF specification does not state that `Opt` fields are
inheritable, in practice there are PDF generators that let annotations
inherit the options from a parent.
This fixes something that I noticed while working with the code in `Catalog.getPageDict` when debugging issue 8088.
Note that while I don't have an example where this patch really matters, given that e.g. `PartialEvaluator.hasBlendModes` depends on the `objId` to avoid cyclic references this patch could potentially help for some PDF files.
As discussed on IRC, we need to check all nodes at the *bottom* of the tree to ensure that we find the correct `Page` dict.
Furthermore, this patch also gets rid of the caching present in a previous version, since it's not clear if that really helps.
Note that this patch purposely adds an `eq` test, using a reduced test-case, so that we can be sure that the algorithm actually finds the correct `Page` dict for each `pageIndex`.
Fixes 8088.
The download button in pdf.js doesn't work in iOS Chrome.
- It appears to be an issue with URLs from URL.createObjectURL.
The URL is correct, but iOS Chrome won't even load the URL
when `a.click()` is called in `download_manager.js`. Even
if you manually visit the URL, you get a blank page.
- Fix this by detecting iOS Chrome and disabling createObjectURL.
The `download_manager.js` `download` method wasn't checking
`PDFJS.disableCreateObjectURL`, so check it there, too.
- Move the navigator.msSaveBlob check earlier, so that
this doesn't change IE10 / IE11 behavior.
- Remove the !URL check since pdf.js has a URL polyfill
now.
[api-minor] Refactor fetching of built-in CMaps to utilize a factory on the `display` side instead, to allow users of the API to provide a custom CMap loading factory (e.g. for use with Node.js)
Currently the built-in CMap files are loaded in `src/core/cmap.js` using `XMLHttpRequest` directly. For some environments that might be a problem, hence this patch refactors that to instead use a factory to load built-in CMaps on the main thread and message the data to the worker thread.
This is inspired by other recent work, e.g. the addition of the `CanvasFactory`, and to a large extent on the IRC discussion starting at http://logs.glob.uno/?c=mozilla%23pdfjs&s=12+Oct+2016&e=12+Oct+2016#c53010.
Previously, we had a function called `getDefaultAppearance`. This name,
however, is misleading as the method gets the normal appearance (in the
`N` entry) and not the default appearance (in the `DA` entry). Moreover,
it was not entirely clear how it works just from reading the code. It
primarily lacks comments and explicit error case handling.
This patch improves the situation by fixing the issues mentioned above
and making this function a proper method of the `Annotation` class, just
like e.g., `setColor` and `setBorderStyle`.
This patch basically reverts one aspect of TrueType (3, 1) cmap parsing to the state prior to PR 4259. After that PR, a number of regressions occurred in this particular code-path, which necessitated a number of follow-ups such as PRs 5703, 5743, and 6425.
The empirical data suggests, at least to me, that we should always prefer a (3, 1) cmap for TrueType fonts when they have an encoding, regardless of the Symbolic font flag.
Obviously this patch passes all unit/font/reference tests locally, and I made sure that all the PRs mentioned above landed with test-cases included.
However, in my opinion, there's still a very real possibility that this patch could potentially cause new regressions.
Given that the PDF file in bug 1337429 has been broken for almost *three* years before anyone noticed, and considering that the code-path in question has been the source of numerous regressions, I do *not* intend to request uplift of this patch to previous Firefox versions (assuming that it's even accepted).
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1337429.
*Yet another thing that I unfortunately missed during review of PR 8023.*
Note that previously, in `make.js` this file was being preprocessed, however as far as I can tell that wasn't actually necessary. Hence this patch just copies the file to the proper output directory.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1338395#c8.
Consume the current character when encountering illegal characters in `Lexer.getObject`, in order to prevent infinite loops during reading of streams (issue 8061)
*Please note:* The rendering of the PDF file in issue 8061 first regressed in PR 7039, and then PR 7493 exacerbated the problem even further by causing an infinite loop.
In this particular case, when errors were encountered inside of the `Lexer.getObject` method *itself*, we didn't advance the stream position. This thus caused an inifinite loop in `parseCMap`, since the exact same character was then parsed over and over again.
Fixes 8061.
[Chromium addon] Prevent errors that break the addon, caused by the `DEFAULT_URL` constant being replaced by a `defaultUrl` viewer configuration parameter (PR 8046 follow-up)
If users want to download, they can quickly click on the Download button
in the newly opened viewer.
The blobUrl logic for Firefox relies on `disableCreateObjectURL` is
never false in Firefox. If the assumption is invalid, then PDF
attachments at the attachment view will not correctly be displayed,
because a data-URL will be generated and `?<filename>` is treated as
part of the data:-URL.
Determine the page rotation at the same place as where the page size is
determined. This allows us to implement custom print page rotation logic
in one place, in the future.
Note that I initially tried to add this as a parameter to the `PDFPageProxy.render` method, such that it could be passed to `PartialEvaluator.getOperatorList`.
However, given all the different code-paths that call `getOperatorList` (there's a bunch only in `annotation.js`), this seemed to very quickly become unwieldy and thus difficult to maintain compared to simply using the existing `evaluatorOptions`.
The download method (and the PDF document properties) detect the
file name using `getPDFFileNameFromURL`. The title ought to also
display the PDF filename if available.
This patch lets the AcroForms example make use of the built-in interactive
forms functionality in PDF.js. This makes the example:
- much easier to understand;
- more feature-complete;
- in sync with the core when new functionality is added;
- similar to the other examples in terms of structure.
Fixes extra canvas create calls.
Fixes unnecessary call of `new DOMCanvasFactory`.
Fixes undefined error of DOMCanvasFactory.
Fixes failures in some of the tests.
Fixes expected behaviour.
Remove unused vars.
The `Driver._cleanup` method is removing all stylesheets between test runs, which causes "TypeError: styleElement.parentNode is null" console errors in `FontLoader.clear`.
As can also be seen during various tests, some of the changes I made in PR 7972 unfortunately causes console errors.
It seems that I didn't test this properly, since it *should* have been obvious to me that while tests are triggered using Node.js, the files in question are run within the *browser*.
