Unfortunately the work-around implemented in PR 12286 didn't actually work in all cases, please refer to the previous commit messages.
To prevent opening of PDF attachments from being completely broken for some users, we'll simply force-download them for now in MOZCENTRAL-builds to unbreak things. (Given that the "Open with" dialog now features a "Open with Firefox"-option, this is less bad than it previously would've been.)
This should provide better filetype detection when downloading PDF attachments in the viewer.
Also, to avoid creating the "is PDF file" regular expression more than once it's extracted into a global constant instead.
This reverts commit 1e5d4b6a80, since it unfortunately doesn't work in all situations.
Please note that I did *successfully* test the patch in a local Firefox build, obviously with an ad blocker installed.
However, I've now tested the *latest* Nightly-build with my default profile, and unfortunately I can still reproduce the bug there!?
Unfortunately e.g. ad blockers can interfere with `window.open` calls, thus preventing PDF attachments from being opened/viewed. For the MOZCENTRAL-build, we can work-around this problem by using a (hidden) link instead.
Without these changes, clicking on the "Open With Different Viewer"-button on the Firefox fallback bar won't actually do anything and the following is printed in the web-console:
```
Uncaught TypeError: (destructured parameter) is undefined
download resource://pdf.js/web/viewer.js:956
response resource://pdf.js/web/viewer.js:1054
listener resource://pdf.js/web/viewer.js:11891
viewer.js:956:1
```
Furthermore, this patch also fixes `PDFViewerApplication.fallback` to pass in an explicit `sourceEventType` when triggering downloading. While this, on its own, would obviously have been sufficient to fix the bug described above, it seems wrong to outright break backwards compatibility of any older `PDFViewerApplication.download` calls.
Now that the `parse` method is simplified we can inline the `setup`
method in the `parse` method since it's only two lines of code. This
avoids some indirection.
Good form type detection is important to get reliable telemetry and to
only show the fallback bar if a form cannot be filled out by the user.
PDF.js only supports AcroForm data, so XFA data is explicitly unsupported
(tracked in issue #2373). However, the previous form type detection
couldn't separate AcroForm and XFA well enough, causing form type
telemetry to be incorrect sometimes and the fallback bar to be shown for
forms that could in fact be filled out by the user.
The solution in this commit is found by studying the specification and
the form documents that are available to us. In a nutshell the rules are:
- There is XFA data if the `XFA` entry is a non-empty array or stream.
- There is AcroForm data if the `Fields` entry is a non-empty array and
it doesn't consist of only document signatures.
The document signatures part was not handled in the old code, causing a
document with only XFA data to also be marked as having AcroForm data.
Moreover, the old code didn't check all the data types.
Now that AcroForm and XFA can be distinguished, the viewer is configured
to only show the fallback bar for documents that only have XFA data. If
a document also has AcroForm data, the viewer can use that to render the
form. We have not found documents where the XFA data was necessary in
that case.
Finally, we include unit tests to ensure that all cases are covered and
move the form type detection out of the `parse` function so that it's
only executed if the document information is actually requested
(potentially making initial parsing a tiny bit faster).
Not only is `catDict` never accessed anymore outside of this file, it
should also never happen since it's internal to the catalog. If data
from it is needed elsewhere, the catalog should provide a getter for it
that can do basic data integrity checks and abstract away any
unnecessary details.
The `AcroForm` entry is part of the catalog, not of the document, so its
logic should be placed there instead. The document should look in the
catalog to fetch it, and not have knowledge of `catDict`, which is a
member internal to the catalog.
Moreover, make the AcroForm member private on the document instance. It's
only used internally and was also never intended to be public. For users
it's exposed by the `getMetadata` API endpoint as `IsAcroFormPresent`.
Only a boolean is exposed, so we now also only store the boolean on the
document instance.
Finally, the annotation code needs access to the full AcroForm
dictionary, so it's updated to fetch the data from the catalog instead
of the document that now only holds the boolean.
The `Collection` entry is part of the catalog, not of the document, so
its logic should be placed there instead. The document should look in the
catalog to fetch it, and not have knowledge of `catDict`, which is a
member internal to the catalog.
Moreover, remove the collection member from the document instance. It's
only used internally and was also never intended to be public. For users
it's exposed by the `getMetadata` API endpoint as `IsCollectionPresent`.
Moving this out of the `parse` function makes sure that the getter is
only executed if the document information is actually requested
(potentially making initial parsing a tiny bit faster).
The `Version` entry is part of the catalog, not of the document, so its
logic should be placed there instead. The document should look in the
catalog to fetch it, and not have knowledge of `catDict`, which is a
member internal to the catalog.
Moreover, make the version member private on the document instance. It's
only used internally and was also never intended to be public. For users
it's exposed by the `getMetadata` API endpoint as `PDFFormatVersion`.
Finally, clarify how the version from the header and the version from
the catalog are treated using a comment.
Even though the code obviously works as-is, given that we have unit-tests for it, it still feels incorrect to just *assume* that the `Catalog`-instance has all of its properties immediately available. Especially when (almost) all of the other handlers, in `src/core/worker.js`, protect their data accesses with appropriate `pdfManager.ensure` calls.
Even though the code obviously works as-is, given that we have unit-tests for it, it still feels incorrect to just *assume* that the `XRef`-instance has all of its properties immediately available. Especially when (almost) all of the other handlers, in `src/core/worker.js`, protect their data accesses with appropriate `pdfManager.ensure` calls.
In `display/canvas.js` the accent offsets must be multiplied by `fontSize` to make the offsets large enough. Another problem is in `core/type1_parser.js` when the Type1 command `seac` is handled. There is an error in the Adobe Type1 spec. See chapter 6 in Type1 Font Format Supplement, which provides an errata: The arguments of `seac` specify the offset of the left side bearing (LSB) points, not the offset of origins. This can be fixed in `core/type1_parser.js` by adding the difference of the LSB values.
Currently there's enough leading padding that the `numPages` span feels somewhat "disconnected" from the `pageNumber` input, which seems unfortunate when they contain related state.
In addition to the unit tests these reference tests make sure that this
document, that triggered some edge cases in our code, can be rendered
and printed successfully now.
The down appearance (`D`) is optional and not available in the document
from #12233, so the checkboxes are never saved/printed as checked
because the checked appearance is based on the export value that is
missing because the `D` entry is not available.
Instead, we should use the normal appearance (`N`) since that one is
required and therefore always available.
Finally, the /Off appearance is optional according to section 12.7.4.2.3
of the specification, so that needs to be taken into account to match
the specification and to fix reference test failures for the
`annotation-button-widget-print` test. That is a file that doesn't
specify an /Off appearance in the normal appearance dictionary.
The helper method `_decodeFormValue` is used to ensure that it happens
in one place. Note that form values are field values, display values
and export values.
Obviously it doesn't make sense to call that method without providing an `AnnotationStorage`-instance, however we should ensure that doing so won't cause errors.
Hence we need to check that `annotationStorage` is actually defined, before attempting to call its `resetModified` method.
This solution is obviously *not* perfect, since printing being cancelled will thus remove the warning as well. However, a similar problem already exists for saving, since the user may cancel that one as well.
All-in-all, since way cannot really detect with absolute certainty that either saving or printing actually finished, this seems good enough for now.
This commit makes the following improvements:
- The code is similar to the other interactive form widgets now, with a
clear note for the only difference.
- Calling `getOrCreateValue` unconditionally ensures that choice widgets
always have a value in the annotation storage. Previously we only
inserted a value in the annotation storage when an option matched or
when a selection was changed. However, this causes breakage when
saving/printing because comboboxes, which we don't fully support yet
but are rendered, might not have a value in storage at all. Their
field value might not match any option since it allows the user to
enter a custom value.
- Calling `getOrCreateValue` unconditionally ensures that forms with
choice widgets no longer always trigger a warning when the user
navigates away from the page. This fixes
https://github.com/mozilla/pdf.js/pull/12241#discussion_r474279654
Given that `renderInteractiveForms` is now enabled by default in "full" viewer, it seems reasonable to enable it by default in the viewer components as well.
Especially considering that it's simple to disable, when creating the affected components, for anyone implementing their own viewer.
The export value is used when the document is saved, so it should also
be used when the document is opened to determine which choice widget
option is selected. The display value is, as the name implies, only to
be used for viewer display purposes and not for other logic.
This makes sure that in the document from #12233 the "Favourite colour"
choice widget is correctly initialized with "Red" instead of "Black"
because the field value is equal to the export value (always the case),
but not the display value (not always the case). Moreover, saving now
also correctly uses the export value and not the display value.
Related to https://bugzilla.mozilla.org/show_bug.cgi?id=1659753
This allows Firefox trigger a "save" event from ctrl/cmd+s or the "Save
Page As" context menu, which in turn lets pdf.js generate a new PDF if
there is form data to save.
I also now use `sourceEventType` on downloads so Firefox can determine if
it should launch the "open with" dialog or "save as" dialog.
With the changes in PR 11077, the zoom dropdown now looks "squashed" in locales with longer than average zoom-strings[1]. The reason is that the zoom-value and the dropdown-icon are too close together, which doesn't look good in affected locales.
To fix this, the following changes are made:
- Increase the calculated dropdown width, in `Toolbar._adjustScaleWidth`, to account for the much wider icon (7 px -> 16 px) and the increased padding.
- Move the dropdown-icon *slightly* outwards, and also *slightly* reduce the left (right in RTL locales) padding of the dropdown-contents.
- Finally, remove the right (left in RTL locales) padding to reduce the chance of the *default* browser dropdown-icon being visible.
---
[1] This affects e.g. the `de` and `nl` locales, but there's probably other examples as well.
The specification states that the field value is `null` if no item is
selected and we didn't handle this case properly. Even though this did
not break the rendering because we always convert the value to an array
and the `includes` check in the display layer would simply not match,
the field value would be `[null]` which is not expected and strange from
an API perspective.
This commit fixes that by ensuring that we return an empty array in
case the field value is `null`. The API therefore still always gives an
array for the field value, but now the code is more specific so that the
value is either an empty array or an array of strings.
This commit follows the same pattern as another unit test in this file
and both reduces existing and future code duplication (since the next
commit will extend this test with an additional input).
This commit includes support for rendering pages in printing mode,
which, when combined with annotation storage data, is useful for testing
if form data is correctly rendered onto the printed canvas.
This is needed for some smoke tests in mozilla central for testing forms
in pdf.js.
Note: AnnotationLayerBuilder.render() doesn't really need to be async, but
we're talking of making the annotation's render functions async, so this
will make that switch easier.
Note that this will increase the run-time of `gulp dist` and `gulp dist-install`, but that's unavoidable given that there's now additional building happening.
This is *similar* to the existing transfer function support for SMasks, but extended to simple image data.
Please note that the extra amount of data now being sent to the worker-thread, for affected /ExtGState entries, is limited to *at most* 4 `Uint8Array`s each with a length of 256 elements.
Refer to https://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#G9.1658137 for additional details.
I completely overlooked the fact that `PartialEvaluator.handleSetFont` also updates the current `state`, which means that currently we're not actually handling font data correctly for cached /ExtGState data. (Thankfully, using /ExtGState to set a font is somewhat rare in practice.)
Currently using a touchscreen with pdf.js doesn't work so well. In Firefox,
with apz.allow_zooming = false (default on current release/beta), it does a
reflow zoom which makes the UI elements bigger. And with apz.allow_zooming = true
(default on current Firefox nightly), or in Chrome, it does a smooth pinch-zoom
but that also scales up the entire UI. Neither of these is a particularly good
experience, so this patch just disables any multi-touch gestures. Touch-based
panning (which involves a single touch point) is left unaffected.
I obviously missed this during review, but currently `PDFViewerApplication._saveInProgress` is reset *synchronously* in `PDFViewerApplication.save`.
That was probably not intended, since it essentially renders the `PDFViewerApplication._saveInProgress` check pointless given that the actual saving is an *asynchronous* operation.
While this will obviously increase the size of the output of `gulp minified`/`gulp minified-es5` *slightly*, the resulting files are still a lot smaller than the non-minified builds.
See https://github.com/terser/terser#minify-options for information about various Terser options.
The original code would get a long sequence of miniscule "tick" values while
pinch-zooming, and each tick value would cause a 1.1x zoom. So even the smallest
pinch gesture on a trackpad would cause high amounts of zoom. This patch
accumulates the wheel deltas until they reach an integer threshold (with a
tweak of the scaling factor to make it feel more natural) at which point it
triggers the zoom based on the integer component of the accumulated delta. The
fractional part is retained in the accumulator.
Some fonts have loca tables that aren't sorted or use 0 as an offset to
signal a missing glyph. This fixes the bad loca tables by sorting them
and then rewriting the loca table and potentially re-ordering the glyf
table to match.
Fixes#11131 and bug 1650302.
- Initialize the `AnnotationStorage`-instance, on `PDFDocumentProxy`, lazily.
- Change the `AnnotationStorage` to use a `Map` internally, rather than a regular Object (simplifies the following points).
- Let `AnnotationStorage.getAll` return `null` when there's no data stored, to avoid unnecessary parsing on the worker-thread. This ought to "just work", since the worker-thread code *should* already handle the `!annotationStorage` case everywhere.
- Add a new `AnnotationStorage.size` getter, to be able to easily tell if there's any data stored.
Prior to PR 11601, the `disableCreateObjectURL` option was present on `getDocument` in the API, since it was (potentially) used when decoding JPEG images natively in the browser. Hence setting this option, which was done automatically using compatibility-code, were in some browsers necessary in order for e.g. JPEG images to be correctly rendered.
The downside of the `disableCreateObjectURL` option is that memory usage increases significantly, since we're forced to build and use `data:` URIs (rather than `blob:` URLs).
However, at this point in time the `disableCreateObjectURL` option is only necessary for *some* (non-essential) functionality in the default viewer; in particular:
- The openfile functionality, used only when manually opening a new file in the default viewer.
- The download functionality, used when downloading either the PDF document itself or its attached files (if such exists).
- The print functionality, in the generic `PDFPrintService` implementation.
Hence neither the general PDF.js library, nor the *basic* functionality of the default viewer, depends on the `disableCreateObjectURL` option any more; which is why I'm thus proposing that we remove the option since using it is a performance footgun.
*Please note:* To not outright break currently "supported" browsers, which lack proper `URL.createObjectURL` support, this patch purposely keeps the compatibility-code to explicitly disable `URL.createObjectURL` usage *only* for browsers which are known to not work correctly.[1]
While it's certainly possible that there's additional, likely older, browsers with broken `URL.createObjectURL` support, the last time that these types of problems were reported was over *three* years ago.[2]
Hence in the *very* unlikely event that additional problems occur, as a result of these changes, we can either add a new case in the compatibility-code or simply declare the affected browser as unsupported.
---
[1] Which are IE11 (see issue 3977), and Google Chrome on iOS (see PR 8081).
[2] Given that `URL.createObjectURL` is used by default, you'd really expect more reports if these problems were widespread.
Issue 4398 was fixed by PR 4437, however a test-case wasn't included as far as I can tell. Given that PR 12186 is now in the process of re-factoring that code, adding a test-case cannot hurt as far as I'm concerned.
With recent changes, these event handlers are now essentially identical. Hence a new helper function is added, to reduce unnecessary duplication (will also be helpful with upcoming changes).
These two classes are unsurprisingly quite similar, and with upcoming changes[1] the amount of (essentially) duplicated code will increase even further.
Notable changes:
- Collect shared functionality in the `BaseTreeViewer` class, reducing both current and future code-duplication.
- Reduce unnecessary duplication in the CSS rules, which will be particularly useful with upcoming changes.
- Tweak the attachmentsView to use links, rather than buttons, to simplify (primarily) the CSS rules.
---
[1] Once API support for "Optional Content" lands, I've got more-or-less finished patches to add viewer support as well.
While the parameter name (clearly) suggests that an `AnnotationStorage`-instance is expected, looking at the only call-sites that include the parameter (i.e. the `PDFPrintServiceFactory` instances) it actually contains just a normal Object.
Hence it seems much more reasonable to actually pass a valid `AnnotationStorage`-instance, as the name suggests, and simply have `PDFPageProxy.render` do the `annotationStorage.getAll()` call. (Since we cannot send an `AnnotationStorage`-instance as-is to the worker-thread, given the "structured clone algorithm".)
Do the `AppOptions.get("printResolution")` lookup once in `web/app.js `, when initializing `PDFPrintServiceFactory`-instances, rather than for every printed page
- Fix the `gulp types` task to run on Windows. Currently this fails, and the solution was to "borrow" the same formatting as used in the `gulp jsdoc` task.
