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.
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
- 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.
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.
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).
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.
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.
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 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.)
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.
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.
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.
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.
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.
*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.
- 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.
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).
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.
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`.
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 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.
*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
*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.
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.
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.
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.
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
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).
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.
- 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.
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.
*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.
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.
- 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.
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.
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.
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.)
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.
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.
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.
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`.
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 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.
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.
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.
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.
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.
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.
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 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.
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.
* 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
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.
- 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.
* 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.
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.
*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.
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
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.
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.
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.
- 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.
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".)
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.
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.
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).
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
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.
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.
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.
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 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.
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