My apologies for not testing this thoroughly, and for causing unnecessary churn in the code!
The last (and only) usage of `MOZ_CENTRAL` was removed in PR 3036, so it's been unused for almost four years now.
If we need to have different code-paths for `FIREFOX`/`MOZCENTRAL` builds, the preprocessor should (and has) been used instead.
See http://eslint.org/docs/rules/brace-style.
Having the opening/closing braces on the same line can often make the code slightly more difficult to read, in particular for `if`/`else if` statements, compared to using new lines.
This patch also, for consistency with `mozilla-central`, enables the [`no-iterator`](http://eslint.org/docs/rules/no-iterator) rule. Note that this rule didn't require a single code change.
When a blob or data-URL is opened with the extension, viewer.html
rewrites the URL. But when the viewer is refreshed (e.g. F5), Chrome
would fail to display the viewer because the extension router was not
set up to recognize such URLs.
Now it is.
The regular expression incorrectly marked a group as capturing.
For `http://example.com/#file.pdf`, the expected result is "file.pdf",
but instead "document.pdf" was returned.
This property was added all the way back in PR 542, but hasn't actually been relied upon ever since PR 692.
Note that there's a `isStream()` utility function which replaced the property years ago, hence the `isStream` property is now dead code.
Other PDF viewers, e.g. Adobe Reader, seem to append `FileAttachment`s to their attachments views.
One obvious difference in PDF.js is that we cannot append all the annotations on document load, since that would require parsing *every* page. Despite that, it still seems like a good idea to add `FileAttachment`s, since it's thus possible to access all the various types of attachments from a single place.
*Note:* With the previous patch we display a notification when a `FileAttachment` is added to the sidebar, which thus makes appending the contents of these annotations to the sidebar slightly more visible/useful.
A longstanding issue with the viewer is that you cannot tell if a PDF document includes an outline and/or attachments without actually opening the sidebar.
This patch contains a suggested solution for that, by displaying an hide-on-interaction notification on the `sidebarToggle` button (and the relevant sidebar view buttons). Note that this was inspired by e.g. the update notification that is displayed on the menu button in Firefox.
For an initial implementation, I've tried to do this in such a way that the notification isn't too distracting. Without being an UX expert, I don't think that we'd want something too in-your-face, in order to keep the viewer toolbars reasonable clean. (We probably do *not* want e.g. an entire notification bar in these situations, since that would take up unnecessary screen space and require actions from the user to close.)
However it's certainly possible that the current notification might simply be *too* inconspicuous to be truly helpful to users, but we could probably iterate on that if the feature itself is deemed useful.
Please see http://eslint.org/docs/rules/no-unused-vars; note that this patch purposely uses the same rule options as in `mozilla-central`, such that it fixes part of issue 7957.
It wasn't, in my opinion, entirely straightforward to enable this rule compared to the already existing rules. In many cases a `var descriptiveName = ...` format was used (more or less) to document the code, and I choose to place the old variable name in a trailing comment to not lose that information.
I welcome feedback on these changes, since it wasn't always entirely easy to know what changes made the most sense in every situation.
Given the nature of `EOF` and `isEOF`, it seems to me that they really ought to be placed in `core/primitives.js` instead.
In general, it doesn't seem great to have to depend on the entire `core/parser.js` file for such simple primitives/helper functions.
In particular, while `core/ps_parser.js` is completely separate from `core/parser.js` with regards to its function, it still depends on the latter for just *one* primitive.
Note that compared to e.g. PR 7389, this will not reduce the number of dependencies for `core/ps_parser`, however the new dependency IMHO makes more sense.
Given that this patch causes a lot of churn in the addon code, I wouldn't really mind if we ultimately decide against doing this and just add a rule exception in mozilla-central instead.[1]
---
[1] Note that I used the ESLint `--fix` option, hence writing this commit message actually took longer time than the creation of the patch :-)
PR 7322 added the `PdfJsNetwork.jsm` file, instead of the general `src/core/network.js` file for the Firefox addon. However, `make.js` wasn't updated to actually stop including the now obsolete network file.
*This fixes a regression from commit c9a0955c9c, i.e. PR 7738.*
Currently if you quickly rotate a document at least *twice*,[1] such that rendering of a page hasn't finished for the first rotation before the last rotation is triggered, the `cssTransform` method can fail to update the page correctly leading to it looking temporarily distorted.
The reason why things break is that previously we stored the `viewport` on the canvas DOM element, meaning that when it was accessed in `cssTransform` is was guaranteed to point to the `viewport` of the `zoomLayer` canvas.
Generally you want to avoid storing data on DOM elements this way, and during the `PDFPageView` refactoring needed to support SVG rendering, the previous `viewport` was instead stored directly on `PDFPageView`.
However, the problem is first of all that the `paintedViewport` only stores the *last* `viewport` computed, and second of all that there're no guarantees that it actually applies to the current `zoomLayer` canvas.
If a document is rotated slowly enough that rendering finishes *before* the next rotation then this problem doesn't exist, but for sufficiently quick rotations rendering will be cancelled at least once and the `paintedViewport` could thus be bogus.
The solution for the above problems is to ensure that we track the correct `viewport` for each DOM element (canvas or svg),[2] which seemed easist to do with a `WeakMap`.[3]
---
[1] I'm able to reproduce this using the `tracemonkey` file, but please note that for pages with few operations, i.e. that render very quickly, the effect may be hard to spot.
[2] One other possible solution that I briefly considered, was to wait until rendering finished before storing the current `viewport`. However, that would have caused issues with rotating a page before the *first* rendering operation had finished.
[3] This regression took me way longer to both figure out, and fix, than I'd like to admit :-)
*Please note: ignoring whitespace changes is most likely necessary for the diff to be readable.*
This patch addresses all the current, in `mozilla-central`, linting failures in the addon. It should thus be possible to change the `.eslintignore` entry for PDF.js in `mozilla-central` from `browser/extensions/pdfjs/**` to `browser/extensions/pdfjs/build/**` and `browser/extensions/pdfjs/web/**` instead.
Note that we cannot, for backwards compatibility reason of the general PDF.js library, at this time make similar changes for files residing in the `build` and `web` directories in `mozilla-central`.