- Place the TypeScript definitions in their own `types` directory, when building `pdfjs-dist`. These should *not* be cluttering the main `build` directory, especially since the generated TypeScript definitions consists of *multiple folders*. (Only if the TypeScript definitions would be concatenated into *a single file*, would placing them directly in `pdfjs-dist/build` be acceptable.)
Over time we used multiple different formats for JSDoc comments. This
commit standardizes those formats to the one we used most often.
Moreover, this removes the example in the outline endpoint documentation
since it now has a proper type definition and it didn't render correctly
in JSDoc.
This commit:
- formats the documentation block according to the standards;
- replaces the callback definitions with the `function` type (we have
that for other definitions already and the callback type was not
rendered correctly by JSDoc);
- synchronizes the type documentation and the class documentation;
- fixes the documentation by making it easier to read and making sure
that all optional properties are indicated as such;
- uses the `@link` tag to indicate links to other code.
The `typestest` still passes and JSDoc now renders this class correctly.
This commit:
- moves the preparation work to a new `typestest-pre` target similar to
how the other targets work;
- moves the `TYPESTEST_DIR` definition to the top of the file like we
did for all other directory variables;
- renames the `TYPES_BUILD_DIR` variable to `TYPES_DIR` since it's
shorter and the naming scheme then corresponds to the other directory
variables;
- switches to `const`/template strings in the types targets where needed;
- converts the `if (err !== null)` check to `if (err)` similar to other
targets.
This patch should *hopefully* remove the intermittent unit-test failure, by using the *same* `optionalContentConfigPromise` for both `renderTask`s and thus get more predictable timing behaviour.
Add a new method to the API to get the optional content configuration. Add
a new render task param that accepts the above configuration.
For now, the optional content is not controllable by the user in
the viewer, but renders with the default configuration in the PDF.
All of the test files added exhibit different uses of optional content.
Fixes#269.
Fix test to work with optional content.
- Change the stopAtErrors test to ensure the operator list has something,
instead of asserting the exact number of operators.
Since the attachment fetching/parsing is already asynchronous, possibly delaying the dispatching of an "attachmentsloaded" event should thus not be a problem in general.
Note that in some cases, i.e. PDF documents with no "regular" attachments and only FileAttachment annotations, we'll thus no longer dispatch an "attachmentsloaded" event when `attachmentsCount === 0` and instead wait for the FileAttachment parsing to finish. (The use of a timeout still guarantees that an "attachmentsloaded" event is *eventually* dispatched though.)
This patch *considerably* simplifies the "attachmentsloaded" event handler, in `PDFSidebar`, since the details are now abstracted away from the consumer.
Finally, add a check in `_appendAttachment` to ensure (indirectly) that the FileAttachment annotation is actually relevant for the active document.
This approach is already used in other parts of the code-base, see e.g. `PDFOutlineViewer`, and has the advantage of only invalidating the DOM once rather than for every attachment item.
This is possible now that most intermittent unit test failures have been
resolved by other patches. There is only one remaining, but it's very
rare and doesn't have to block this update anymore.
This might make debugging intermittent failures a bit easier in the
future because it allows us to spot unexpected differences in the number
of tests being run and allows us to run the tests locally in the same
order in case of intermittent failures.
This reverts commit 50f73092e1. This
causes an inconsistency with the integrated find bar that should be
discussed more before moving on with this (refer to PR #12141).
These errors can/will occur if data is still loading when the document is destroyed, which is the case in the API unit-tests that load the `tracemonkey.pdf` file.
While this patch prevents these kind of problems, and thus allows us to update Jasmine again, I cannot help but thinking that it's slightly "hacky". Basically, we'll simply catch and ignore (some) rejected promises once the document is destroyed and/or its data loading is aborted. However, I don't *think* that these changes should cause issues in general, since we don't really care about errors once document destruction has started (note e.g. the fair number of `catch` handlers ignoring `AbortException`s already).
*The [api-minor] label probably ought to have been added to the original PR, given the changes to the `createAnnotationLayerBuilder` signature (if nothing else).*
This patch fixes the following things:
- Let the `AnnotationLayer.render` method create an `AnnotationStorage`-instance if none was provided, thus making the parameter *properly* optional. This not only fixes the reference tests, it also prevents issues when the viewer components are used.
- Stop exporting `AnnotationStorage` in the official API, i.e. the `src/pdf.js` file, since it's no longer necessary given the change above. Generally speaking, unless absolutely necessary we probably shouldn't export unused things in the API.
- Fix a number of JSDocs `typedef`s, in `src/display/` and `web/` code, to actually account for the new `annotationStorage` parameter.
- Update `web/interfaces.js` to account for the changes in `createAnnotationLayerBuilder`.
- Initialize the storage, in `AnnotationStorage`, using `Object.create(null)` rather than `{}` (which is the PDF.js default).
The f1040-annotations reftest started hanging after #12107. We traced
this to `TypeError: can't access property "getOrCreateValue", storage is
undefined`.
We essentially need to add `annotationStorage` to the parameters in
test/driver.js.
This PR adds typescript definitions from the JSDoc already present.
It adds a new gulp-target 'types' that calls 'tsc', the typescript
compiler, to create the definitions.
To use the definitions, users can simply do the following:
```
import {getDocument, GlobalWorkerOptions} from "pdfjs-dist";
import pdfjsWorker from "pdfjs-dist/build/pdf.worker.entry";
GlobalWorkerOptions.workerSrc = pdfjsWorker;
const pdf = await getDocument("file:///some.pdf").promise;
```
Co-authored-by: @oBusk
Co-authored-by: @tamuratak
These changes improves the existing search functionality triggered when
the URL contains a `#search` hash.
In addition to performing the actual search immediately like before,
this will ensure the search text field inside the find bar gets
populated with the text currently being searched for.
Besides making general sense, as far as I can tell, this patch should also prevent *one* source of `Uncaught (in promise) ...` exceptions.
Unfortunately `reason instanceof AbortError` doesn't work here, since `AbortError` isn't actually defined in browsers; note how even the DOM specification contains an example using the `name` property: https://dom.spec.whatwg.org/#aborting-ongoing-activities
This patch prevents the following errors from being logged in the console, when the unit-tests are running:
- Firefox: `Uncaught (in promise) DOMException: The operation was aborted.`
- Chrome: `Uncaught (in promise) DOMException: The user aborted a request.`
Looking carefully at this code, you'll notice that the `loadDocument` function has no less than *three* Promise handling functions. This obviously makes no sense, since a Promise can only have one resolve and one reject handler.
Hence the final `onFailure`-case is unreachable, which only serves to add confusion when reading the code. Note that this code has been re-factored more than once over the years, but it seems as if this may even have been incorrect already in PR 3310 (and no-one have noticed for seven years :-).
There's quite frankly no particular reason to special-case Type3-fonts with image resources, which are very rare anyway, now that we have a general mechanism for sending/receiving images globally.
While the `CharProcs` streams of Type3-fonts *usually* don't rely on dependencies, such as e.g. images, it does happen in some cases.
Currently any dependencies are simply appended to the parent operatorList, which in practice means *only* the operatorList of the *first* page where the Type3-font is being used.
However, there's one thing that's slightly unfortunate with that approach: Since fonts are global to the PDF document, we really ought to ensure that any Type3 dependencies are appended to the operatorList of *all* pages where the Type3-font is being used. Otherwise there's a theoretical risk that, if one page has its rendering paused, another page may try to use a Type3-font whose dependencies are not yet fully resolved. In that case there would be errors, since Type3 operatorLists are executed synchronously.
Hence this patch, which ensures that all relevant pages will have Type3 dependencies appended to the main operatorList. (Note here that the `OperatorList.addDependencies` method, via `OperatorList.addDependency`, ensures that a dependency is only added *once* to any operatorList.)
Finally, these changes also remove the need for the "waiting for the main-thread"-hack that was added to `PartialEvaluator.buildPaintImageXObject` as part of fixing issue 10717.
Jasmine >= 3.6.0 causes intermittent test failures because of random
task abortions.
Puppeteer >= 4.0.0 causes ENOTEMPTY/EBUSY errors during shutdown on the
Windows bot.
Moreover, `jasmine-core` is a dependency of `jasmine` so it doesn't have
to be required separately.
This patch should *hopefully* remove the `Unhandled promise rejection: ...` errors, by returning the "final" promise. Also, by pausing/delaying of rendering slightly the likelihood of the test failing in the first place should thus be reduced.
This uses a similar format to the reference-test logging, and will help determine in *exactly* which browser the failure occurred (since the tests run concurrently).
- Given the `DefaultExternalServices` implementation, the `PDFViewerApplication.supportsDocumentFonts` getter is guaranteed to be defined and we can thus remove some (now) unnecessary `PDFJSDev` checks from the `webViewerInitialized` function.
- By slightly tweaking the "pdfBugEnabled" definition in `web/app_options`, similar to the existing ones for "workerSrc" and "cMapUrl", we can remove some `PDFJSDev` checks from the `PDFViewerApplication._parseHashParameters` method.
The current behavior for `getPagesOverview` assumes we want to only
auto-rotate if:
- `enablePrintAutoRotate` is `true`
- `isFirstPagePortrait !== isPortraitOrientation(size)`
This second check is what is breaking #9297. The two PDFs linked have a
landscape orientation first page, as well as subsequent pages. Since
`false === false`, we print portrait.
Let's drop the comparison with `isFirstPagePortrait`, and print
landscape if `!isPortraitOrientation(size)`.
Fixes#9297.
When the old `Dict.getAll()` method was removed, it was replaced with a `Dict.getKeys()` call and `Dict.get(...)` calls (in a loop).
While this pattern obviously makes a lot of sense in many cases, there's some instances where we actually want the *raw* `Dict` values (i.e. `Ref`s where applicable). In those cases, `Dict.getRaw(...)` calls are instead used within the loop. However, by introducing a new `Dict.getRawValues()` method we can reduce the number of (strictly unnecessary) function calls by simply getting the *raw* `Dict` values directly.
Using a `Map` instead of an `Object` provides some advantages such as
cheaper ways to get the size of the cache, to find out if an entry is
contained in the cache and to iterate over the cache. Moreover, we can
clear and re-use the same `Map` object now instead of creating a new
one.
Since this method calls `Dict.get` to fetch data, there could thus be `Error`s thrown in corrupt PDF documents when attempting to resolve an indirect object.
To ensure that this won't ever become a problem, we change the method to be `async` such that a rejected Promise would be returned and general OperatorList parsing won't break.
- Replace the existing loops with `for...of` variants instead.
- Make use of `continue`, to reduce indentation and to make the code (slightly) easier to follow, when checking `/Resources` entries.
This case should no longer happen, given the `instanceof Ref` branch just above (added in PR 6971).
Also, I've run the entire test-suite locally with `continue` replaced by `throw new Error(...)` and didn't find any problems.
Given that this method is used during what's essentially a *pre*-parsing stage, before the actual OperatorList parsing occurs, on second thought it doesn't seem at all necessary to warn and trigger fallback in cases where there's lookup errors.
*Please note:* Any any errors will still be either suppressed or thrown, according to the `ignoreErrors` option, during the *actual* OperatorList parsing.
It turns out that `getTextContent` suffers from *similar* problems with repeated GStates as `getOperatorList`; please see the previous patch.
While only `/ExtGState` resources containing Fonts will actually be *parsed* by `PartialEvaluator.getTextContent`, we're still forced to fetch/validate repeated `/ExtGState` resources even though *most* of them won't affect the textContent (since they mostly contain purely graphical state).
With these changes we also no longer need to immediately reset the current text-state when encountering a `setGState` operator, which may thus improve text-selection in some cases.
This patch will help pathological cases the most, with issue 2813 being a particularily problematic example. While there's only *four* `/ExtGState` resources, there's a total `29062` of `setGState` operators. Even though parsing of a single `/ExtGState` resource is quite fast, having to re-parse them thousands of times does add up quite significantly.
For simplicity we'll only cache "simple" `/ExtGState` resource, since e.g. the general `SMask` case cannot be easily cached (without re-factoring other code, which may have undesirable effects on general parsing).
By caching "simple" `/ExtGState` resource, we thus improve performance by:
- Not having to fetch/validate/parse the same `/ExtGState` data over and over.
- Handling of repeated `setGState` operators becomes *synchronous* during the `OperatorList` building, instead of having to defer to the event-loop/microtask-queue since the `/ExtGState` parsing is done asynchronously.
---
Obviously I had intended to include (standard) benchmark results with this patch, but for reasons I don't understand the test run-time (even with `master`) of the document in issue 2813 is *a lot* slower than in the development viewer (making normal benchmarking infeasible).
However, testing this manually in the development viewer (using `pdfBug=Stats`) shows a *reduction* of `~10 %` in the rendering time of the PDF document in issue 2813.
Stop special-casing the (very unlikely) "no `/XObject` found"-scenario, when parsing `OPS.paintXObject` operators, in `PartialEvaluator.{getOperatorList, getTextContent}`
Originally there weren't any (generally) good ways to handle errors gracefully, on the worker-side, however that's no longer the case and we can simply fallback to the existing `ignoreErrors` functionality instead.
Also, please note that the "no `/XObject` found"-scenario should be *extremely* unlikely in practice and would only occur in corrupt/broken documents.
Note that the `PartialEvaluator.getOperatorList` case is especially bad currently, since we'll simply (attempt to) send the data as-is to the main-thread. This is quite bad, since in a corrupt/broken document the data *could* contain anything and e.g. be unclonable (which would cause breaking errors).
Also, we're (obviously) not attempting to do anything with this "raw" `OPS.paintXObject` data on the main-thread and simply ensuring that we never send it definately seems like the correct approach.
This special-case was added in PR 1992, however it became unnecessary with the changes in PR 4824 since all of the ColorSpace parsing is now done on the worker-thread (with only RGB-data being sent to the main-thread).
Originally ColorSpaces were only *partially* parsed on the worker-thread, to obtain an IR-format which was sent to the main-thread. This had the somewhat unfortunate side-effect of causing the majority of the (potentially heavy) ColorSpace parsing to happen on the main-thread.
Hence PR 4824 which, among other things, changed ColorSpaces to be *fully* parsed on the worker-thread with only RGB-data being sent to the main-thread.
While it thus originally was necessary to have `ColorSpace.{parseToIR, fromIR}` methods, to handle the worker/main-thread split, that's no longer the case and we can thus reduce all of the ColorSpace parsing to one method instead.
Currently, when parsing a ColorSpace, we call `ColorSpace.parseToIR` which parses the ColorSpace-data from the document and then creates the IR-format. We then, immediately, call `ColorSpace.fromIR` which parses the IR-format and then finally creates the actual ColorSpace.[1]
All-in-all, this leads to a fair amount of unnecessary indirection which also (in my opinion) makes the code less clear.
Obviously these changes are not really expected to have a significant effect on performance, especially with the recently added caching of ColorSpaces, however there'll now be strictly fewer function calls, less memory allocated, and overall less parsing required during ColorSpace-handling.
---
[1] For ICCBased ColorSpaces, given the validation necessary, this currently even leads to parsing an /Alternate ColorSpace *twice*.
Given the dummy-methods on `DefaultExternalServices`, there's no longer any compelling reason not to (attempt to) report telemetry unconditionally.
The only larger change consists of moving the `KNOWN_VERSIONS` and `KNOWN_GENERATORS` arrays ouf of the `PDFViewerApplication._initializeMetadata` method.
*Please note:* Most of this patch consists of whitespace-only changes.
There's a few things that could be improved in the current implementation, such as:
- It's currently necessary to *both* manually track the `featureId`s which should trigger delayed fallback, as well as manually report telemetry in affected cases.
Obviously there's only two call-sites as of now (forms and javaScript), but it still feels somewhat error-prone especially if more cases were to be added in the future. To address this, this patch adds a new (private) method which abstracts away these details from the call-sites.
- Generally, it also seems nice to reduce *and* simplify the amount of state we need to store on `PDFViewerApplication` in order to support the "delayedFallback" functionality.
Also, having to *manually* work with the "delayedFallback"-array in multiple places feels cumbersome and makes e.g. the `PDFViewerApplication.fallback` method less clear as to its behaviour.
- Having code *outside* of `PDFViewerApplication`, i.e. in the event handlers, directly access properties which are marked as "private" via a leading underscore doesn't seem that great in general.
Furthermore, having the event handlers directly deal with that should be internal state also seem unfortunate. To address this, the patch will instead make use of a new `PDFViewerApplication.triggerDelayedFallback` callback.
- There's at least one code-path in the viewer, see `PDFViewerApplication.error`, where `fallback` can be called without an argument.
