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.
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.
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.
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).
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.
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.
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.
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>
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.
*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.
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.
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.
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.
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.