The main changes in this patch are that we now use [classes](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Classes) instead of our previous "class-like" functions, and also use the more compact [object shorthand notation](https://developer.mozilla.org/en/docs/Web/JavaScript/Reference/Operators/Object_initializer#New_notations_in_ECMAScript_2015).
A couple of functions were also converted to [arrow functions](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Functions/Arrow_functions), to reduced usages of `bind(this)` and `var self = this`.
One caveat with ES6 classes is that it's not (yet) possible to define private constants/helper functions within them, which is why the `NetworkManagerClosure` was kept to not change the visibility of those constant/functions.
Besides testing in Firefox Nightly 53, this patch has also been tested in Firefox ESR 45 and SeaMonkey 2.46.
However, I'd gladly welcome help with testing the patch more, to ensure that nothing has gone wrong during the refactoring.
Fixes the first bullet point of issue 7957.
http://eslint.org/docs/rules/no-lone-blocks
Note that we currently have no code that violates this rule in the source files, but it seems that the built files are possibly affected (see issue 7957).
Please see http://eslint.org/docs/rules/spaced-comment.
Note that the exceptions added for `line` comments are intended to still allow use of the old preprocessor without linting errors.
Also, I took the opportunity to improve the grammar slightly (w.r.t. capitalization and punctuation) for comments touched in the patch.
Further adjust the heuristics used to detect OpenType font files with CFF data, to ensure that all Type0 fonts are handled the same way regardless of font Subtype (issue 7901)
Rename `annotation_layer_spec.js` to `annotation_spec.js` to better describe what is actually tested, and simplify the `FileAttachmentAnnotation` unit-test to avoid having to use the entire API in the test
Every other unit-test in `annotation_spec.js` is already only testing the annotation code. Hence it seems unnecessarily convoluted to make use of the API here, when we can (fairly) simply provide the necessary data explicitly as in all the other annotation unit-test.
Currently you have to open the files to be certain which tests each one will run, which we can avoid by appending the directory names to the file names of the tests themselves.
We're currently making use of `uniquePrefix`/`idCounters` in multiple files, to create unique object id's, and adding a new occurrence of them requires some care to ensure that an object id isn't accidentally reused.
Furthermore, having to pass around multiple parameters as we currently do seem like something you want to avoid.
Instead, this patch adds a factory which means that there's only *one* thing that needs to be passed around. And since it's now only necessary to call a method in order to obtain a unique object id, the details are thus abstracted away at the call-sites which avoids accidental reuse of object id's.
To test that this works as expected a very simple `Page` unit-test is added, and the existing `Annotation layer` tests are also adjusted slightly.
Using `else` after `return` is not necessary, and can often lead to unnecessarily cluttered code. By using the `no-else-return` rule in ESLint we can avoid this pattern, see http://eslint.org/docs/rules/no-else-return.
According to https://wiki.mozilla.org/RapidRelease/Calendar#Past_branch_dates: The *last* ESR version of Firefox 38 was released in April this year, and since June the only available ESR version has been based on Firefox 45.
Now that Seamonkey has *finally* released a new version, i.e. 2.46 which should correspond to Firefox 49, there doesn't seem to be any reason to keep the fallback code around in the addon anymore.
[api-minor] Ensure that the `getDocument` Promise is rejected if the `loadingTask` is destroyed, or an `Error` is thrown, inside of the `onPassword` callback (issue 7806)
We're already passing in a, currently unused, `PdfManager` instance when initializing the `XRef`. To avoid having to pass a single `password` parameter around, we could thus simply get the `password` through the `PdfManager` instance instead.
This patch also removes the `UpdatePassword` message, in favour of using the `sendWithPromise` method of `MessageHandler`.
Furthermore, the patch also refactors the `BasePdfManager_updatePassword`/`BasePdfManager_passwordChanged` methods (in pdf_manager.js), and the `pdfManagerReady` function (in worker.js).
I recently happened to look at the code I wrote for PR 5964, which fixed [bug 1157493](https://bugzilla.mozilla.org/show_bug.cgi?id=1157493), and I quickly realized that the solution is way too simplistic.
The fact that only using the `length` of a `Differences` array worked seems more like a happy accident for a particular set of font data, but could just as easily be incorrect for other PDF files.
Note that in practice, the case where the `Encoding` entry is a regular `Dict` (and not a `Ref` or `Name`) is very rare, hence I don't think that we really need to worry about having to reparse this data.
Also, the performance of this code-block is quite a bit better by updating the `hash` with the data from the *entire* `Differences` array, instead of at every loop iteration.
Let `finishPaintTask` in pdf_page_view.js return a promise instead, to avoid having to throw in the `paintTask.promise` rejection handler, and don't reject the `PDFPageView_draw` promise when rendering is `cancelled`
Changing this particular code makes me somewhat nervous about regressions, since PR 5770 necessitated the follow-up PR 6270.
However, the patch passes all tests added in those PRs (and obviously all other tests). Furthermore, I've manually checked all the issues/bugs referenced in PRs 5770 and 6270 without finding any issues.
**Please note:** This patch fixes *only* the font bug, not the SVG conversion, present on pages two and three of the PDF file in issue 7901.
As mentioned on IRC yesterday, we currently throw even when rendering is `cancelled`, which is annoying when the devtools are active. Furthermore, since `cancelled` isn't really an error, rejecting the `PDFPageView_draw` promise seems somewhat strange in that case.
Modern browsers support styling radio buttons and checkboxes with CSS.
This makes the implementation much easier, and the fallback for older
browsers is still decent.
I haven't got an example where the current code breaks, but given all the previous cases we've seen where PDF generators use indirect objects in Arrays it makes sense to fix this pro-actively.
I've modified the relevant unit-tests slightly, and they would *not* pass without the code changes in this patch.
*Note:* `Dict_getArray` only dereferences Array elements on the "top-level", to avoid recursion issues. Furthermore if you have to loop through the Array at the call-site anyway, then using `Dict_get` in combination with `XRef_fetchIfRef` is a tiny bit more efficient.