It's currently possible (although maybe somewhat unlikely) that such a call *could* be overridden by the `featureId` of a pending "delayedFallback" call, thus not reporting the *correct* fallback reason.
- The "delayedFallback"-state weren't being reset on document close (which shouldn't affect Firefox, but nonetheless it ought to be fixed).
Note how the `getFontID`-method in `src/core/fonts.js` is *completely* global, rather than properly tied to the current document. This means that if you repeatedly open and parse/render, and then close, even the *same* PDF document the `fontID`s will still be incremented continuously.
For comparison the `createObjId` method, on `idFactory`, will always create a *consistent* id, assuming of course that the document and its pages are parsed/rendered in the same order.
In order to address this inconsistency, it thus seems reasonable to add a new `createFontId` method on the `idFactory` and use that when obtaining `fontID`s. (When the current `getFontID` method was added the `idFactory` didn't actually exist yet, which explains why the code looks the way it does.)
*Please note:* Since the document id is (still) part of the `loadedName`, it's thus not possible for different documents to have identical font names.
This removes additional `// eslint-disable-next-line no-shadow` usage, which our old pseudo-classes necessitated.
Most of the re-formatting changes, after the `class` definitions and methods were fixed, were done automatically by Prettier.
*Please note:* I'm purposely not doing any `var` to `let`/`const` conversion here, since it's generally better to (if possible) do that automatically on e.g. a directory basis instead.
This will simplify the `class` conversion in the next patch, and with modern JavaScript the moved code is still limited to the current module scope.
*Please note:* For improved consistency with our usual formatting, the `TILING_PATTERN`/`SHADING_PATTERN` constants where re-factored slightly.
This removes additional `// eslint-disable-next-line no-shadow` usage, which our old pseudo-classes necessitated.
*Please note:* I'm purposely not doing any `var` to `let`/`const` conversion here, since it's generally better to (if possible) do that automatically on e.g. a directory basis instead.
*Yet another instalment in the never-ending series of things that you think of __after__ a patch has landed.*
Since `Function`s are only cached by reference, we thus don't need to allocate storage for names in `LocalFunctionCache` instances. Obviously the effect of these changes are *really tiny*, but it seems reasonable in principle to avoid allocating data structures that are guaranteed to be unused.
Note that compared other structures, such as e.g. Images and ColorSpaces, `Function`s are not referred to by name, which however does bring the advantage of being able to share the cache for an *entire* page.
Furthermore, similar to ColorSpaces, the parsing of individual `Function`s are generally fast enough to not really warrant trying to cache them in any "smarter" way than by reference. (Hence trying to do caching similar to e.g. Fonts would most likely be a losing proposition, given the amount of data lookup/parsing that'd be required.)
Originally I tried implementing this similar to e.g. the recently added ColorSpace caching (and in a couple of different ways), however it unfortunately turned out to be quite ugly/unwieldy given the sheer number of functions/methods where you'd thus need to pass in a `LocalFunctionCache` instance. (Also, the affected functions/methods didn't exactly have short signatures as-is.)
After going back and forth on this for a while it seemed to me that the simplest, or least "invasive" if you will, solution would be if each `PartialEvaluator` instance had its *own* `PDFFunctionFactory` instance (since the latter is already passed to all of the required code). This way each `PDFFunctionFactory` instances could have a local `Function` cache, without it being necessary to provide a `LocalFunctionCache` instance manually at every `PDFFunctionFactory.{create, createFromArray}` call-site.
Obviously, with this patch, there's now (potentially) more `PDFFunctionFactory` instances than before when the entire document shared just one. However, each such instance is really quite small and it's also tied to a `PartialEvaluator` instance and those are *not* kept alive and/or cached. To reduce the impact of these changes, I've tried to make as many of these structures as possible *lazily initialized*, specifically:
- The `PDFFunctionFactory`, on `PartialEvaluator` instances, since not all kinds of general parsing actually requires it. For example: `getTextContent` calls won't cause any `Function` to be parsed, and even some `getOperatorList` calls won't trigger `Function` parsing (if a page contains e.g. no Patterns or "complex" ColorSpaces).
- The `LocalFunctionCache`, on `PDFFunctionFactory` instances, since only certain parsing requires it. Generally speaking, only e.g. Patterns, "complex" ColorSpaces, and/or (some) SoftMasks will trigger any `Function` parsing.
To put these changes into perspective, when loading/rendering all (14) pages of the default `tracemonkey.pdf` file there's now a total of 6 `PDFFunctionFactory` and 1 `LocalFunctionCache` instances created thanks to the lazy initialization.
(If you instead would keep the document-"global" `PDFFunctionFactory` instance and pass around `LocalFunctionCache` instances everywhere, the numbers for the `tracemonkey.pdf` file would be instead be something like 1 `PDFFunctionFactory` and 6 `LocalFunctionCache` instances.)
All-in-all, I thus don't think that the `PDFFunctionFactory` changes should be generally problematic.
With these changes, we can also modify (some) call-sites to pass in a `Reference` rather than the actual `Function` data. This is nice since `Function`s can also be `Streams`, which are not cached on the `XRef` instance (given their potential size), and this way we can avoid unnecessary lookups and thus save some additional time/resources.
Obviously I had intended to include (standard) benchmark results with these changes, but for reasons I don't really understand the test run-time (even with `master`) of the document in issue 2541 is quite a bit slower than in the development viewer.
However, logging the time it takes for the relevant `PDFFunctionFactory`/`PDFFunction ` parsing shows that it takes *approximately* `0.5 ms` for the `Function` in question. Looking up a cached `Function`, on the other hand, is *one order of magnitude faster* which does add up when the same `Function` is invoked close to 2000 times.
This moves, and slightly simplifies, code that's currently residing in the unit-test utils into the actual library, such that it's bundled with `GENERIC`-builds and used in e.g. the API-code.
As an added bonus, this also brings out-of-the-box support for CMaps in e.g. the Node.js examples.
The default viewer, and thus Firefox, depends on the `RenderTask.onContinue` functionality to pause/continue rendering (such that the most visible page always renders first).
Despite this functionality thus being very important, it has however never actually been tested *at all* as far as I can tell. Hence this patch which adds a new boolean `renderTaskOnContinue` parameter (`false` by default), that can be used to force a reference-test to use the `RenderTask.onContinue` code-path in the `InternalRenderTask` class.
Note that I purposely made this new reference-test behaviour *optional*, since I didn't want to negatively affect the general runtime of the tests (given that there's a slight delay added to the rendering). Also, for e.g. benchmarking you'd most likely want to stay away from the `RenderTask.onContinue` functionality for similar reasons.
This should reduce the possibility of accidentally truncating some inline images, while *not* causing the "EI" detection to become significantly slower.[1]
There's obviously a possibility that these added checks are not sufficient to catch *every* single case of "EI" sequences within the actual inline image data, but without specific test-cases I decided against over-engineering the solution here.
*Please note:* The interpolation issues are somewhat orthogonal to the main issue here, which is the truncated image, and it's already tracked elsewhere.
---
[1] I've looked at the issue a few times, and this is the first approach that I was able to come up with that didn't cause *unacceptable* performance regressions in e.g. issue 2618.
Versions of `needle` prior to `2.5.0` cannot cope with redirects (as documented: https://github.com/tomas/needle/issues/312).
This prevents prebuilt `canvas` binaries from being downloaded on MacOS,
requiring the global install of its dependencies.
Updating Needle restores it to functionality, addressing this. It also avoids
the need to add `request` to `package.json`; it also obsoletes https://github.com/mozilla/pdf.js/pull/12018.
With the changes made in PR 9659, `ColorSpace.fromIR` no longer takes a second `pdfFunctionFactory` parameter and there's thus one call-site that can be simplified.
This patch contains the following *notable* improvements:
- Changes the `ColorSpace.parse` call-sites to, where possible, pass in a reference rather than actual ColorSpace data (necessary for the next point).
- Adds (local) caching of `ColorSpace`s by `Ref`, when applicable, in addition the caching by name. This (generally) improves `ColorSpace` caching for e.g. the SMask code-paths.
- Extends the (local) `ColorSpace` caching to also apply when handling Images and Patterns, thus further reducing unneeded re-parsing.
- Adds a new `ColorSpace.parseAsync` method, almost identical to the existing `ColorSpace.parse` one, but returning a Promise instead (this simplifies some code in the `PartialEvaluator`).
This will allow caching of ColorSpaces by either `Name` *or* `Ref`, which doesn't really make sense for images, thus allowing (better) caching for ColorSpaces used with e.g. Images and Patterns.
Defining this *inline* in the "constructor" looks slightly weird (I really don't know why I wrote it like that originally), and it can simply be changed to a regular method instead.
Tweak the `QueueOptimizer` to recognize `OPS.paintImageMaskXObject` operators as *repeated* when the "skew" transformation matrix elements are non-zero (issue 8078)
As can be seen in the code there's a handful of places where this structure needs to be iterated, something that becomes cumbersome when dealing with `Object`s. Hence, by changing this to a `Map` instead we can both simplify the code and avoid creating unnecessary closures.
Particularily the `PDFPageProxy._tryCleanup` method becomes a lot more readable, at least in my opinion.
Finally, since this property is intended to be "private" the name is adjusted to reflect that.
This patch aims to simplify the `PDFPageProxy._destroy` method, by:
- Replacing the unnecessary `forEach` with a "regular" `for`-loop instead.
- Use a more appropriate variable name, since `intentState.renderTasks` contain instances of `InternalRenderTask`.
- Move the "is rendering completed"-handling to a new `InternalRenderTask.completed` getter, to abstract away some (mostly) internal `InternalRenderTask` state.
In hindsight, using the `for (let [key, value] of myMap) { ... }`-format when we don't care about the `key` probably wasn't such a great idea. Since `Map`s have explicit support for iterating either `key`s or `value`s, we should probably use that instead here.
*First of all, I should mention that my understanding of the finer details of the `QueueOptimizer` (and its related `CanvasGraphics` methods) is somewhat limited.*
Hence I'm not sure if there's actually a very good reason for *only* considering ImageMasks where the "skew" transformation matrix elements are zero as *repeated*, however simply looking at the code I just don't see why these elements cannot be non-zero as long as they are *all identical* for the ImageMasks.
Furthermore, looking at the *group* case (which is what we're currently falling back to), there's no particular limitation placed upon the transformation matrix elements.
While this patch obviously isn't enough to *completely* fix the issue, since there should be a visible Pattern rendered as well[1], it seem (at least to me) like enough of an improvement that submitting this is justified.
With these changes the referenced PDF document will no longer hang the *entire* browser, and rendering also finishes in a *reasonable* time (< 10 seconds for me) which seem fine given the *huge* number of identical inline images present.[2]
---
[1] Temporarily changing the Pattern to a solid color *does* render the correct/expected area, which suggests that the remaining problem is a pre-existing issue related to the Pattern-handling itself rather than the `QueueOptimizer` functionality.
[2] The document isn't exactly rendered immediately in e.g. Adobe Reader either.
This removes multiple instances of `// eslint-disable-next-line no-shadow`, which our old pseudo-classes necessitated.
*Please note:* I'm purposely not doing any `var` to `let`/`const` conversion here, since it's generally better to (if possible) do that automatically on e.g. a directory basis instead.
This removes one instance of `// eslint-disable-next-line no-shadow`, which our old pseudo-classes necessitated.
*Please note:* I'm purposely not doing any `var` to `let`/`const` conversion here, since it's generally better to (if possible) do that automatically on e.g. a directory basis instead.
The `isNaN` check is obviously redundant, since `NaN` is the only value that isn't equal to itself; see https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/NaN#Examples
The `parseFloat`/`parseInt` comparison would make sense if the `value` ever contains a String, which however is never actually the case. Besides looking through the code, I've also run the entire test-suite locally with `assert(typeof value === "number", "encodeNumber");` added at the top of the method and there were no failures.
Hence we can simplify the "is integer" check a bit in the `CFFCompiler.encodeNumber` method.
Because of a really stupid `Promise`-related mistake on my part, when re-factoring `PDFImage.buildImage` during the `NativeImageDecoder` removal, we're no longer re-throwing errors occuring during image parsing/decoding as intended.
The result is that some (fairly) corrupt documents will never finish loading, and unfortunately there were apparently no sufficiently corrupt images in the test-suite to catch this.
For now we need to use a Babel-plugin, since Webpack 4.x doesn't seem to support it yet. (Most likely we'll have to update to Webpack 5, once that becomes available, in order for this to be directly supported. This is thus also blocked on removing the `webpack-stream` package.)
While the `??` operator will thus always be transpiled by Babel, even in modern builds, simply supporting it for development purposes seems like a step in the right direction.
https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Operators/Nullish_coalescing_operator
Since this is completely internal functionality, and furthermore limited to the worker-thread, this change should thus not have any observable effect for e.g. an API-user.
Apparently I completely overlooked the fact that with the changes in PR 11069 these properties became *completely* unused, and consequently they thus ought to be removed.
Since there's (essentially) no tests for the SVG-backend, these changes didn't make in into PR 11912 when the code in the `src/display/canvas.js` file was modified.
Since this helper function is no longer used anywhere in the main code-base, but only in a couple of unit-tests, it's thus being moved to a more appropriate spot.
Finally, the implementation of `isEmptyObj` is also tweaked slightly by removing the manual loop.
Compared to regular `Object`s, `Map`s have a number of advantageous properties: Of particular importance in this case is the built-in iteration support, and that determining if the structure is empty is easy.
Compared to regular `Object`s, `Map`s have a number of advantageous properties: Of particular importance in this case is the built-in iteration support, and that determining if the structure is empty is easy.
Compared to regular `Object`s, `Map`s (and `Set`s) have a number of advantageous properties: Of particular importance in this case is the built-in iteration support, and that determining if the structure is empty is easy.
The `RefSet` primitive predates ES6, so that most likely explains why an
object is used internally to track the entries. However, nowadays we can
use built-in JavaScript sets for this purpose. Built-in types are often
more efficient/optimized and using it makes the code a bit more clear
since we don't have to assign `true` to keys anymore just to indicate
their presence.
Both of the removed methods were added in PR 2719, however they are no longer used:
- It appears that `hasPendingRequests` was never used at all, even from the beginning.
- The only general PDF.js library usage of `abortAllRequests` was removed in PR 6879, which is now four years ago. (Originally the Firefox-specific network implementation, see https://searchfox.org/mozilla-central/source/browser/extensions/pdfjs/content/PdfJsNetwork.jsm, was shared with the `src/display/network.js` file and *there* this method is used. However, since all of the Firefox-specific code now lives directly in mozilla-central, that's not relevant for the removal in this patch.)
After PRs 10727 and 11912, the code responsible for sending the decoded image data to the main-thread has now become a fair bit more involved the previously.
To reduce the amount of duplication here, the actual code responsible for sending the data is thus extracted into a new helper method instead.
Avoid calling Math.pow if possible when calculating the transfer
function of the CalRGB color space since calling Math.pow is expensive.
If the value of color is larger than the threshold, 0.99554525,
the final result of the transform is larger that 254.5
since ((1 + 0.055) * 0.99554525 ** (1 / 2.4) - 0.055) * 255 === 254.50000003134699
In the old code the use of an Array meant that we had to *manually* track the `numChunksLoaded` property, given that simply using the Array `length` wouldn't have worked since there's no guarantee that the data is loaded in order when e.g. range requests are in use.
Tracking closely related state *separately* in this manner never seem like a good idea, and we can now instead utilize a Set to avoid that.
On ISO/IEC 10918-6:2013 (E), section 6.1: (http://www.itu.int/rec/T-REC-T.872-201206-I/en)
"Images encoded with three components are assumed to be RGB data encoded as YCbCr unless the image contains an APP14 marker segment as specified in 6.5.3, in which case the colour encoding is considered either RGB or YCbCr according to the application data of the APP14 marker segment"
But common jpeg libraries consider RGB too if components index are ASCII R (0x52), G (0x47) and B (0x42): https://stackoverflow.com/questions/50798014/determining-color-space-for-jpeg/50861048
Issue #11931
Using `require.resolve("worker-loader")` to check if `worker-loader` is installed causes webpack to include `worker-loader` in the output bundle, which is not the intended effect. Aside from increasing the bundle size unnecessarily, it also causes errors for webpack configs with targets that don't have node's built-in modules.
These errors can be fixed by configuring webpack `externals` to exclude `worker-loader`, but it's more difficult to figure out this solution than to figure out that `worker-loader` needs to be installed (even without this explicit error message).
To solve this, the explicit check for `worker-loader` has been removed. An alternative solution would be to use webpack's `resolveWeak`. Documentation has also been added in `examples/webpack` to help users.
