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.
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.
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
```
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.)
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).
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.
This file (currently) contains not only DOM-specific helper functions/classes, but is used generally for various helper code relevant for main-thread functionality.
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.
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.
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.
*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.
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.
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.
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.
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 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).
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.
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.
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.
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).
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.
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).