*Please note that most of the necessary code adjustments were made in PR 7890.*
ESLint has a number of advantageous properties, compared to JSHint. Among those are:
- The ability to find subtle bugs, thanks to more rules (e.g. PR 7881).
- Much more customizable in general, and many rules allow fine-tuned behaviour rather than the just the on/off rules in JSHint.
- Many more rules that can help developers avoid bugs, and a lot of rules that can be used to enforce a consistent coding style. The latter should be particularily useful for new contributors (and reduce the amount of stylistic review comments necessary).
- The ability to easily specify exactly what rules to use/not to use, as opposed to JSHint which has a default set. *Note:* in future JSHint version some of the rules we depend on will be removed, according to warnings in http://jshint.com/docs/options/, so we wouldn't be able to update without losing lint coverage.
- More easily disable one, or more, rules temporarily. In JSHint this requires using a numeric code, which isn't very user friendly, whereas in ESLint the rule name is simply used instead.
By default there's no rules enabled in ESLint, but there are some default rule sets available. However, to prevent linting failures if we update ESLint in the future, it seemed easier to just explicitly specify what rules we want.
Obviously this makes the ESLint config file somewhat bigger than the old JSHint config file, but given how rarely that one has been updated over the years I don't think that matters too much.
I've tried, to the best of my ability, to ensure that we enable the same rules for ESLint that we had for JSHint. Furthermore, I've also enabled a number of rules that seemed to make sense, both to catch possible errors *and* various style guide violations.
Despite the ESLint README claiming that it's slower that JSHint, https://github.com/eslint/eslint#how-does-eslint-performance-compare-to-jshint, locally this patch actually reduces the runtime for `gulp` lint (by approximately 20-25%).
A couple of stylistic rules that would have been nice to enable, but where our code currently differs to much to make it feasible:
- `comma-dangle`, controls trailing commas in Objects and Arrays (among others).
- `object-curly-spacing`, controls spacing inside of Objects.
- `spaced-comment`, used to enforce spaces after `//` and `/*. (This is made difficult by the fact that there's still some usage of the old preprocessor left.)
Rules that I indend to look into possibly enabling in follow-ups, if it seems to make sense: `no-else-return`, `no-lonely-if`, `brace-style` with the `allowSingleLine` parameter removed.
Useful links:
- http://eslint.org/docs/user-guide/configuring
- http://eslint.org/docs/rules/
Move the `Preferences` initialization/fetching code to the top of `PDFViewerApplication.initialize`, to enable using them when initializing e.g. the viewer components
Ideally we'd remove the 'localized' event from the `eventBus`, but for backwards compatibility we keep it in `GENERIC` builds.
Note that while we want to ensure that the direction attribute of the HTML is updated as soon as the `localized` Promise is resolved, we purposely wait until the viewer has been initialized to ensure that the 'localized' event will always be dispatched.
With `bindEvents()` now being called after the viewer has been initialized, we no longer need to have `PDFViewerApplication.initialized` checks in the event handler functions.
Furthermore by moving the `window.addEventListener`s to a helper method, `PDFViewerApplication.initialized` checks are no longer necessary in the event handlers, hence we thus address part of issue 7797 here as well.
Note that in quick testing using `console.time/timeEnd`, both locally and with the Firefox addon, the total run time of the *entire* `PDFViewerApplication.initialize` function does not seem to change with this patch.
This patch refactors the `CropBox` code to combine fetching and
validation code in a getter, like we already did for the `MediaBox`
property. Combined with variable name changes, this improves readability
of the code and makes the `view` getter simpler as well.
I've not actually, thus far, come across a PDF file that this patch fixes. However, given the string of recent patches that has fixed issues with indirect objects in arrays, I think that it makes sense to proactively avoid any issues in this code.
I just realized that none of our current unit-tests cover this particular part of the Page Label parsing code, hence this patch adjusts an existing test PDF to include a "St" entry in the Page Label dictionary.
This patch avoids the creation of extra arrays when initializing an
array with default (zero) values. Doing this additionally makes the code
more readable by allocating enough space for the number of color
components.
It seems that for normal web pages, at least in Firefox, the keyboard shortcuts <kbd>Ctrl</kbd> + <kbd>Up</kbd>/<kbd>Down</kbd> are functionally equivalent to <kbd>Home</kbd>/<kbd>End</kbd>. This is obviously an edge-case, but can be easily implemented by using the same logic as we do for <kbd>Home</kbd>/<kbd>End</kbd>.
Fixes 7852.
*Please note:* I'm finding it slightly difficult to interpret issue 7852, and bug 1285719, since among other things: the title includes the word "reverse" with no other mention of it, and the STR makes reference to print preview which doesn't seem applicable to the PDF viewer.
However, compared to regular web pages in Firefox, I think the behavior of this patch makes sense here.
For Arabic characters, the Unicode character codes are mapped to Unicode
character types using the character codes for indexing. However, the
character code 0x061D is undefined (and therefore invalid) in the
Unicode standard. The imported list does not contain this entry, but not
having it in the list breaks indexing for items after it. Therefore, put
an empty string on its position to make indexing work properly and issue
a warning in the unlikely event that we encounter this character.
This patch moves the user agent checks to the top of the file to reduce
duplication and to provide a clear overview of which user agent we are
detecting.
Moreover, we extract inline user agent checks as well and use existing
checks in more places. Finally, we fix the indenting in one place for
consistency.
*This patch fixes something that I noticed while debugging https://bugzilla.mozilla.org/show_bug.cgi?id=1308536.*
The PDF file contains a font called "NuptialScript", which unfortunately is not embedded. Since that is a non-standard font we will not be able to render it entirely correct. However, by adding "NuptialScript" to the `getNonStdFontMap`, we can at least improve the rendering slightly by using an italic (serif) fallback font.
This patch adds support for non-embedded Arial Black fonts, that use a `Arial-Black...` format for the font names.
Also, this patch changes `canvas.js` such that we always render Arial Black fonts with the maximum weight, which actually improves a number of existing test-cases. This should thus explain the test "failures", which are clear improvements compared with e.g. Adobe Reader.
Fixes 7835.
For consistency, I also renamed the `FIREFOX/MOZCENTRAL` sessionStorage key, but given that sessionStorage is a lot less permanent than localStorage it didn't seem necessary to migrate any existing values.