Since *inline* images, i.e. those defined inside of `/Contents` streams, are by their very definition page-specific it thus seem like a good idea to actually enforce that they won't accidentally end up in the `GlobalImageCache`.
When converting this file to use standard `import`/`export` statements, I sorted the exports in the same order as the imports to simplify things.
However, looking at the list of `export`ed properties it probably doesn't hurt to add a couple of comments to clarify from where specifically the `export`s originated.
By updating to the new major version of Acorn, we'll get support for newer ECMAScript features as they become available (although some features are currently also blocked by ESLint support and/or SystemJS usage).
Please see https://github.com/acornjs/acorn/releases/tag/7.2.0 for details.
It turns out that `getTextContent` suffers from *similar* problems with repeated images as `getOperatorList`; please see the previous patch.
While only `/XObject` resources of the `Form`-type will actually be *parsed* in `PartialEvaluator.getTextContent`, since those are the only ones that may contain text, we're still forced to fetch repeated image resources where the name differs (but not the reference).
Obviously it's less bad in this case, since we're not actually parsing `/XObject`s of e.g. the `Image`-type. However, you still want to avoid even fetching the data whenever possible, since `Stream`s are not cached on the `XRef` instance (given their potential size) and the lookup can thus be somewhat expensive in general.
To address these issues, we can simply replace the exiting name-only caching in `PartialEvaluator.getTextContent` with a new cache backed by `LocalImageCache` instead.
Currently the local `imageCache`, as used in `PartialEvaluator.getOperatorList`, will miss certain cases of repeated images because the caching is *only* done by name (usually using a format such as e.g. "Im0", "Im1", ...).
However, in some PDF documents the `/XObject` dictionaries many contain hundreds (or even thousands) of distinctly named images, despite them referring to only a handful of actual image objects (via the XRef table).
With these changes we'll now cache *local* images using both name and (where applicable) reference, thus improving re-usage of images resources even further.
This patch was tested using the PDF file from [bug 857031](https://bugzilla.mozilla.org/show_bug.cgi?id=857031), i.e. https://bug857031.bmoattachments.org/attachment.cgi?id=732270, with the following manifest file:
```
[
{ "id": "bug857031",
"file": "../web/pdfs/bug857031.pdf",
"md5": "",
"rounds": 250,
"lastPage": 1,
"type": "eq"
}
]
```
which gave the following results when comparing this patch against the `master` branch:
```
-- Grouped By browser, page, stat --
browser | page | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ---- | ------------ | ----- | ------------ | ----------- | --- | ----- | -------------
firefox | 0 | Overall | 250 | 2749 | 2656 | -93 | -3.38 | faster
firefox | 0 | Page Request | 250 | 3 | 4 | 1 | 50.14 | slower
firefox | 0 | Rendering | 250 | 2746 | 2652 | -94 | -3.44 | faster
```
While this is certainly an improvement, since we now avoid re-parsing ~1000 images on the first page, all of the image resources are small enough that the total rendering time doesn't improve that much in this particular case.
In pathological cases, such as e.g. the PDF document in issue 4958, the improvements with this patch can be very significant. Looking for example at page 2, from issue 4958, the rendering time drops from ~60 seconds with `master` to ~30 seconds with this patch (obviously still slow, but it really showcases the potential of this patch nicely).
Finally, note that there's also potential for additional improvements by re-using `LocalImageCache` instances for e.g. /XObject data of the `Form`-type. However, given that recent changes in this area I purposely didn't want to complicate *this* patch more than necessary.
In the PDF file from the issue below, the fill alpha (`ca`) is set
before drawing the circles using the `setGState` operator. Doing so
causes the global alpha to be set on the canvas' context for the canvas
back-end, but this was not handled in the SVG back-end. This patch fixes
that by taking the fill opacity into account when drawing shading
patterns in the same way as done elsewhere so it is only included if the
value is non-default.
Fixes#11812.
When "Cleanup" is triggered, you obviously need to remove all globally cached data on *both* the main- and worker-threads.
However, the current the implementation of the `GlobalImageCache.clear` method also means that we lose *all* information about which images were cached and not just their data. This thus has the somewhat unfortunate side-effect of requiring images, which were previously known to be "global", to *again* having to reach `NUM_PAGES_THRESHOLD` before being cached again.
To avoid doing unnecessary parsing after "Cleanup", we can thus let `GlobalImageCache.clear` keep track of which images were cached while still removing their actual data. This should not have any significant impact on memory usage, since the only extra thing being kept is a `RefSetCache` (essentially an Object) with a couple of `Set`s containing only integers.
With the changes made in the previous patch, the `web/app_options.js` file no longer depends on anything *except* files residing in the `web/` folder. Hence the `gulp default_preferences` task can now be further simplified and thus becomes even faster than before; see also PR 11724.
With the changes in previous patches, the `disableCreateObjectURL` option/functionality is no longer used for anything in the API and/or in the Worker code.
Note however that there's some functionality, mainly related to file loading/downloading, in the GENERIC version of the default viewer which still depends on this option.
Hence the `disableCreateObjectURL` option (and related compatibility code) is moved into the viewer, see e.g. `web/app_options.js`, such that it's still available in the default viewer.
Currently some JPEG images are decoded by the built-in PDF.js decoder in `src/core/jpg.js`, while others attempt to use the browser JPEG decoder. This inconsistency seem unfortunate for a number of reasons:
- It adds, compared to the other image formats supported in the PDF specification, a fair amount of code/complexity to the image handling in the PDF.js library.
- The PDF specification support JPEG images with features, e.g. certain ColorSpaces, that browsers are unable to decode natively. Hence, determining if a JPEG image is possible to decode natively in the browser require a non-trivial amount of parsing. In particular, we're parsing (part of) the raw JPEG data to extract certain marker data and we also need to parse the ColorSpace for the JPEG image.
- While some JPEG images may, for all intents and purposes, appear to be natively supported there's still cases where the browser may fail to decode some JPEG images. In order to support those cases, we've had to implement a fallback to the PDF.js JPEG decoder if there's any issues during the native decoding. This also means that it's no longer possible to simply send the JPEG image to the main-thread and continue parsing, but you now need to actually wait for the main-thread to indicate success/failure first.
In practice this means that there's a code-path where the worker-thread is forced to wait for the main-thread, while the reverse should *always* be the case.
- The native decoding, for anything except the *simplest* of JPEG images, result in increased peak memory usage because there's a handful of short-lived copies of the JPEG data (see PR 11707).
Furthermore this also leads to data being *parsed* on the main-thread, rather than the worker-thread, which you usually want to avoid for e.g. performance and UI-reponsiveness reasons.
- Not all environments, e.g. Node.js, fully support native JPEG decoding. This has, historically, lead to some issues and support requests.
- Different browsers may use different JPEG decoders, possibly leading to images being rendered slightly differently depending on the platform/browser where the PDF.js library is used.
Originally the implementation in `src/core/jpg.js` were unable to handle all of the JPEG images in the test-suite, but over the last couple of years I've fixed (hopefully) all of those issues.
At this point in time, there's two kinds of failure with this patch:
- Changes which are basically imperceivable to the naked eye, where some pixels in the images are essentially off-by-one (in all components), which could probably be attributed to things such as different rounding behaviour in the browser/PDF.js JPEG decoder.
This type of "failure" accounts for the *vast* majority of the total number of changes in the reference tests.
- Changes where the JPEG images now looks *ever so slightly* blurrier than with the native browser decoder. For quite some time I've just assumed that this pointed to a general deficiency in the `src/core/jpg.js` implementation, however I've discovered when comparing two viewers side-by-side that the differences vanish at higher zoom levels (usually around 200% is enough).
Basically if you disable [this downscaling in canvas.js](https://github.com/mozilla/pdf.js/blob/8fb82e939cf0c8618a4e775ff17fc96f726872b5/src/display/canvas.js#L2356-L2395), which is what happens when zooming in, the differences simply vanish!
Hence I'm pretty satisfied that there's no significant problems with the `src/core/jpg.js` implementation, and the problems are rather tied to the general quality of the downscaling algorithm used. It could even be seen as a positive that *all* images now share the same downscaling behaviour, since this actually fixes one old bug; see issue 7041.
Currently image resources, as opposed to e.g. font resources, are handled exclusively on a page-specific basis. Generally speaking this makes sense, since pages are separate from each other, however there's PDF documents where many (or even all) pages actually references exactly the same image resources (through the XRef table). Hence, in some cases, we're decoding the *same* images over and over for every page which is obviously slow and wasting both CPU and memory resources better used elsewhere.[1]
Obviously we cannot simply treat all image resources as-if they're used throughout the entire PDF document, since that would end up increasing memory usage too much.[2]
However, by introducing a `GlobalImageCache` in the worker we can track image resources that appear on more than one page. Hence we can switch image resources from being page-specific to being document-specific, once the image resource has been seen on more than a certain number of pages.
In many cases, such as e.g. the referenced issue, this patch will thus lead to reduced memory usage for image resources. Scrolling through all pages of the document, there's now only a few main-thread copies of the same image data, as opposed to one for each rendered page (i.e. there could theoretically be *twenty* copies of the image data).
While this obviously benefit both CPU and memory usage in this case, for *very* large image data this patch *may* possibly increase persistent main-thread memory usage a tiny bit. Thus to avoid negatively affecting memory usage too much in general, particularly on the main-thread, the `GlobalImageCache` will *only* cache a certain number of image resources at the document level and simply fallback to the default behaviour.
Unfortunately the asynchronous nature of the code, with ranged/streamed loading of data, actually makes all of this much more complicated than if all data could be assumed to be immediately available.[3]
*Please note:* The patch will lead to *small* movement in some existing test-cases, since we're now using the built-in PDF.js JPEG decoder more. This was done in order to simplify the overall implementation, especially on the main-thread, by limiting it to only the `OPS.paintImageXObject` operator.
---
[1] There's e.g. PDF documents that use the same image as background on all pages.
[2] Given that data stored in the `commonObjs`, on the main-thread, are only cleared manually through `PDFDocumentProxy.cleanup`. This as opposed to data stored in the `objs` of each page, which is automatically removed when the page is cleaned-up e.g. by being evicted from the cache in the default viewer.
[3] If the latter case were true, we could simply check for repeat images *before* parsing started and thus avoid handling *any* duplicate image resources.
With these changes SystemJS is now only used, during development, on the worker-thread and in the unit/font-tests, since Firefox is currently missing support for worker modules; please see https://bugzilla.mozilla.org/show_bug.cgi?id=1247687
Hence all the JavaScript files in the `web/` and `src/display/` folders are now loaded *natively* by the browser (during development) using standard `import` statements/calls, thanks to a nice `import-maps` polyfill.
*Please note:* As soon as https://bugzilla.mozilla.org/show_bug.cgi?id=1247687 is fixed in Firefox, we should be able to remove all traces of SystemJS and thus finally be able to use every possible modern JavaScript feature.
As part of reducing our reliance on SystemJS in the development viewer, this patch replaces usage of `require` statements with modern standards `import`/`export` statements instead.
If we want to try and move forward with reducing usage of SystemJS, we don't have much choice but to make these kind changes (despite what prior test-results showed, however I'm no longer able to reproduce the issues locally).
As part of trying to reduce the usage of SystemJS in the development viewer, this patch is a necessary step that will allow removal of some `require` statements.
Currently this uses `SystemJS.import` in non-PRODUCTION mode, but it should be possible to replace those with standard *dynamic* `import` calls in the future.
Given that the `getPDFFileNameFromURL` helper function has a specific code-path for handling non-string inputs, this empty string fallback really isn't necessary at the call-site in `web/pdf_document_properties.js`.
Apparently the old link format used in MOZCENTRAL-builds, with the blob URL separated from the filename with a `?` character violates the specification; see https://bugzilla.mozilla.org/show_bug.cgi?id=1632644#c5
Obviously just removing the `?`-part of the URL would have worked, but that would also have meant that we'd no longer be able to provide the correct filename when the user attempts to download the opened PDF attachment.
To fix this we'll instead append the filename in the hash-part of the URL, which however required using a *custom* hash-parameter to avoid triggering the fallback "named destination" code-paths in the viewer.
Note that only changing the `web/pdf_attachment_viewer.js` file wasn't sufficient to fix the bug, and we also need to tweak the `webViewerInitialized` function in `web/app.js` since MOZCENTRAL-builds used to ignore *everything* in the URL hash.
This particular code is very old, but changing it *should* be completely safe given that the `PDFViewerApplication.setTitleUsingUrl` method since some time now stores both the original URL (in `this.url`) as well as one without the hash (in `this.baseUrl`). The latter one is already used everywhere where it matters, so this change seem fine to me.
This patch thus restores the original behaviour for PDF attachments in the MOZCENTRAL-build, by once again allowing them to be opened *directly* in the browser without downloading. (The fallback added in PR 11845 is obviously kept, since it seems generally useful to have.)
Originally the `default_preferences.json` file was checked into the repository, and we thus needed to load it in non-PRODUCTION mode (which was originally done asynchronously using `XMLHttpRequest`). Over the years a lot has changed and the `default_preferences.json` file is now built, by the `gulp default_preferences` task, from the `web/app_options.js` file. Hence it's no longer necessary, in non-PRODUCTION mode, to use SystemJS here since we can simply use a standard `import` statement instead.
Note how e.g. `web/app.js` already imports from `web/app_options.js` in the same exact way that `web/preferences.js` now does, hence this patch will *not* result in any significant changes in the built/bundled viewer file.
This is another (small) part in trying to reduce usage of SystemJS, with the goal of hopefully getting rid of it completely. (I've started working on this, and doing so has identified a number of problem areas; this patch addresses one of them.)
This replaces some additional `require`/`exports` usage with standard `import`/`export` statements instead.
Hence another, small, part in the effort to reduce the reliance on SystemJS-specific functionality in the development viewer.
By re-factoring the existing gulp tasks, most of the code can be re-used for both the existing `gulp minified` as well as the new `gulp minified-es5` task.
While working on PR 11872, it occurred to me that it probably wouldn't be a bad idea to change the `_parsedAnnotations` getter to handle errors individually for each annotation. This way, one broken/corrupt annotation won't prevent the rest of them from being e.g. fetched through the API.
Given that the `PDFLinkService.setHash` method itself if completely synchronous, moving the handling of "nameddest" to occur last *shouldn't* cause any problems (famous last words).
This way the destination will still override any previous parameter, such as e.g. the "page", as expected. Furthermore, given that the `PDFLinkService.navigateTo` method is asynchronous that should provide additional guarantees that the "nameddest" parameter is always respected.
As sort-of expected, this fairly innocent looking change also required some tweaks in the `PDFHistory` to prevent dummy history entires upon document load (only an issue when both "page" *and* "nameddest" parameters are provided in the hash).
This rule complements the existing `accessor-pairs` nicely, and ensures that a getter/setter pair is always consistently ordered.
Please find additional details about this rule at https://eslint.org/docs/rules/grouped-accessor-pairs
Having `assert` calls without a message string isn't very helpful when debugging, and it turns out that it's easy enough to make use of ESLint to enforce better `assert` call-sites.
In a couple of cases the `assert` calls were changed to "regular" throwing of errors instead, since that seemed more appropriate.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/no-restricted-syntax
This should ensure that a page will always render successfully, even if there's errors during the Annotation fetching/parsing.
Additionally the `OperatorList.addOpList` method is also adjusted to ignore invalid data, to make it slightly more robust.
All of the other *similar* helper functions only target one file per function, and there's no particular reason for this one to be different.
This patch will simplify future changes, e.g. experimenting with using `gulp watch` instead of SystemJS for the development viewer.
Not only is there no code depending on it now, the actual task itself doesn't even make sense as-is. Note that it uses the default `DEFINES` configuration *unaltered*, which is neither useful nor correct since the resulting build thus won't make sense without an actual built target set.
Given that none of the PDF.js contributors know React, maintaining and/or providing supporting for the example isn't really feasible unfortunately.
Even something as simple as running/testing the example becomes difficult for anyone completely unfamiliar with React, and furthermore:
- It's very difficult to tell if the example demonstrates React best-practices, since the PDF.js contributors don't know React.
- We also have no reasonable way of keeping the example up-to-date with changes in React.
- The React example, in its current form, is even *hard-coding* the PDF.js version to a now unsupported version.
- The example is currently triggering "fake worker" usage, see issue 11729, which is really *really* bad. Note that the "fake worker" functionality is *only* intended as a fallback, and it should absolutely *not* under any circumstances be advertised and certainly shouldn't be triggered in official PDF.js examples.