Fixes 7760.
Currently if you try to enable SVG rendering in the addons, a `TypeError` is thrown by the browser since we have code that depends on what `paintOnCanvas`/`paintOnSvg` (should) return.
For commands with with too few arguments, clear out `args` if it's an Array instead of replacing it with `null` in `EvaluatorPreprocessor_read` (issue 7804)
For `PartialEvaluator_getTextContent`, the same `args` Array should be re-used for every `EvaluatorPreprocessor_read` call. Hence we want to ensure that it's not accidentally replaced with `null` in `EvaluatorPreprocessor_read`, since otherwise corrupt PDF files (with too few arguments for certain commands) will cause errors in `PartialEvaluator_getTextContent`.
Perhaps a micro-optimization, but this patch also changes two `!args` comparisons to `args === null`, since that should be a tiny bit more efficient.
According to e.g. issue 7754, it appears that the current `isSafari` check is failing in newer version of the browser. Despite the fact that checking the userAgent is an anti-pattern, which should be avoided, it's currently the simplest solution.
By only allowing very specific type of `JavaScript` actions, and also utilizing the existing `URL` validation, this patch shouldn't pose too much risk.
Fixes one of the points in issue 3897 (with the PDF file taken from issue 3438).
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=843699 (probably, since that bug doesn't contain a test-case).
Given that the `customScaleOption` should already be hidden, provided that the browser supports that, this patch also prevents it from being accessible via the keyboard.
As far as my testing goes in various browsers, this doesn't seem to have any ill effects, and note that we're already explicitly ignoring the `custom` value in the `select` event handler.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1315608.
This patch resolves the responsiveness issues for the toolbar in the
viewer. Depending on the language (for example the Dutch language),
elements could overlap when the viewport size is reduced.
The main issue here is that the CSS rules are unnecessarily complex and
handle lots of different cases (LTR/RTL, displacements for specific
viewport widths, et cetera). By removing this complexity and letting the
browser handle the responsiveness, we not only get simpler CSS rules and
HTML mark-up, but the responsiveness issues are mostly fixed at the same
time. We no longer have to position the elements manually (by setting
their `left` attribute value) anymore.
It seems that certain bad PDF generators can create badly encoded "Prefix" entries for Page Labels, one example being http://ukjewishfilm.org/wp-content/uploads/2015/09/Jewish-Film-Festival-Programme-ONLINE.pdf.
Unfortunately I didn't come across such a PDF file while adding the API support for Page Labels, but with them now being used in the viewer I just found this issue. With this patch, we now display the Page Labels in the same way as Adobe Reader.
The original code is difficult to read and, more importantly, performs
actions that are not described in the specification. It replaces empty
names with a backtick and an index, but this behavior is not described
in the specification. While the specification is not entirely clear
about what should happen in this case, it does specify that the `T`
field is optional and that multiple field dictionaries may have the same
fully qualified name, so to achieve this it makes the most sense to
ignore missing `T` fields during construction of the field name. This is
the most specification-compliant solution and, judging by opened issue #6623, also the required and expected behavior.
I got tired of staring at a bunch of localization warnings every time that I open the console, hence this patch adds the missing translations to the Swedish locale.
- Renamed startPrint to performPrint to emphasize that the method
does not start the print process (preparing pages for the printer),
but that it does the actual printing (sending pages off to the
printer).
- Put performPrint in the PDFPrintService, so that it can be
overridden if needed.
- Move the global scratchCanvas to PDFPrintService. This is mainly to
make it easier to reason about the state of scratchCanvas. In practice
there is no difference because only one PDFPrintService instance can
be instantiated at any given time.
- Move all logic of using the rendered page to one location.
This makes it easier to replace the printing logic later, when I add
special handling to out-of-process frames in the Chrome extension.
Make sure that the print service is stopped as soon as possible when
aborted, and that it is not possible for a (slow) promise to
accidentally wipe the state of a print job that was started later.
Remove/deprecate specifying a pageNumber directly after the hash symbol (#), to improve compatibility since other PDF viewers don't support this form (issue 7746)
There's no mention of our `#{pagenum}` form in http://www.adobe.com/content/dam/Adobe/en/devnet/acrobat/pdfs/pdf_open_parameters.pdf, and Adobe Reader doesn't seem to support it either.
Hence this patch removes support for it in the extensions, but keeps it in the `GENERIC` build with a deprecation warning and a fallback to handle it as a destination.
Fixes 7746.
This patch implements the page label functionality in a similar way as Adobe Reader.
For documents with page labels, if a non-existent page label is entered we'll try to fallback to the page number instead.
The patch also includes a preference (`disablePageLabels`), to make it easy to opt-out of using page labels if the user/implementor so wishes.
The way that `get/set currentPageLabel` is implemented in `PDFViewer`, is as wrappers for the corresponding `get/set currentPageNumber` functions, since that seemed like the cleanest solution.
The page labels are purposely *only* added to the page controls in the viewer UI, and not stored in e.g. the `ViewHistory`. Since doing so would mean adding unnecessary code complexity, without any real added value, and would also mean delaying the inital loading of PDF documents.
Note that this patch will ignore page labels if they are identical to standard page numbering, since in this case displaying the page labels adds no value (but only UI noise). The reason for handling this case specially, is that in practice a surprising number of PDF files include "pointless" page labels.
The following reasoning was used for deciding to remove the "Page: " label, and replace it with a tooltip, from the main toolbar:
- We have no other visible labels in the *main* toolbar (e.g. the Zoom dropdown doesn't have a label, but only a tooltip).
- We already hide the "Page: " label when the viewer is narrow.
- The varying width of the "Page: " label in different locales is already causing issues for many languages, with overlap in the main toolbar as a result.
Trying to create responsive CSS styles that works well in all locales is already difficult, and if we add support for page labels that will only further compound the issues.
- Some PDF viewers (e.g. Adobe Reader, pdfium in Chrome) doesn't show labels in the UI by default.
In some PDF files, the first element (i.e. the one containing either a `Ref` or a `Number` pointing to a page) of the explicit destination Array may be bogus.
One such example is actually the file `pdf.pdf` in the test-suite, where some destinations are incompletely specified. One such example being the `G1.998360` destination whose explicit destination Array contains `[null, /XYZ, 54, 488, null]`, i.e. the destination page is `null`.
Hence this patch adds a bit more validation for that case. It also adds an additional check to ensure that the resulting `pageNumber` is non-negative, and finally a couple more error messages for existing cases of malformed data.
In general we neither want, nor can, support arbitrary `Launch` actions. But in practice, all the cases we've seen so far just contains relative URLs to other PDF files. Building on PR 7689, we can thus at least support basic `Launch` actions.
Note that in `FIREFOX/MOZCENTRAL/CHROME` builds of the standard viewer the `docBaseUrl` parameter will be set by default, since in that case it makes sense to use the current URL as a base.
For the `GENERIC` viewer, or the API itself, it doesn't make sense to try and set the `docBaseUrl` by default. However, custom deployments/implementations may still find the parameter useful.
Note that this will automatically reject any relative URL.
To make the API more useful to consumers, URLs that are rejected will be available via the `unsafeUrl` property in the data object returned by `PDFPageProxy_getAnnotations`.
The patch also adds a bit more validation of the data for `Named` actions.
This not only reduces code duplication, but it also allow us to easily support the same kind of URLs we currently do for Link annotations in the Outline as well.
This requires the `run-sequence` dependency because Gulp does not have a
way to run sequences of tasks inside a Gulp task. Gulp 4.0 will have
support for this, but until that is released this is the recommended way
to implement it.
The outline toggle button has a feature where it can be double-clicked
to expand/collapse all items shown therein. Although this is described
in the FAQ, can go potentially unnoticed. This, however, being a useful
feature, advertise on the tool tip itself.
l10n translation for en-US and IDs updated.
Signed-off-by: Jeenu Viswambharan <jeenuv@gmail.com>
Each well-formed SVG image has the following structure:
SVG element
- Definitions element
- Root group
- Other group 1
- ...
- Other group n
This patch factors out initialization code into a private method in such
as way that the creation of this structure is clear from the code. The
root group is the replacement for the parent group from before. We need
this group as we cannot apply the viewport transform on the SVG element
itself (this caused issues in Chrome). If other code appends groups to
the SVG image, in reality it is appending those groups to the root
group, but this detail is abstracted away by this patch.
This patch ensures that we only create transformation groups when it is
actually required and that we re-use transform groups as much as possible.
It reduces the number of transform groups for the Tracemonkey paper from
2790 to 1271, thereby making the DOM much lighter and rendering/scrolling
smoother. Moreover, it simplifies the code and prevents duplication.
Finally, we issue a warning when an unimplemented graphic state is
encountered. Before, this was ignored silently, making debugging harder.
This patch:
- resolves a warning in the console about missing character encoding;
- makes the viewer use the same background color and PDF file as the
regular viewer;
- simplifies the example to bring it more in line with the other
examples.
Let `Parser_makeFilter` pass in the `DecodeParms` data to various image `Stream`s, instead of re-fetching it in various `[...]Stream.prototype.ensureBuffer` methods
In `Parser_filter` the `DecodeParms` data is fetched and passed to `Parser_makeFilter`, where we also make sure that a `Ref` is resolved to a direct object.
We can thus pass this along to the various image `Stream` constructors, to avoid the current situation where we lookup/resolve data that is already available.
Note also that we currently do *not* handle the case where `DecodeParms` is an Array entirely correct in the various image `Stream`s, and this patch fixes that for free.
While the array argument to TJ should only contain strings and numbers, other
unfortunate items are found in PDFs in the wild, e.g.:
[(Grandes) 0.0 Tc
-250.0 (Client\350les,) 0.0 Tc
-250.0 (Financements) 0.0 Tc
-250.0 (et) 0.0 Tc
-250.0 (March\351s) ] TJ
getOperatorList already properly ignores any non-string, non-numeric values in
TJ arrays; without this patch to getTextContent, returned text items can have
NaN widths due to calculations being applied to those non-numeric values.
For PDF files with multiple `/Filter`s, where the `/Length` entry is zero, we fail to render the file correctly. The reason is that `maybeLength` is `null` for the every filter except the first, and `!maybeLength` is thus truthy.
Hence it seems that we should completely ignore the `/Length` entry and also explicitly check `maybeLength === 0`.
Note that I've not (yet) come across a PDF file with this issue in the wild, but given all the stupid things PDF generators do I wouldn't be surprised if such a file actually exists. In order to prevent a possible future bug, I'm submitting this patch which includes a hand-edited PDF file that we currently cannot render correctly (but e.g. Adobe Reader can).
Even though this patch passes all tests (unit/font/reference) locally, including the new ones that I added in PR 7621, I'm still a bit nervous about modifying the code that choose the fallback encoding for fonts without an `/Encoding` entry.
Note that over the years this code has been changed on a number of occasions, see a possibly incomplete [list here], to deal with various cases of incorrect font data.
According to the PDF specification, see http://www.adobe.com/content/dam/Adobe/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#G8.1904184, it seems that we should fallback to the `StandardEncoding` for Nonsymbolic fonts.
There's obviously a risk that fixing this particular issue *could* break other PDF files for which we don't have tests. However I've tried to change the logic as little as possible in this patch, to hopefully reduce possible breakage.
Based on debugging numerous font issue, it seems that a lot of fonts actually set the Symbolic flag, even when they are in fact *not* Symbolic. Fonts actually marked as Nonsymbolic seem to be somewhat less common, which I hope should reduce the risk of the patch somewhat.
Fixes 7580.
2016-09-13 14:07:16 +02:00
1797 changed files with 429724 additions and 101517 deletions
@ -4,11 +4,11 @@ The issues are used to track both bugs filed by users and specific work items fo
If the issue is related to errors produced by a specific PDF, please always include the PDF by providing a URL where contributors can download it. Without a PDF for reproduction, such issues will be closed. We understand that many PDFs contain sensitive information, however having a PDF is essential to resolving the issue and building our regression testing suite. If possible, try creating a reduced example exhibiting the problem but not containing sensitive data. Also small PDFs are best suited for our regression testing. If an important issue only shows on sensitive PDFs, contributors might be willing to accept these PDFs via a secure exchange.