By preserving the exception type, more fine-grained error handling can be performed via client-side logic (e.g. redirect to a search page if a PDF is not found, or to a ticket system in case of invalid PDF files).
The original exception is now re-thrown.
Fixes#11658
We already rendered the name for radio buttons, but it was missing for
all other interactive form elements. This commit adds that so that
values entered in form elements can be read based on the element name.
To save time or resources during development it can be useful to run
tests only in Firefox. Previously this could be done by editing the
browser manifest file, but since that file is no longer used for
Puppeteer, this command line flag replaces it. For example, executing
`gulp unittest --noChrome` will only run the unit tests in Firefox.
This commit replaces our own infrastructure for handling browsers during
tests with Puppeteer. Using our own infrastructure for this had a few
downsides:
- It has proven to not always be reliable, especially when closing the
browser, causing failures on the bots because browsers were still
running even though they should have been stopped. Puppeteer should do
a better job with this because it uses the browser's test built-in
instrumentation tools for this (the devtools protocol) which our code
didn't. This also means that we don't have to pass
parameters/preferences to tweak browser behavior anymore.
- It requires the browsers under test to be installed on the system,
whereas Puppeteer downloads the browsers before the test. This means
that setup is much easier (no more manual installations and browser
manifest files) as well as testing with different browser versions
(since they can be provisioned on demand). Moreover, this ensures that
contributors always run the tests in both Firefox and Chrome,
regardless of which browsers they have installed locally.
- It's all code we have to maintain, so Puppeteer abstracts away how the
browsers start/stop for us so we don't have to keep that code.
By default, Puppeteer only installs one browser during installation,
hence the need for a post-install script to install the second browser.
This requires `cross-env` to make passing the environment variable work
on both Linux and Windows.
The other testing code already uses the name of the browser as the
unique identifier, so I don't see a good reason to not use that for
identifying browsers to quit as well. Doing so simplifies the (already
somewhat complex) testing logic and ensures that we can use existing
functionality (such as the `getSession` function) to retrieve sessions.
Given that the `NativeImageDecoder.{isSupported, isDecodable}` methods require both dictionary lookups *and* ColorSpace parsing, in hindsight it actually seems more reasonable to the `JpegStream.maybeValidDimensions` checks *first*.
This is a simple work-around for https://bugzilla.mozilla.org/show_bug.cgi?id=1632644 which was caused by platform changes in Firefox. Ideally the Firefox bug should still be fixed, but these PDF.js changes seem generally useful to prevent both current and future issues here.
There's no particular reason for using the PDF.js helper function `createObjectURL` here, given that the relevant code-path is already guarded by multiple "disableCreateObjectURL" option checks.
Given that `URL.createObjectURL` is assumed to always be available in MOZCENTRAL builds, note the existing usage in the file, there's no reason to depend on the PDF.js helper function `createObjectURL` at all here.
Furthermore this patch also changes `DownloadManager.downloadData` to actually revoke the `blobUrl` after downloading has completed, which is similar to the existing code in `DownloadManager.download`.
*This patch implements https://github.com/mozilla/pdf.js/pull/11777#issuecomment-609741348*
This extends the work from PR 11773 and 11777 further, by immediately releasing the `font.data` property once the font been attached to the DOM. By not unnecessarily holding onto this data on the main-thread, we'll thus reduce the memory usage of fonts even further (especially beneficial in longer documents with composite fonts).
The new behaviour is controlled by the recently added `fontExtraProperties` API option (adding a new option just for this patch didn't seem necessary), since there's one edge-case in the SVG renderer where the `font.data` property is necessary (see the `pdf2svg` example).
Note that while the default viewer does run clean-up with an idle timeout, that timeout will be reset whenever rendering occurs *or* when scrolling happens in the viewer. In practice this means that unless the user doesn't interact with the viewer in *any* way during an extended period of time, currently set to 30 seconds, the `PDFDocumentProxy.cleanup` method will never be called and font resources will thus not be cleaned-up.
This is necessary in order to support cases where the default viewer is embedded in a *dynamically* created <iframe> element.
In order to also support a use-case where there's *multiple* <iframe> elements (containing default viewers) on the same page, the "webviewerloaded" event now includes a `source` detail parameter such that it's possible to associate an event with the relevant DOM element.
As evident from the code, `PageViewport` only supports[1] `rotation` values which are a multiple of 90 degrees. Besides it being somewhat difficult to imagine meaningful use-cases for a non-multiple of 90 degrees `rotation`, the code also becomes both simpler and more efficient by not having to consider arbitrary `rotation` values.
However, any invalid rotation will *silently* fallback to assume zero `rotation` which probably isn't great for e.g. `PDFPageProxy.getViewport` in the API. Hence this patch, which will now enforce that only valid `rotation` values are accepted.
---
[1] As far as I can tell, from looking through the history, nothing else has ever been supported either.
Hopefully we don't need the *latest* Node.js releases for the unit-tests to work on Travis, and by using "long-term support" releases instead we should be able to avoid these types of sudden failures in the future as well.
Change the "download" keyboard shortcut (Ctrl+S) handling, in GENERIC/CHROME builds, to utilize the `EventBus` (issue 11657); add a new "openfile" keyboard shortcut (Ctrl+O), in GENERIC builds
Somewhat surprisingly, despite the GENERIC viewer implementing "openfile" support, there's never been a keyboard shortcut available. Similar to the previous patch, this utilizes the `EventBus` for consistency with the `Toolbar`/`SecondaryToolbar` buttons.
*Please note:* This patch should NOT be construed as carte blanche to simply convert all of the code in `webViewerKeyDown`, or elsewhere, to make use of the `EventBus` instead of direct function calls.
Any further changes, along the lines in this patch, would need to be evaluated on a case-by-case basis to determine if they are actually wanted, given that many/most existing cases in `webViewerKeyDown` should already be *indirectly* observable through the `EventBus` instance.
This improves the consistency of the "download" handling, in the default viewer, such that the `Toolbar`/`SecondaryToolbar` buttons *and* the keyboard shortcut are now handled in the same way (using the `EventBus`).
Given that the "download" keyboard shortcut handling is limited to GENERIC/CHROME builds and that the issue does raise a valid point about only being able to observe *some* downloads, these changes seem acceptable in this particular case.
Finally the pre-processor condition is adjusted to *explicitly*, rather than implicitly, list the affected build targets.
*Please note:* This patch should NOT be construed as carte blanche to simply convert all of the code in `webViewerKeyDown`, or elsewhere, to make use of the `EventBus` instead of direct function calls.
Any further changes, along the lines in this patch, would need to be evaluated on a case-by-case basis to determine if they are actually wanted, given that many/most existing cases in `webViewerKeyDown` should already be *indirectly* observable through the `EventBus` instance.
These two `AppOptions` are only defined in GENERIC builds, hence it's completely unnecessary to check them in the extension builds (e.g. MOZCENTRAL and CHROME).
Also, simply let the "printResolution" option be defined in all builds since it's being accessed in `web/firefox_print_service.js` as well.
Rather than creating a new `Stream` just to validate the OS/2 TrueType table, it's simpler/better to just pass in a reference to the font data and use that instead (similar to other TrueType helper functions).
There's a handful of cases in the code where the intention is simply to advance the `Stream` position, but rather than only doing that the code instead fetches/computes a Uint16 value (and without using the result for anything).
There's a handful of cases in the code where the intention is simply to advance the `Stream` position, but rather than only doing that the code instead fetches the bytes in question (and without using the result for anything).
*Please note:* These changes were done automatically, using the `gulp lint --fix` command.
This rule is already enabled in mozilla-central, see https://searchfox.org/mozilla-central/rev/567b68b8ff4b6d607ba34a6f1926873d21a7b4d7/tools/lint/eslint/eslint-plugin-mozilla/lib/configs/recommended.js#103-104
The main advantage, besides improved consistency, of this rule is that it reduces the size of the code (by 3 bytes for each case). In the PDF.js code-base there's close to 8000 instances being fixed by the `dot-notation` ESLint rule, which end up reducing the size of even the *built* files significantly; the total size of the `gulp mozcentral` build target changes from `3 247 456` to `3 224 278` bytes, which is a *reduction* of `23 178` bytes (or ~0.7%) for a completely mechanical change.
A large number of these changes affect the (large) lookup tables used on the worker-thread, but given that they are still initialized lazily I don't *think* that the new formatting this patch introduces should undo any of the improvements from PR 6915.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/dot-notation
What the user did:
Open the PDF Viewer in Chrome;
Mouse click into the “Page number” input field;
What they saw:
A pop-up list with seemingly random numbers
What you were expecting to see:
Nothing
What they see is the Chrome “Autofill” feature at work – that is suggesting values that you have previously entered into number fields in forms, as possible values you may want to enter into this field. The list has nothing to do with the PDF currently open but the user does not know this.
- Add a reduced test-case for issue 11768, to prevent future regressions.
(Given that PR 11769 is only a work-around, rather than a proper solution, it may not be entirely accurate for the issue to be closed as fixed.)
- Add more validation of the charCode, as found by the heuristics, in `PartialEvaluator._buildSimpleFontToUnicode` to prevent future issues.
Some of the code in `src/core/jpg.js` is fairly old, and has with time become unnecessary when the surrounding code has been updated to handle various types of JPEG corruption.
In particular the `if (!marker || marker <= 0xff00) { ... }` branch is now dead code, since:
- The `!marker` case can no longer happen, since we would already have broken out of the loop thanks to the `!fileMarker` branch a handful of lines above.
- The `marker <= 0xff00` case can also no longer happen, since the `findNextFileMarker` function validate markers much more thoroughly (by checking `marker >= 0xffc0 && marker <= 0xfffe`). Hence we'd again have broken out of the loop via the `!fileMarker` branch above when no valid marker was found.
This patch fixes yet another instalment in the never-ending series of "what the *bleep* was I thinking", by changing the `PDFDocumentProxy.getViewerPreferences` method to return `null` by default.
Not only is this method now consistent with many other API methods, for the data not present case, but it also avoids having to e.g. loop through an object to check if it's actually empty (note the old unit-test).
Please note that these changes were done automatically, using `gulp lint --fix`.
Given that the major version number was increased, there's a fair number of (primarily whitespace) changes; please see https://prettier.io/blog/2020/03/21/2.0.0.html
In order to reduce the size of these changes somewhat, this patch maintains the old "arrowParens" style for now (once mozilla-central updates Prettier we can simply choose the same formatting, assuming it will differ here).
- Since the Gitpod contributing work-flow is, from the PDF.js project perspective, completely unsupported we don't want to mention it in such a highly visible way as the README file.
- Since Gitpod is a commercial service, we probably want to avoid *potentially* being seen as advertising and/or endorsing it by mentioning it (somewhat prominently) in the PDF.js README file.
- By leaving the actual Gitpod files in the repository, for now, this should thus avoid outright breaking things for any existing users.
With two kind of builds now being produced, with/without translation/polyfills, it's unfortunately somewhat easy for users to accidentally pick the wrong one.
In the case where a user would attempt to use a modern build of PDF.js in an older browser, such as e.g. IE11, the failure would be immediate when the code is loaded (given the use of unsupported ECMAScript features).
However in some browsers/environments, a modern PDF.js build may load correctly and thus *appear* to function, only to fail for e.g. certain API calls. To hopefully lessen the support burden, and to try and improve things overall, this patch adds additional checks to ensure that a modern build of PDF.js cannot be used in browsers/environments which lack native support for `Promise.allSettled`.[1] Hence we'll fail early, with an error message telling users to pick an ES5-compatible build instead.
*Please note:* While it's probably too early to tell if this will be a widespread issue, it's possible that this is the sort of patch that *may* warrant being `git cherry-pick`ed onto the current beta version (v2.4.456).
---
[1] This was a fairly recent addition to the web platform, see https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Promise/allSettled#Browser_compatibility
*Please note:* Most of the necessary API work was done in PR 10033, and the only remaining thing to do here was to implement it in the viewer.
The new preference should thus allow e.g. enterprise users to disable copying in the viewer, for PDF documents whose permissions specify that.
In order to simplify things the "copy"-permission was implemented using CSS, as suggested in https://bugzilla.mozilla.org/show_bug.cgi?id=792816#c55, which should hopefully suffice.[1]
The advantage of this approach, as opposed to e.g. disabling the `textLayer` completely, is first of all that it ensures that searching still works correctly even in copy-protected documents. Secondly this also greatly simplifies the overall implementation, since it doesn't require a lot of code for something that's disabled by default.
---
[1] As the discussion in the bug shows, this kind of copy-protection is not very strong and is also generally easy to remove/circumvent in various ways. Hence a simple solution, targeting "regular"-users rather than "power"-users is hopefully deemed acceptable here.
At this point in time, compared to when the "ignore single-char" code was added, we *should* generally be doing a much better job of combining text into as few chunks as possible.
However, there's still bad cases where we're not able to combine text as much as one would like, which is why I'm *not* proposing to simply measure/scale all text. Instead this patch will to only measure/scale single-char text in cases where the horizontal/vertical scale is off significantly, since that's were you'd expect bad text-selection behaviour otherwise.
Note that most of the movement caused by this patch is with Type3 fonts, which is a somewhat special font type and one where our current text-selection behaviour is probably the least good.
[api-minor] Change `Font.exportData` to, by default, stop exporting properties which are completely unused on the main-thread and/or in the API (PR 11773 follow-up)
For years now, the `Font.exportData` method has (because of its previous implementation) been exporting many properties despite them being completely unused on the main-thread and/or in the API.
This is unfortunate, since among those properties there's a number of potentially very large data-structures, containing e.g. Arrays and Objects, which thus have to be first structured cloned and then stored on the main-thread.
With the changes in this patch, we'll thus by default save memory for *every* `Font` instance created (there can be a lot in longer documents). The memory savings obviously depends a lot on the actual font data, but some approximate figures are: For non-embedded fonts it can save a couple of kilobytes, for simple embedded fonts a handful of kilobytes, and for composite fonts the size of this auxiliary can even be larger than the actual font program itself.
All-in-all, there's no good reason to keep exporting these properties by default when they're unused. However, since we cannot be sure that every property is unused in custom implementations of the PDF.js library, this patch adds a new `getDocument` option (named `fontExtraProperties`) that still allows access to the following properties:
- "cMap": An internal data structure, only used with composite fonts and never really intended to be exposed on the main-thread and/or in the API.
Note also that the `CMap`/`IdentityCMap` classes are a lot more complex than simple Objects, but only their "internal" properties survive the structured cloning used to send data to the main-thread. Given that CMaps can often be *very* large, not exporting them can also save a fair bit of memory.
- "defaultEncoding": An internal property used with simple fonts, and used when building the glyph mapping on the worker-thread. Considering how complex that topic is, and given that not all font types are handled identically, exposing this on the main-thread and/or in the API most likely isn't useful.
- "differences": An internal property used with simple fonts, and used when building the glyph mapping on the worker-thread. Considering how complex that topic is, and given that not all font types are handled identically, exposing this on the main-thread and/or in the API most likely isn't useful.
- "isSymbolicFont": An internal property, used during font parsing and building of the glyph mapping on the worker-thread.
- "seacMap": An internal map, only potentially used with *some* Type1/CFF fonts and never intended to be exposed in the API. The existing `Font.{charToGlyph, charToGlyphs}` functionality already takes this data into account when handling text.
- "toFontChar": The glyph map, necessary for mapping characters to glyphs in the font, which is built upon the various encoding information contained in the font dictionary and/or font program. This is not directly used on the main-thread and/or in the API.
- "toUnicode": The unicode map, necessary for text-extraction to work correctly, which is built upon the ToUnicode/CMap information contained in the font dictionary, but not directly used on the main-thread and/or in the API.
- "vmetrics": An array of width data used with fonts which are composite *and* vertical, but not directly used on the main-thread and/or in the API.
- "widths": An array of width data used with most fonts, but not directly used on the main-thread and/or in the API.
By slicing the Uint8Array directly, rather than using the prototype and a `call` invocation, the runtime of `decryptAscii` is decreased slightly (~30% based on quick logging).
The `decryptAscii` function is still less efficient than `decrypt`, however ASCII encoded Type1 font programs are sufficiently rare that it probably doesn't matter much (we've only seen *two* examples, issue 4630 and 11740).
The PDF document, in the referenced issue, actually contains ASCII-encoded Type1 data which we currently *incorrectly* identify as binary.
According to the specification, see https://www-cdf.fnal.gov/offline/PostScript/T1_SPEC.PDF#[{%22num%22%3A203%2C%22gen%22%3A0}%2C{%22name%22%3A%22XYZ%22}%2C87%2C452%2Cnull], the current checks are insufficient to decide between binary/ASCII encoded Type1 font programs.
In preparation for the next patch, this changes the signature of `TranslatedFont` to take an object rather than individual parameters. This also, in my opinion, makes the call-sites easier to read since it essentially provides a small bit of documentation of the arguments.
Finally, since it was necessary to touch `TranslatedFont` anyway it seemed like a good idea to also convert it to a proper `class`.
Given that the `vertical` property is always accessed on the main-thread, ensuring that the property is explicitly defined seems like the correct thing to do since it also avoids boolean casting elsewhere in the code-base.
With `Font.exportData` now only exporting white-listed properties, there should no longer be any reason to not use standard shadowing in the `Font.spaceWidth` method.
Furthermore, considering the amount of other changes to the code-base over the years it's not even clear to me that the special-case was necessary any more (regardless of the preceding patches).
A number of *internal* font properties, which only make sense on the worker-thread, were previously exported. Some of these properties could also contain potentially large Arrays/Objects, which thus unnecessarily increases memory usage since we're forced to copy these to the main-thread and also store them there.
This patch stops exporting the following font properties:
- "_shadowWidth": An internal property, which was never intended to be exported.
- "charsCache": An internal cache, which was never intended to be exported and doesn't make any sense on the main-thread. Furthermore, by the time `Font.exportData` is called it's usually `undefined` or a mostly empty Object as well.
- "cidEncoding": An internal property used with (some) composite fonts.
As can be seen in the `PartialEvaluator.translateFont` method, `cidEncoding` will only be assigned a value when the font dictionary has an "Encoding" entry which is a `Name` (and not in the `Stream` case, since those obviously cannot be cloned).
All-in-all this property doesn't really make sense on the main-thread and/or in the API, and note also that the resulting `cMap` property is (partially) available already.
- "fallbackToUnicode": An internal map, part of the heuristics used to improve text-selection in (some) badly generated PDF documents with simple fonts. This was never intended to be exposed on the main-thread and/or in the API.
- "glyphCache": An internal cache, which was never intended to be exported and which doesn't make any sense on the main-thread. Furthermore, by the time `Font.exportData` is called it's usually a mostly empty Object as well.
- "isOpenType": An internal property, used only during font parsing on the worker-thread. In the *very* unlikely event that an API consumer actually needs that information, then `fontType` should be a (generally) much better property to use.
Finally, in the (hopefully) unlikely event that any of these properties become necessary on the main-thread, re-adding them to the white-list is easy to do.
This patch addresses an existing, and very long standing, TODO in the code such that it's no longer possible to send arbitrary/unnecessary font properties to the main-thread.
Furthermore, by having a white-list it's also very easy to see *exactly* which font properties are being exported.
Please note that in its current form, the list of exported properties contains *every* possible enumerable property that may exist in a `Font` instance.
In practice no single font will contain *all* of these properties, and e.g. embedded/non-embedded/Type3 fonts will all differ slightly with respect to what properties are being defined. Hence why only explicitly set properties are included in the exported data, to avoid half of them being `undefined`, which however should not be a problem for any existing consumer (since they'd already need to handle those cases).
Since a fair number of these font properties are completely *internal* functionality, and doesn't make any sense to expose on the main-thread and/or in the API, follow-up patch(es) will be required to trim down the list. (I purposely included all properties here for brevity and future documentation purposes.)
- Use template strings when printing document/viewer information in `_initializeMetadata`, since the old format feels overly verbose.
Also, get the WebGL state from the `BaseViewer` instance[1] rather than the `AppOptions`. Since the `AppOptions` value could theoretically have been changed (by the user) after the viewer components were initialized, it seems much more useful to print the *actual* value that'll be used during rendering.
- Change `_initializePdfHistory` to actually do the "is embedded"-check first, in accordance with the comment and given that the "disableHistory" option usually shouldn't be set.
---
[1] Admittedly reaching into the `BaseViewer` instance and just grabbing the value perhaps isn't a great approach overall, but given that the WebGL-backend isn't even on by default this probably doesn't matter too much.
Over time, with more and more API-functionality added, the `PDFViewerApplication.load` method has become quite large and complex. In an attempt to improve the current situation somewhat, this patch moves the fetching and initialization of "auto print" out into its own (private) helper method instead.
Over time, with more and more API-functionality added, the `PDFViewerApplication.load` method has become quite large and complex. In an attempt to improve the current situation somewhat, this patch moves the fetching and initialization of "metadata" out into its own (private) helper method instead.
Over time, with more and more API-functionality added, the `PDFViewerApplication.load` method has become quite large and complex. In an attempt to improve the current situation somewhat, this patch moves the fetching and initialization of "page labels" out into its own (private) helper method instead.
I completely overlooked the fact that we had *one* occurrence of an asynchronous `mkdirp` call in the gulpfile, which thus breaks since the package now uses Promises rather than a callback function; sorry about that!
*Please note:* This patch on its own is *not* sufficient to address the underlying problem in the referenced issue, hence why no test-case is included since the *actual* bug still needs to be fixed.
As can be seen in the specification, https://tc39.es/ecma262/#sec-string.fromcodepoint, `String.fromCodePoint` will throw a RangeError for invalid code points.
In the event that a CMap, in a composite font, contains invalid data and/or we fail to parse it correctly, it's thus possible that the glyph mapping that we build end up with entires that cause `String.fromCodePoint` to throw and thus `Font.charToGlyph` to break.
If that happens, as is the case in issue 11768, significant portions of a page/document may fail to render which seems very unfortunate.
While this patch doesn't fix the underlying problem, it's hopefully deemed useful not only for the referenced issue but also to prevent similar bugs in the future.
With two kind of builds now being produced, with/without translation/polyfills, it's unfortunately somewhat easy for users to accidentally pick the wrong one.
In the case where a user would attempt to use a modern build of PDF.js in an older browser, such as e.g. IE11, the failure would be immediate when the code is loaded (given the use of unsupported ECMAScript features).
However in some browsers/environments, in particular Node.js, a modern PDF.js build may load correctly and thus *appear* to function, only to fail for e.g. certain API calls. To hopefully lessen the support burden, and to try and improve things overall, this patch adds checks to ensure that a modern build of PDF.js cannot be used in browsers/environments which lack native support for critical functionality (such as e.g. `ReadableStream`). Hence we'll fail early, with an error message telling users to pick an ES5-compatible build instead.
To ensure that we actually test things better especially w.r.t. usage of the PDF.js library in Node.js environments, the `gulp npm-test` task as used by Node.js/Travis was changed (back) to test an ES5-compatible build.
(Since the bots still test the code as-is, without transpilation/polyfills, this shouldn't really be a problem as far as I can tell.)
As part of these changes there's now both `gulp lib` and `gulp lib-es5` build targets, similar to e.g. the generic builds, which thanks to some re-factoring only required adding a small amount of code.
*Please note:* While it's probably too early to tell if this will be a widespread issue, it's possible that this is the sort of patch that *may* warrant being `git cherry-pick`ed onto the current beta version (v2.4.456).
This avoids unnecessary duplication of many images, thus reducing the size of PDF.js image resources slightly.
Note that since the images should only be flipped horizontally, this required specifying the horizontal/vertical scaling separately for the hiDPI-images.
This functionality was only added to the default viewer for backwards compatibility and to support the various PDF viewer tests in mozilla-central, with the intention to eventually remove it completely.
While the different mozilla-central tests cannot be *easily* converted from DOM events, it's however possible to limit that functionality to only MOZCENTRAL builds *and* when tests are running.
Rather than depending of the re-dispatching of internal events to the DOM, the default viewer can instead be used in e.g. the following way:
```javascript
document.addEventListener("webviewerloaded", function() {
PDFViewerApplication.initializedPromise.then(function() {
// The viewer has now been initialized, and its properties can be accessed.
PDFViewerApplication.eventBus.on("pagerendered", function(event) {
console.log("Has rendered page number: " + event.pageNumber);
});
});
});
```
The correct/intended way of working with the "viewer components" is by providing an `EventBus` instance upon initialization, and the `getGlobalEventBus` was only added for backwards compatibility.
Note, for example, that using `getGlobalEventBus` doesn't really work at all well with a use-case where there's *multiple* `PDFViewer` instances on a one page, since it may then be difficult/impossible to tell which viewer a particular event originated from.
All of the "viewer components" examples have been previously updated, such that there's no longer any code/examples which relies on the now removed `getGlobalEventBus` functionality.
The `sizes` property doesn't appear to have been used ever since the code was first split into main/worker-threads, which is so many years ago that I wasn't able to easily find exactly in which PR/commit it became unused.
The `encoding` property is always assigned the `properties.baseEncoding` value, however the `PartialEvaluator` doesn't actually compute/set that value any more. Again it was difficult to determine when it became unused, but it's been that way for years.
This looks entirely like something which was left-over from debugging, and that line hasn't been touched since PR 4515, especially considering that the corresponding branch in `FirefoxBrowser` doesn't print anything.
Given the way that "classes" were previously implemented in PDF.js, using regular functions and closures, there's a fair number of false positives when the `no-shadow` ESLint rule was enabled.
Note that while *some* of these `eslint-disable` statements can be removed if/when the relevant code is converted to proper `class`es, we'll probably never be able to get rid of all of them given our naming/coding conventions (however I don't really see this being a problem).
This rule is *not* currently enabled in mozilla-central, but it appears commented out[1] in the ESLint definition file; see https://searchfox.org/mozilla-central/rev/c80fa7258c935223fe319c5345b58eae85d4c6ae/tools/lint/eslint/eslint-plugin-mozilla/lib/configs/recommended.js#238-239
Unfortunately this rule is, for fairly obvious reasons, impossible to `--fix` automatically (even partially) and each case thus required careful manual analysis.
Hence this ESLint rule is, by some margin, probably the most difficult one that we've enabled thus far. However, using this rule does seem like a good idea in general since allowing variable shadowing could lead to subtle (and difficult to find) bugs or at the very least confusing code.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/no-shadow
---
[1] Most likely, a very large number of lint errors have prevented this rule from being enabled thus far.
Fixes#11718 in which the `ff` ligature glyph is at index zero in a CFF font. Beacuse this is a CIDFont, glyph names are CIDs, which are integers. Thus the string `".notdef"` is not correct. The rest of the charset data is already parsed correctly as integers when the boolean argument `cid` is true.
*This is part of a series of patches that will try to split PR 11566 into smaller chunks, to make reviewing more feasible.*
Once all the code has been fixed, we'll be able to eventually enable the ESLint no-shadow rule; see https://eslint.org/docs/rules/no-shadow
*This is part of a series of patches that will try to split PR 11566 into smaller chunks, to make reviewing more feasible.*
Once all the code has been fixed, we'll be able to eventually enable the ESLint no-shadow rule; see https://eslint.org/docs/rules/no-shadow
Please note that the `setPDFNetworkStreamFactory` functionality isn't exposed in the public API, i.e. not listed among the exports in the `src/pdf.js` file, and that even if it were it wouldn't really be useful considering that none of the `PDFNetworkStream`/`PDFFetchStream`/`PDFNodeStream` classes are exported either.
These methods were deprecated already in PDF.js version `2.1.266`, see PRs 10246 and 10369, and were converted to throw `Error`s upon invocation in PDF.js version `2.4.456`, see PR 11219.
Hence it ought to be possible to remove these methods now.
Given that none of the relevant options are marked as optional in the code/JSDocs, and that the `PDFDocumentProperties` class is specific to the default viewer (and not exposed as part of the viewer components), there's no good reason as far as I can tell for these checks.
The `BaseViewer.setDocument` method in particular is necessary for rendering to start when the viewer loads, hence you obviously want that to happen as soon as possible and without any unnecessary delays.
Unfortunately *some* API calls need to be done before that, note existing comments, however the `ViewHistory` initialization (and subsequent fetching of data) in particular can be moved slightly without any adverse effects.
As part of testing I've used logging with `performance.now()` inserted in various parts of this code, and there's *obviously* no discernible changes between `master` and this patch for e.g. rendering starting in the viewer.
*Note:* The vast majority of this patch is simple indentation changes, which were forced by Prettier (and done automatically with `gulp lint --fix`).
- Ensure that `database.files` actually contains an Array, rather than some arbitrary data.
- Only try to lookup an existing entry when the `database` existed on load, since there's obviously nothing to find when `database.files = []` was set (this case is very common in the MOZCENTRAL build since `sessionStorage` is being used there).
*This is part of a series of patches that will try to split PR 11566 into smaller chunks, to make reviewing more feasible.*
Once all the code has been fixed, we'll be able to eventually enable the ESLint no-shadow rule; see https://eslint.org/docs/rules/no-shadow
When JPEG images are decoded by the browser, on the main-thread, there's a handful of short-lived copies of the image data; see https://github.com/mozilla/pdf.js/blob/c3f4690bde8137d80c74203b1ad91476fc2ca160/src/display/api.js#L2364-L2408
That code thus becomes quite problematic for very big JPEG images, since it increases peak memory usage a lot during decoding. In the referenced issue there's a couple of JPEG images whose dimensions are `10006 x 7088` (i.e. ~68 mega-pixels), which causes the *peak* memory usage to increase by close to `1 GB` (i.e. one giga-byte) in my testing.
By letting the PDF.js JPEG decoder, rather than the browser, handle very large images the *peak* memory usage is considerably reduced and the allocated memory also seem to be reclaimed faster.
*Please note:* This will lead to movement in some existing `eq` tests.
Letting Travis update npm packages can lead to unexpected and completely unrelated failures for any PR and/or merge, see e.g. 11719, if there's ever backwards *incompatible* changes when a package is updated.
The *exact* package versions are specified in `package-lock.json`, and they should thus be used when running tests. Note that the bots won't update npm packages, and neither should Travis as far as I'm concerned.
This property has never been documented and/or *intentionally* exposed through the API, instead the `PDFPageProxy.pageNumber` property is the documented/intended API to use here.
Hence pageIndex is changed to a "private" property on `PDFPageProxy` instances, and internal API functionality is also updated to *consistently* use `this._pageIndex` rather than a mix of formats.
The viewer doesn't currently support executing of any JavaScript, as found in some PDF documents, for security reasons. (Although there's a small hack to at least provide basic support for automatic printing on document load, without running scripts.)
However, in the event that the browser doesn't support printing we're not run *any* of this code. In particular that means that we're also not displaying the "Warning: JavaScript is not supported" message, which seems strange since JavaScript found in a PDF document can really contain *anything* (and not only printing instructions).
It thus seem reasonable, as far as I'm concerned, to always display the JavaScript-warning *even* if printing happens to be unsupported.
Don't accidentally accept invalid glyphNames which *appear* to follow the Cdd{d}/cdd{d} format in `PartialEvaluator._buildSimpleFontToUnicode` (issue 11697)
The /Differences array of the problematic font contains a `/c.1` entry, which is consequently detected as a *possible* Cdd{d}/cdd{d} glyphName by the existing heuristics.
Because of how the base 10 conversion is implemented, which is necessary for the base 16 special case, the parsed charCode becomes `0.1` thus causing `String.fromCodePoint` to throw since that obviously isn't a valid code point.
To fix the referenced issue, and to hopefully prevent similar ones in the future, the patch adds *additional* validation of the charCode found by the heuristics.
*This is part of a series of patches that will try to split PR 11566 into smaller chunks, to make reviewing more feasible.*
Once all the code has been fixed, we'll be able to eventually enable the ESLint `no-shadow` rule; see https://eslint.org/docs/rules/no-shadow
Trying to enable the ESLint rule `no-shadow`, against the `master` branch, would result in a fair number of errors in the `Glyph` class in `src/core/fonts.js`.
Since the glyphs are exposed through the API, we can't very well change the `isSpace` property on `Glyph` instances. Thus the best approach seems, at least to me, to simply rename the `isSpace` helper function to `isWhiteSpace` which shouldn't cause any issues given that it's only used in the `src/core/` folder.
Rather than duplicating the lookup and caching in multiple files, it seems easier to simply move all of this functionality into `src/shared/util.js` instead.
This will also help avoid a bunch of ESLint errors once the `no-shadow` rule is eventually enabled.
While Firefox originally used transition effects for browser UI toolbar buttons, that was removed years ago in https://bugzilla.mozilla.org/show_bug.cgi?id=1393057
Since the PDF.js viewer toolbar transitions were likely based on the Firefox ones, it seems reasonable that these transition effects are removed from PDF.js as well. Besides removing a bunch of CSS, this also makes the toolbar feel ever so slightly more "snappy" without these delays on mouse interaction.
(In order to make it more feasible to modernize/improve the viewer UI, trying to clean-up/simplify existing rules iteratively seems like the most reasonble way to make any progress here w.r.t. being able to test/review things.)