The issue tracking system is designed to record a single technical problem. A bug report is something where a developer/contributor can work on. The GitHub issue tracker is not a good place for general, not well thought out or unworkable ideas. Most likely a discussion-type issue will not be addressed for a long time or closed as invalid. The best place is our dev-pdf-js@lists.mozilla.org mailing list. You can subscribe to it using http://lists.mozilla.org or Google Groups. This way you will reach not only developers. As an alternative, you can join our weekly engineering meeting to discuss new ideas for the project.
The issue tracking system is designed to record a single technical problem. A bug report is something where a developer/contributor can work on. The GitHub issue tracker is not a good place for general, not well thought out or unworkable ideas. Most likely a discussion-type issue will not be addressed for a long time or closed as invalid. The best place for general discussions is our Matrix room at https://chat.mozilla.org/#/room/#pdfjs:mozilla.org.
If you are developing a custom solution, first check the examples at https://github.com/mozilla/pdf.js#learning and search existing issues. If this does not help, please prepare a short well-documented example that demonstrates the problem and make it accessible online on your website, JS Bin, GitHub, etc. before opening a new issue or contacting us on the IRC channel -- keep in mind that just code snippets won't help us troubleshoot the problem.
If you are developing a custom solution, first check the examples at https://github.com/mozilla/pdf.js#learning and search existing issues. If this does not help, please prepare a short well-documented example that demonstrates the problem and make it accessible online on your website, JS Bin, GitHub, etc. before opening a new issue or contacting us in the Matrix room -- keep in mind that just code snippets won't help us troubleshoot the problem.
Note that the translations for PDF.js in the `l10n` folder are synchronized with the Aurora branch of Mozilla Firefox. This means that we will only accept pull requests that add strings currently missing in the Aurora branch (because it will take at least six weeks before the most recent translations are in the Aurora branch), but keep in mind that the changes will be overwritten when we synchronize again.
Note that the translations for PDF.js in the `l10n` folder are imported from the Nightly channel of Mozilla Firefox, such that we don't have to maintain them ourselves. This means that we will not accept pull requests that add new languages and/or modify existing translations, unless the corresponding changes have been made in Mozilla Firefox first.
This repository is governed by Mozilla's code of conduct and etiquette guidelines.
For more details, please read the
[Mozilla Community Participation Guidelines](https://www.mozilla.org/about/governance/policies/participation/).
## How to Report
For more information on how to report violations of the Community Participation Guidelines, please read our '[How to Report](https://www.mozilla.org/about/governance/policies/participation/reporting/)' page.
<!--
## Project Specific Etiquette
In some cases, there will be additional project etiquette i.e.: (https://bugzilla.mozilla.org/page.cgi?id=etiquette.html).
PDF.js is built into version 19+ of Firefox, however one extension is still available:
+ [Development Version](http://mozilla.github.io/pdf.js/extensions/firefox/pdf.js.xpi) - This extension is mainly intended for developers/testers, and it is updated every time new code is merged into the PDF.js codebase. It should be quite stable, but might break from time to time.
+ Please note that the extension is *not* guaranteed to be compatible with Firefox versions that are *older* than the current ESR version, see the [Release Calendar](https://wiki.mozilla.org/RapidRelease/Calendar#Past_branch_dates).
+ The extension should also work in Seamonkey, provided that it is based on a Firefox version as above (see [Which version of Firefox does SeaMonkey 2.x correspond with?](https://wiki.mozilla.org/SeaMonkey/FAQ#General)), but we do *not* guarantee compatibility.
PDF.js is built into version 19+ of Firefox.
#### Chrome
@ -51,44 +48,53 @@ directory `build/chromium`.
To get a local copy of the current code, clone it using git:
$ git clone git://github.com/mozilla/pdf.js.git
$ git clone https://github.com/mozilla/pdf.js.git
$ cd pdf.js
Next, install Node.js via the [official package](http://nodejs.org) or via
Next, install Node.js via the [official package](https://nodejs.org) or via
[nvm](https://github.com/creationix/nvm). You need to install the gulp package
globally (see also [gulp's getting started](https://github.com/gulpjs/gulp/blob/master/docs/getting-started.md#getting-started)):
globally (see also [gulp's getting started](https://github.com/gulpjs/gulp/tree/master/docs/getting-started)):
$ npm install -g gulp-cli
$ npm install -g gulp-cli@^2.3.0
If you prefer to not install `gulp-cli` globally, you have to prefix all the `gulp` commands with `npx` (for example, `npx gulp server` instead of `gulp server`).
If everything worked out, install all dependencies for PDF.js:
$ npm install
Finally you need to start a local web server as some browsers do not allow opening
PDF files using a file:// URL. Run
> [!NOTE]
> On MacOS M1/M2 you may see some `node-gyp`-related errors when running `npm install`. This is because one of our dependencies, `"canvas"`, does not provide pre-built binaries for this platform and instead `npm` will try to build it from source. Please make sure to first install the necessary native dependencies using `brew`: https://github.com/Automattic/node-canvas#compiling.
Finally, you need to start a local web server as some browsers do not allow opening
PDF files using a `file://` URL. Run:
$ gulp server
and then you can open
and then you can open:
+ http://localhost:8888/web/viewer.html
It is also possible to view all test PDF files on the right side by opening
Please keep in mind that this assumes the latest version of Mozilla Firefox; refer to [Building PDF.js](https://github.com/mozilla/pdf.js/blob/master/README.md#building-pdfjs) for non-development usage of the PDF.js library.
It is also possible to view all test PDF files on the right side by opening:
+ http://localhost:8888/test/pdfs/?frame
## Building PDF.js
In order to bundle all `src/` files into two productions scripts and build the generic
viewer, issue:
In order to bundle all `src/` files into two production scripts and build the generic
viewer, run:
$ gulp generic
This will generate `pdf.js` and `pdf.worker.js` in the `build/generic/build/` directory.
If you need to support older browsers, run:
$ gulp generic-legacy
This will generate `pdf.js` and `pdf.worker.js` in the `build/generic/build/` directory (respectively `build/generic-legacy/build/`).