When reftest analyzer shows magnified pixels, there is a seemingly random offset between the mouse position and the magnified position. The reason for this is that reftest analyzer assumes all images have 800 * 1000 pixels but actually the test images have varying sizes.
*This patch fixes something that's annoyed me every now and then over the years, when debugging/fixing corrupt PDF documents.*
For corrupt PDF documents where `Lexer.getHexString` encounters invalid characters, there's very rarely just a handful of them. In practice it's not uncommon for there to be many hundreds, or even many thousands, invalid hex characters found.
Not only is the resulting console warning spam utterly useless in these cases, there's often enough of it that performance may even suffer; hence this patch which limits the amount of messages that any one `Lexer.getHexString` invocation may print.
The PDF document in question is *corrupt*, since it contains an XObject with a truncated dictionary and where the stream contents start without a "stream" operator.
For reasons that I don't even pretend to understand, the `textLayerFactory` property is determined for *every single* page in the PDF document.
Given that the `TextLayerMode` should be consistent for *all* pages in a document, we obviously could/should define `textLayerFactory` just once instead.
There's a couple of issues with this functionality:
- The respective `PromiseCapability` instances are not being reset, in `BaseViewer._resetView`, when the document is closed which is inconsistent with all other state.
- While the default viewer depends on these promises, and they thus ought to be considered part of e.g. the `PDFViewer` API-surface, they're not really defined in a particularily user-visible way (being that they're attached to the `BaseViewer` instance *inline* in `BaseViewer.setDocument`).
- There's some internal `BaseViewer` state, e.g. `BaseViewer._pageViewsReady`, which is tracked manually and could instead be tracked indirectly via the relevant `PromiseCapability`, thus reducing the need to track state *twice* since that's always best to avoid.
*Please note:* In the existing implementation, these promises are not defined *until* the `BaseViewer.setDocument` method has been called.
While it would've been simple to lift that restriction in this patch, I'm purposely choosing *not* to do so since this ensures that any Promise handlers added inside of `BaseViewer.setDocument` are always invoked *before* any external ones (and keeping that behaviour seems generally reasonable).
The only reason for the `return undefined;` lines was to appease the ESLint `consistent-return` rule, but that's not actually necessary if you make use of the fact that the method is `async` and that we can thus await the Promise rather than returning it.
Note that `Dict.set` will only be called with values returned through `Parser.getObj`, and thus indirectly via `Lexer.getObj`. Since neither of those methods will ever return `undefined`, we can simply assert that that's the case when inserting data into the `Dict` and thus get rid of `in` checks when doing the data lookups.
In this case, since `Dict.set` is fairly hot, the patch utilizes an *inline check* and when necessary a direct call to `unreachable` to not affect performance of `gulp server/test` too much (rather than always just calling `assert`).
For very large and complex PDF files this will help performance *slightly*, since `Dict.{get, getAsync, has}` is called *a lot* during parsing in the worker.
This patch was tested using the PDF file from issue 2618, i.e. http://bugzilla-attachments.gnome.org/attachment.cgi?id=226471, with the following manifest file:
```
[
{ "id": "issue2618",
"file": "../web/pdfs/issue2618.pdf",
"md5": "",
"rounds": 250,
"type": "eq"
}
]
```
which gave the following results when comparing this patch against the `master` branch:
```
-- Grouped By browser, stat --
browser | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ------------ | ----- | ------------ | ----------- | --- | ----- | -------------
Firefox | Overall | 250 | 2838 | 2820 | -18 | -0.65 | faster
Firefox | Page Request | 250 | 1 | 2 | 0 | 11.92 | slower
Firefox | Rendering | 250 | 2837 | 2818 | -19 | -0.65 | faster
```
This patch deprecates the existing `getOpenActionDestination` API method, in favor of a better and more general `getOpenAction` method instead. (For now JavaScript actions, related to printing, are still handled as before.)
By clearly separating "regular" Print actions from the JavaScript handling, it's thus possible to get rid of the somewhat annoying and strictly incorrect warning when the viewer loads.
It occured to me that similar to the `getGlobalEventBus` function, it's probably a good idea to *also* notify users of the fact that `eventBusDispatchToDOM` is now deprecated.
Rather than depending of the re-dispatching of internal events to the DOM, the default viewer can instead be used in e.g. the following way:
```javascript
document.addEventListener("webviewerloaded", function() {
PDFViewerApplication.initializedPromise.then(function() {
// The viewer has now been initialized, and its properties can be accessed.
PDFViewerApplication.eventBus.on("pagerendered", function(event) {
console.log("Has rendered page number: " + event.pageNumber);
});
});
});
```
I overlooked this in PR 11631; sorry about that!
Also, ensure that `EventBus` instances *always* track "external" events using a boolean regardless of the actual option value.
This changes the dropdown icon from being set using the `background` CSS property, to being set with `::after` which is *similar* to all the other toolbar button icons (which use `::before`).
Also tweaks the dropdown `background-color` on `:hover` slightly, since the other changes made it too light.
Fixes#11477
The PDF draws many space characters but the embedded fonts don't have a glyph named `space`, so `.notdef` should be drawn instead. PDF.js assumed that Type1 fonts define `.notdef` as the first glyph (index 0). However, now the fonts have the glyph `A` at index 0 and `.notdef` is the last one, so `A` appears where spaces are expected.
Because the rest of the font machinery in `core/fonts.js` assumes `.notdef` is at index zero, it's easiest to modify `core/type1_parser.js` so that it "repairs" fonts and makes sure `.notdef` is at index 0.
The PDF document in question is *corrupt*, since it contains multiple instances of incorrect operators.
We obviously don't want to slow down parsing of *all* documents (since most are valid), just to accommodate a particular bad PDF generator, hence the reason for the inline check before calling the `ensureStateFont` method.
It's no longer necessary to special-case this getter in the `GenericFontLoader` case, since the GENERIC build hasn't been using `mozPrintCallback` for years now (furthermore Firefox 63 is really old as well).
Add a `ignoreDestinationZoom` option/preference to allow users to preserve the current zoom level when navigating to internal destinations (issue 5064, 11606)
Since the goal has always been, essentially since the `EventBus` abstraction was added, to remove all dispatching of DOM events[1] from the viewer components this patch tries to address one thing that came up when updating the examples:
The DOM events are always dispatched last, and it's thus guaranteed that all internal event listeners have been invoked first.
However, there's no such guarantees with the general `EventBus` functionality and the order in which event listeners are invoked is *not* specified. With the promotion of the `EventBus` in the examples, over DOM events, it seems like a good idea to at least *try* to keep this ordering invariant[2] intact.
Obviously this won't prevent anyone from manually calling the new *internal* viewer component methods on the `EventBus`, but hopefully that won't be too common since any existing third-party code would obviously use the `on`/`off` methods and that all of the examples shows the *correct* usage (which should be similarily documented on the "Third party viewer usage" Wiki-page).
---
[1] Looking at the various Firefox-tests, I'm not sure that it'll be possible to (easily) re-write all of them to not rely on DOM events (since getting access to `PDFViewerApplication` might be generally difficult/messy depending on scopes).
In any case, even if technically feasible, it would most likely add *a lot* of complication that may not be desireable in the various Firefox-tests. All-in-all, I'd be fine with keeping the DOM events only for the `MOZCENTRAL` target and gated on `Cu.isInAutomation` (or similar) rather than a preference.
[2] I wouldn't expect any *real* bugs in a custom implementation, simply based on event ordering, but it nonetheless seem like a good idea if any "external" events are still handled last.
To avoid outright breaking third-party usages of the "viewer components" the `getGlobalEventBus` functionality is left intact, but a deprecation message is printed if the function is invoked.
The various examples are updated to *explicitly* initialize an `EventBus` instance, and provide that when initializing the relevant viewer components.
This patch extends the existing heuristics, which are really the best that we can do in general for these kinds of non-embedded *and* non-standard fonts.
Furthermore, this patch also tries to improve the copy-and-paste behaviour for non-embedded Wingdings fonts by also using the `ZapfDingbatsEncoding` in this case.
*Note:* I'm not sure that adding additional tests for Wingdings fonts matters that much, given how limited our "support" for them really is.
Given that all of these primitives implement caching, to avoid unnecessarily duplicating those objects *a lot* during parsing, it would thus be good to actually enforce usage of `Cmd.get()`/`Name.get()`/`Ref.get()` in the code-base.
Luckily it turns out that there's an ESLint rule, which is fairly easy to use, that can be used to disallow arbitrary JavaScript syntax.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/no-restricted-syntax
*This whole patch feels somewhat arbitrary, and I'd be slightly worried about possibly breaking something else.*
To limit the impact of these changes, we only re-parse JPEG images using a reduced `scanLines` value if and only if: An unexpected EOI (End of Image) marker was encountered during decoding of Scan data *and* the "actual" `scanLines` value is at least one order of magnitude smaller than expected.
In some cases PDF documents can contain JPEG images that the native browser decoder cannot handle, e.g. images with DNL (Define Number of Lines) markers or images where the SOF (Start of Frame) marker contains a wildly incorrect `scanLines` parameter.
Currently, for "simple" JPEG images, we're relying on native image decoding to *fail* before falling back to the implementation in `src/core/jpg.js`. In some cases, note e.g. issue 10880, the native image decoder doesn't outright fail and thus some images may not render.
In an attempt to improve the current situation, this patch adds additional validation of the JPEG image SOF data to force the use of `src/core/jpg.js` directly in cases where the native JPEG decoder cannot be trusted to do the right thing.
The only way to implement this is unfortunately to parse the *beginning* of the JPEG image data, looking for a SOF marker. To limit the impact of this extra parsing, the result is cached on the `JpegStream` instance and this code is only run for images which passed all of the pre-existing "can the JPEG image be natively rendered and/or decoded" checks.
---
*Slightly off-topic:* Working on this *really* makes me start questioning if native rendering/decoding of JPEG images is actually a good idea.
There's certain kinds of JPEG images not supported natively, and all of the validation which is now necessary isn't "free". At this point, in the `NativeImageDecoder`, we're having to check for certain properties in the image dictionary, parse the `ColorSpace`, and finally read the actual image data to find the SOF marker.
Furthermore, we cannot just send the image to the main-thread and be done in the "JpegStream" case, but we also need to wait for rendering to complete (or fail) before continuing with other parsing.
In the "JpegDecode" case we're even having to parse part of the image on the main-thread, which seems completely at odds with the principle of doing all heavy parsing in the Worker, and there's also a couple of potentially large (temporary) allocations/copies of TypedArray data involved as well.
Given that this is completely unused, and that a "normal" function call may be a *tiny* bit more efficient, there's no good reason as far as I can tell to keep it.
Rather than adding two buttons, it seems easier to simply link to the relevant section of the README instead (since it also means fewer things to keep up-to-date).
This complements the existing `PDFViewerApplication.initialized` boolean property, and may be helpful for custom implementations of the default viewer. This will thus provide users of the default viewer an alternative to setting the preference to dispatch events to the DOM (and listen for the "localized" event), since they can instead use:
```javascript
document.addEventListener("webviewerloaded", function() {
PDFViewerApplication.initializedPromise.then(function() {
// The viewer has now been initialized.
})
});
```
Note that in order to avoid manually tracking the initialization state *twice*, this implementation purposely uses the `PromiseCapability` functionality to handle both `PDFViewerApplication.initialized` and `PDFViewerApplication.initializedPromise` internally.
*I've had this patch locally for awhile, but have apparently missed to upstream it.*
This simplifies enabling of new ESLint rules, since most of them support automatic fixing of errors, without having to edit `gulpfile.js` or manually invoke ESLint directly.
With the way that the `default_preferences.json` file is now generated at build time, the `gulp lint` task is now noticeably slower than before. This slowdown has been, and still is, somewhat annoying during the deployment of new ESLint rules.
Hence this patch, which moves the `chromium/preferences_schema.json` validation from `gulp lint` and into a new `gulp lint-chromium` task instead. *Obviously* this new task is run as part of the `gulp npm-test` task, and thus through `npm test` on Node.js/Travis, such that it's still being tested as before.
Please note that these changes do *not* affect the *public* interface of the `Metadata` class, but only touches internal structures.[1]
These changes were prompted by looking at the `getAll` method, which simply returns the "private" metadata object to the consumer. This seems wrong conceptually, since it allows way too easy/accidental changes to the internal parsed metadata.
As part of fixing this, the internal metadata was changed to use a `Map` rather than a plain Object.
---
[1] Basically, we shouldn't need to worry about someone depending on internal implementation details.
- Remove the "capturing group" in the regular expression that removes leading "junk" from the raw metadata, since it's not necessary here (it's simply a case of too much copy-pasting in a prior patch).
According to [MDN](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Guide/Regular_Expressions/Cheatsheet#Groups_and_ranges) you want to, for performance reasons, avoid "capturing groups" unless actually needed.
- Add inline comments to document a bunch of magic values in the code.
Given that all `TypedArray` polyfills were removed in PDF.js version `2.0`, since native support is now required, this branch has been dead code for awhile.
In practice it's extremely rare[1] for the padding to be zero in *all* components, hence it seems better to just set it directly rather than creating a temporary variable and checking for the "no padding"-case.
---
[1] In the `tracemonkey.pdf` file that only happens with `0.08%` of all text elements.
Over the years there's been a fair number of issues/PRs opened, where people have wanted to add `hasOwnProperty` checks in (hot) loops in the font parsing code. This has always been rejected, since we don't want to risk reducing performance in the Firefox PDF viewer simply because some users of the general PDF.js library are *incorrectly* extending the `Array.prototype` with enumerable properties.
With this patch the general PDF.js library will now fail immediately with a hopefully useful Error message, rather than having (some) fonts fail to render, when the `Array.prototype` is incorrectly extended.
Note that I did consider making this a warning, but ultimately decided against it since it's first of all possible to disable those (with the `verbosity` parameter). Secondly, even when printed, warnings can be easy to overlook and finally a warning may also *seem* OK to ignore (as opposed to an actual Error).
This patch contains some *much* needed clean-up of, and improvements to, this old code thus addressing a number of issues:
- Set more reasonable *default* values for the widths, in `web/viewer.css`, since the current ones are actually too small even for the (default) `en-US` locale.
This obviously result in a slightly larger zoom dropdown width for many locales, but the more consistent UI does look good to me.
- Stop setting the `min-width`/`max-width` and just use `width` instead.
- Set an explicit `height` of the zoom dropdown, in an attempt to get Google Chrome to display it with the same height as the toolbar buttons.
- Remove additional `Element.setAttribute("style", ...)` usage.
- Actually check *all* of the predefined l10n strings, since the old implementation (implicitly) assumed that the currently selected one was the longest (note e.g. the `ja-JP` locale where one string is considerably longer than the rest).
- Stop invalidating the DOM multiple times when doing the measurements. This was achieved by using a temporary in-memory `canvas`, and we now only need to query the DOM once in order to get the current font properties of the zoom dropdown.
This patch makes the following changes, to improve these API methods:
- Let `PDFPageProxy.cleanup` return a boolean indicating if clean-up actually happened, since ongoing rendering will block clean-up.
Besides being used in other parts of this patch, it seems that an API user may also be interested in the return value given that clean-up isn't *guaranteed* to happen.
- Let `PDFDocumentProxy.cleanup` return the promise indicating when clean-up is finished.
- Improve the JSDoc comment for `PDFDocumentProxy.cleanup` to mention that clean-up is triggered on *both* threads (without going into unnecessary specifics regarding what *exactly* said data actually is).
Add a note in the JSDoc comment about not calling this method when rendering is ongoing.
- Change `WorkerTransport.startCleanup` to throw an `Error` if it's called when rendering is ongoing, to prevent rendering from breaking.
Please note that this won't stop *worker-thread* clean-up from happening (since there's no general "something is rendering"-flag), however I'm not sure if that's really a problem; but please don't quote me on that :-)
All of the caches that's being cleared in `Catalog.cleanup`, on the worker-thread, *should* be re-filled automatically even if cleared *during* parsing/rendering, and the only thing that probably happens is that e.g. font data would have to be re-parsed.
On the main-thread, on the other hand, clearing the caches is more-or-less guaranteed to cause rendering errors, since the rendering code in `src/display/canvas.js` isn't able to re-request any image/font data that's suddenly being pulled out from under it.
- Last, but not least, add a couple of basic unit-tests for the clean-up functionality.
While it would be nice to change the `PDFFormatVersion` property, as returned through `PDFDocumentProxy.getMetadata`, to a number (rather than a string) that would unfortunately be a breaking API change.