Both scripts are needed but only `pdf.js` needs to be included since `pdf.worker.js` will
be loaded by `pdf.js`. If you want to support more browsers than Firefox you'll also need
to include `compatibility.js` from `build/generic/web/`. The PDF.js files are large and
should be minified for production.
be loaded by `pdf.js`. The PDF.js files are large and should be minified for production.
## Using PDF.js in a web application
@ -97,55 +103,44 @@ or to build it from source. We supply pre-built versions for usage with NPM and
the `pdfjs-dist` name. For more information and examples please refer to the
[wiki page](https://github.com/mozilla/pdf.js/wiki/Setup-pdf.js-in-a-website) on this subject.
## Including via a CDN
PDF.js is hosted on several free CDNs:
- https://www.jsdelivr.com/package/npm/pdfjs-dist
- https://cdnjs.com/libraries/pdf.js
- https://unpkg.com/pdfjs-dist/
## Learning
You can play with the PDF.js API directly from your browser through the live
demos below:
You can play with the PDF.js API directly from your browser using the live demos below:
More examples can be found in the [examples folder](https://github.com/mozilla/pdf.js/tree/master/examples/). Some of them are using the pdfjs-dist package, which can be built and installed in this repo directory via `gulp dist-install` command.
For an introduction to the PDF.js code, check out the presentation by our
PDF.js heavily relies on the use of [Promises](https://developer.mozilla.org/docs/Web/JavaScript/Reference/Global_Objects/Promise). If promises are new to you, it's recommended you become familiar with them before continuing on.
This tutorial shows how PDF.js can be used as a library in a web browser.
[examples/](https://github.com/mozilla/pdf.js/tree/master/examples) provides more examples, including usage in Node.js (at [examples/node/](https://github.com/mozilla/pdf.js/tree/master/examples/node)).
### Document
The object structure of PDF.js loosely follows the structure of an actual PDF. At the top level there is a document object. From the document, more information and individual pages can be fetched. To get the document:
```js
PDFJS.getDocument('helloworld.pdf')
pdfjsLib.getDocument('helloworld.pdf')
```
Remember though that PDF.js uses promises, so the above will return a promise that is resolved with the document object.
Remember though that PDF.js uses promises, and the above will return a `PDFDocumentLoadingTask` instance that has a `promise` property which is resolved with the document object.
Each PDF page has its own viewport which defines the size in pixels(72DPI) and initial rotation. By default the viewport is scaled to the original size of the PDF, but this can be changed by modifying the viewport. When the viewport is created an initial transformation matrix will also be created that takes into account the desired scale, rotation, and it transforms the coordinate system (the 0,0 point in PDF documents the bottom-left whereas canvas 0,0 is top-left).
Each PDF page has its own viewport which defines the size in pixels(72DPI) and initial rotation. By default the viewport is scaled to the original size of the PDF, but this can be changed by modifying the viewport. When the viewport is created, an initial transformation matrix will also be created that takes into account the desired scale, rotation, and it transforms the coordinate system (the 0,0 point in PDF documents the bottom-left whereas canvas 0,0 is top-left).
```js
var scale = 1.5;
var viewport = page.getViewport(scale);
var viewport = page.getViewport({ scale: scale, });
// Support HiDPI-screens.
var outputScale = window.devicePixelRatio || 1;
var canvas = document.getElementById('the-canvas');
@ -21,7 +21,7 @@ Before downloading PDF.js please take a moment to understand the different layer
<tbody>
<tr>
<td>Core</td>
<td>The core layer is where a binary PDF is parsed and interpreted. This layer is the foundation for all subsequent layers. It is not documented here because using it directly is considered an advanced usage and the API is likely to change. For an example of using the core layer see the [PDF Object Browser](https://github.com/brendandahl/pdf.js.utils/tree/master/browser)
<td>The core layer is where a binary PDF is parsed and interpreted. This layer is the foundation for all subsequent layers. It is not documented here because using it directly is considered an advanced usage and the API is likely to change. For an example of using the core layer see the <ahref="https://github.com/brendandahl/pdf.js.utils/tree/master/browser">PDF Object Browser</a>.
</td>
</tr>
<tr>
@ -30,85 +30,98 @@ Before downloading PDF.js please take a moment to understand the different layer
</tr>
<tr>
<td>Viewer</td>
<td>The viewer is built on the display layer and is the UI for PDF viewer in Firefox and the other browser extensions within the project. It can be a good starting point for building your own viewer. *However, we do ask if you plan to embed the viewer in your own site, that it not just be an unmodified version. Please re-skin it or build upon it.*</td>
<td>The viewer is built on the display layer and is the UI for PDF viewer in Firefox and the other browser extensions within the project. It can be a good starting point for building your own viewer. <em>However, we do ask if you plan to embed the viewer in your own site, that it not just be an unmodified version. Please re-skin it or build upon it.</em></td>
</tr>
</tbody>
</table>
## Download
Please refer to [this wiki page](https://github.com/mozilla/pdf.js/wiki/Frequently-Asked-Questions#faq-support) for information about supported browsers.
<divclass="row">
<divclass="col-md-6">
<h3>Pre-built</h3>
<divclass="col-md-4">
<h3>Prebuilt (modern browsers)</h3>
<p>
Includes the generic build of PDF.js and the viewer.
└── package.json - package definition and dependencies
```
## Trying the Viewer
With the prebuilt or source version open `web/viewer.html` in a browser and the test pdf should load. Note: the worker is not enabled for file:// urls, so use a server. If you're using the source build and have node, you can run `gulp server`.
With the prebuilt or source version, open `web/viewer.html` in a browser and the test pdf should load. Note: the worker is not enabled for file:// urls, so use a server. If you're using the source build and have node, you can run `gulp server`.
## More Information
For a further walkthrough of a minimal viewer see the hello world example. More documentation can be found in our [wiki](https://github.com/mozilla/pdf.js/wiki) too.
For a further walkthrough of a minimal viewer, see the hello world example. More documentation can be found in our [wiki](https://github.com/mozilla/pdf.js/wiki) too.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