However, it does seem like a good idea to at least *validate* the PDF header version on the worker-thread, rather than potentially returning an arbitrary string.
This should hopefully be useful in environments where restrictive CSPs are in effect.
In most cases the replacement is entirely straighforward, and there's only a couple of special cases:
- For the `src/display/font_loader.js` and `web/pdf_outline_viewer.js `cases, since the elements aren't appended to the document yet, it shouldn't matter if the style properties are set one-by-one rather than all at once.
- For the `web/debugger.js` case, there's really no need to set the `padding` inline at all and the definition was simply moved to `web/viewer.css` instead.
*Please note:* There's still *a single* case left, in `web/toolbar.js` for setting the width of the zoom dropdown, which is left intact for now.
The reasons are that this particular case shouldn't matter for users of the general PDF.js library, and that it'd make a lot more sense to just try and re-factor that very old code anyway (thus fixing the `setAttribute` usage in the process).
Having thought *briefly* about using `css-vars-ponyfill`, I'm no longer convinced that it'd be a good idea. The reason is that if we actually want to properly support CSS variables, then that functionality should be available in *all* of our CSS files.
Note in particular the `pdf_viewer.css` file that's built as part of the `COMPONENTS` target, in which case I really cannot see how a rewrite-at-the-client solution would ever be guaranteed to always work correctly and without accidentally touching other CSS in the surrounding application.
All-in-all, simply re-writing the CSS variables at build-time seems much easier and is thus the approach taken in this patch; courtesy of https://github.com/MadLittleMods/postcss-css-variables
By using its `preserve` option, the built files will thus include *both* a fallback and a modern `var(...)` format[1]. As a proof-of-concept this patch removes a couple of manually added fallback values, and converts an additional sidebar related property to use a CSS variable.
---
[1] Comparing the `master` branch with this patch, when using `gulp generic`, produces the following diff for the built `web/viewer.css` file:
```diff
@@ -408,6 +408,7 @@
:root {
--sidebar-width: 200px;
+ --sidebar-transition-duration: 200ms;
}
* {
@@ -550,27 +551,28 @@
position: absolute;
top: 32px;
bottom: 0;
- width: 200px; /* Here, and elsewhere below, keep the constant value for compatibility
- with older browsers that lack support for CSS variables. */
+ width: 200px;
width: var(--sidebar-width);
visibility: hidden;
z-index: 100;
border-top: 1px solid rgba(51, 51, 51, 1);
-webkit-transition-duration: 200ms;
transition-duration: 200ms;
+ -webkit-transition-duration: var(--sidebar-transition-duration);
+ transition-duration: var(--sidebar-transition-duration);
-webkit-transition-timing-function: ease;
transition-timing-function: ease;
}
html[dir='ltr'] #sidebarContainer {
-webkit-transition-property: left;
transition-property: left;
- left: -200px;
+ left: calc(-1 * 200px);
left: calc(-1 * var(--sidebar-width));
}
html[dir='rtl'] #sidebarContainer {
-webkit-transition-property: right;
transition-property: right;
- right: -200px;
+ right: calc(-1 * 200px);
right: calc(-1 * var(--sidebar-width));
}
@@ -640,6 +642,8 @@
#viewerContainer:not(.pdfPresentationMode) {
-webkit-transition-duration: 200ms;
transition-duration: 200ms;
+ -webkit-transition-duration: var(--sidebar-transition-duration);
+ transition-duration: var(--sidebar-transition-duration);
-webkit-transition-timing-function: ease;
transition-timing-function: ease;
}
```
Based on the PDF spec, with `v` operator, current point should be used as the first control point of the curve.
Do not overwrite current point before an SVG curve is built, so it can b actually used as first control point.
Since this rule is now enforced in the entire `/web` folder, there's no good reason for it to remain disabled in this file. Most likely, it's added to reduce the number of linting errors when we started enforcing block-scoped variables.
As can be seen in the code, the `xScaleBlockOffset` typed array doesn't depend on the actual image data but only on the width and x-scale. The width is obviously consistent for an image, and it turns out that in practice the `componentScaleX` is quite often identical between two (or more) adjacent image components.
All-in-all it's thus not necessary to *unconditionally* re-compute the `xScaleBlockOffset` when getting the JPEG image data.
While avoiding, in many cases, one or more loops can never be a bad thing these changes are unfortunately completely dominated by the rest of the JpegImage code and consequently doesn't really show up in benchmark results. *Hence I'd understand if this patch is ultimately deemed not necessary.*
*Note:* This is inspired by PR 5473, which made similar changes for another kind of JPEG data.
Since the implementation in `src/core/jpg.js` only supports 8-bit data, as opposed to similar code in `src/core/colorspace.js`, the computations can be further simplified since the `scale` is always constant.
By updating the coefficients, effectively inlining the `scale`, we'll thus avoid *four* multiplications for each loop iteration.
Unfortunately I wasn't able, based on a quick look through the test-files, to find a sufficiently *large* CMYK JPEG image in order for these changes to really show up in benchmark results. However, when testing the `cmykjpeg.pdf` manually there's a total of `120 000` fewer multiplication with this patch.
There's no particular reason for using the PDF.js helper function `createObjectURL`[1] in the debugger, instead of the native `URL.createObjectURL` directly, for a couple of reasons:
- The relevant code-path only applies to fonts loaded with the Font Loading API, and this isn't supported in IE anyway.
- The debugger can, since quite some time, not even be loaded in IE any more.
- General support for IE is now limited, and there's no guaratee that everything actually works.
---
[1] It provides a fallback for browsers with broken `Blob` support, which as usual means Internet Explorer :-P
In the current `AnnotationLayer` implementation, Popup annotations require that the parent annotation have already been rendered (otherwise they're simply ignored).
Usually the annotations are ordered, in the `/Annots` array, in such a way that this isn't a problem, however there's obviously no guarantee that all PDF generators actually do so. Hence we simply ensure, when rendering the `AnnotationLayer`, that the Popup annotations are handled last.
Interestingly the viewer already seem to work correctly as-is, with workers disabled and a non-standard `verbosity` level.
Hence this is possibly Node.js specific, but given that the issue is lacking *both* the PDF file in question and a runnable test-case, so this patch is essentially a best-effort guess at what the problem could be.
- Re-factor the "incorrect encoding" check, since this can be easily achieved using the general `findNextFileMarker` helper function (with a suitable `startPos` argument).
- Tweak a condition, to make it easier to see that the end of the data has been reached.
- Add a reference test for issue 1877, since it's what prompted the "incorrect encoding" check.
The other image decoders, i.e. the JBIG2 and JPEG 2000 ones, are using the common helper function `readUint16`. Most likely, the only reason that the JPEG decoder is doing it this way is because it originated *outside* of the PDF.js library.
Hence we can simply re-factor `src/core/jpg.js` to use the common `readUint16` helper function, which is especially nice given that the functionality was essentially *duplicated* in the code.
This covers cases that the `--fix` command couldn't deal with, and in a few cases (notably `src/core/jbig2.js`) the code was changed to use block-scoped variables instead.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/prefer-const
With the recent introduction of Prettier this sort of mass enabling of ESLint rules becomes a lot easier, since the code will be automatically reformatted as necessary to account for e.g. changed line lengths.
Note that this patch is generated automatically, by using the ESLint `--fix` argument, and will thus require some additional clean-up (which is done separately).
The debugging hash parameters[1] are intended to facilitate access to various tools/settings in PRODUCTION builds, protected by the `pdfBugEnabled` preference.
At this point, the remaining debugging hash parameters are mainly intended to allow access to the `PDFBug` tools and/or to quickly toggle certain larger features.
The "useOnlyCssZoom" functionality doesn't really seem to fit in with the rest of these hash parameters, since:
- This is, comparatively speaking, a minor viewer-specific feature.
- The zooming implementation will (almost) always fallback to CSS-only zooming, for any document, once the canvases becomes large enough. Hence, the majority of the CSS zooming feature can still be tested *directly* in any build of the viewer.
- After the initial implementation, years ago, the CSS-only zooming code in question hasn't changed much (or even at all), i.e. it doesn't seem like an active development target.[2]
- If the "useOnlyCssZoom" functionality was added today, it's unlikely that a hash parameter would've been added.
- Last, but not least, there's also a `useOnlyCssZoom` preference hence toggling this functionality shouldn't be too difficult (e.g. if someone needs to hack on it).
All in all, I'm thus suggesting that we remove the "useOnlyCssZoom" hash parameter.
---
[1] Originally these hash parameters could be used directly in any build, which was bad since it would allow any link to potentially disable functionality and/or reduce performance.
[2] If it had seen active development over the years, I'd be *much* more inclined to keep the hash parameter.
For people running e.g. Firefox with the `pdfBugEnabled` preference set, to allow quick access to debugging tools, this method will obviously run for every opened PDF file. However, in most cases the URL hash is empty and we can thus skip most of the parsing and simply return early instead.
These two cases should have been whitelisted prior to re-formatting respectively had the comments fixed afterwards, however I unfortunately missed them because of the massive size of the diff.
Fixes#11403
The PDF uses the non-embedded Type1 font Helvetica. Character codes 194 and 160 (`Â` and `NBSP`) are encoded as `.notdef`. We shouldn't show those glyphs because it seems that Acrobat Reader doesn't draw glyphs that are named `.notdef` in fonts like this.
In addition to testing `glyphName === ".notdef"`, we must test also `glyphName === ""` because the name `""` is used in `core/encodings.js` for undefined glyphs in encodings like `WinAnsiEncoding`.
The solution above hides the `Â` characters but now the replacement character (space) appears to be too wide. I found out that PDF.js ignores font's `Widths` array if the font has no `FontDescriptor` entry. That happens in #11403, so the default widths of Helvetica were used as specified in `core/metrics.js` and `.nodef` got a width of 333. The correct width is 0 as specified by the `Widths` array in the PDF. Thus we must never ignore `Widths`.
This removes a couple of, thanks to preceeding code, unnecessary `typeof PDFJSDev` checks, and also fixes a couple of incorrectly implemented (my fault) checks intended for `TESTING` builds.
After PR 9566, which removed all of the old Firefox extension code, the `FIREFOX` build flag is no longer used for anything.
It thus seems to me that it should be removed, for a couple of reasons:
- It's simply dead code now, which only serves to add confusion when looking at the `PDFJSDev` calls.
- It used to be that `MOZCENTRAL` and `FIREFOX` was *almost* always used together. However, ever since PR 9566 there's obviously been no effort put into keeping the `FIREFOX` build flags up to date.
- In the event that a new, Webextension based, Firefox addon is created in the future you'd still need to audit all `MOZCENTRAL` (and possibly `CHROME`) build flags to see what'd make sense for the addon.
While only the `MOZCENTRAL` builds will actually do anything meaningful with the telemetry data, none of the code in question actually runs *at all* in e.g. development mode.[1]
This seems bad since it essentially means that this code is completely untested, despite being quite important for the built-in Firefox PDF viewer, and this thus ought to be fixed.
In this case, the explanation for the current state of the code should be "for historical reasons". Before the viewer was split into the current components and before the pre-processor was improved, back when all code resided in the `web/viewer.js` file, the telemetry reporting was done with *direct* `FirefoxCom` calls. However, with the dummy `DefaultExternalServices.reportTelemetry` method there's nothing actually preventing attempting to report telemetry in any type of build.
NOTE: By running this code in GENERIC builds as well, in addition to just locally, the *viewer* part of telemetry reporting becomes tested e.g. in preview builds too which should help with reviewing.
---
[1] When fixing bug 1606566, I had to edit the relevant `PDFJSDev` checks to be able to actually test the changes locally.
With https://bugzilla.mozilla.org/show_bug.cgi?id=844349 now being fixed in Firefox, the textLayer will now actually stay hidden as intended regardless of the browser settings.
Hence it should no longer be necessary to display the fallback bar, nor print a warning in the console, for documents which contains a textLayer.
Besides removing the `supportsDocumentColors` methods in the default viewer, we can also remove a now unused l10n string.
This particular JSDoc comment is fairly old and it also contains some now unrelated/confusing information.
The only way to *guarantee* that the PDF.js library works as expected is to correctly set the global `workerSrc`[1], hence giving the impression that the option isn't strictly necessary is thus incorrect.
---
[1] Since advertising the fallbackWorkerSrc functionality definitely seems like the *wrong* thing to do.
This way we'll benefit from the existing font caching, and can thus avoid re-creating a fallback font over and over again during parsing.
(Thece changes necessitated the previous patch, since otherwise breakage could occur e.g. with fake workers.)
This is beneficial in situations where the Worker is being re-used, for example with fake workers, since it ensures that things like font resources are actually released.
This rule is already enabled in mozilla-central, and helps avoid some confusing formatting, see https://searchfox.org/mozilla-central/rev/9e45d74b956be046e5021a746b0c8912f1c27318/tools/lint/eslint/eslint-plugin-mozilla/lib/configs/recommended.js#209-210
With the recent introduction of Prettier some of the existing nested ternary statements became even more difficult to read, since any possibly helpful indentation was removed.
This particular ESLint rule wasn't entirely straightforward to enable, and I do recognize that there's a certain amount of subjectivity in the changes being made. Generally, the changes in this patch fall into three categories:
- Cases where a value is only clamped to a certain range (the easiest ones to update).
- Cases where the values involved are "simple", such as Numbers and Strings, which are re-factored to initialize the variable with the *default* value and only update it when necessary by using `if`/`else if` statements.
- Cases with more complex and/or larger values, such as TypedArrays, which are re-factored to let the variable be (implicitly) undefined and where all values are then set through `if`/`else if`/`else` statements.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/no-nested-ternary
Note that this will still allow the FBF tests to run locally, and also on the bots when invoked with `test`/`browsertest` (to not lose all the FBF test-coverage), but will no longer prevent `makeref` from running successfully on the bots.
As described in the issue, having a DOM element with `id=page2` (or any other number) will automatically cause that element to become linkable through the URL hash. That's currently leading to some confusing and outright wrong behaviour, since it obviously only works for pages that have been loaded and rendered.
For PDF documents the only officially supported way to reference a particular page through the URL hash is using the `#page=2` format, which also works for all pages regardless if they're loaded or not.
As far as I can tell there's nothing in the PDF.js default viewer that actually depends on the page/thumbnail `id` at this point in time, hence why I believe that this removal ought to be safe.
Just as a pre-caution this patch adds an `aria-label` to the page canvas, similar to the thumbnail canvas/image, to at least keep this information in the DOM.
Currently *all* users of `pdfjs-dist` are forced to install the `webpack` and `worker-loader` packages, despite the fact that they are *only* relevant if the `webpack.js` file is being used (with a custom Webpack build).
This really doesn't seem great, especially since those packages are the only remaining dependencies in the `pdfjs-dist` library, and it thus seem more reasonable overall that Webpack users handle those dependencies themselves.
To prevent unnecessarily cryptic runtime failures, when people update to newer `pdfjs-dist` versions, the `webpack.js` file was updated to explicitly check for the existence of the `worker-loader` package and error otherwise.
Furthermore, note that `webpack` was only listed as a dependency because of the `worker-loader` package itself (see issue 9248).
Obviously these changes may not be seen as great by Webpack users who rely on `pdfjs-dist`, since it forces them to handle the dependencies themselves, however it should improve things considerably for "general" users of `pdfjs-dist` by not burdening them with unnecessary dependencies.
These sort of changes are also in line with other recent changes, see PR 11418, which removed built-in fake worker loader code for specific JS builders/bundlers/frameworks. This work was prompted not only by a desire to simplify/clean-up old code, but also to lessen future support burden since the PDF.js contributors cannot be assumed to be experts in various JS bundlers.
In order to eventually get rid of SystemJS and start using native `import`s instead, we'll need to provide "complete" file identifiers since otherwise there'll be MIME type errors when attempting to use `import`.
When working on PR 11463 I couldn't help thinking that the `Array.prototype.some` callback function, used when determining the generator, was somewhat difficult to read with its partly unused and strangely named parameters.
This is not only useful to have one format for consistency, but also to
be able to quickly search for colors for e.g., finding duplicates or
when tweaking the CSS for custom deployments.
For reasons that I now cannot even begin to understand, the non-standard SegoeUISymbol font was placed in the `getStdFontMap`. That honestly makes no sense, hence this patch which does what I *should* have done from the start.
This covers the handful of cases that the `--fix` command couldn't deal with, and the changes aren't just fixing the linting errors but attempt to slightly improve the relevant code.
Please find additional details about the ESLint rule at https://eslint.org/docs/rules/prefer-const
Note that this patch is generated automatically, by using the ESLint `--fix` argument, and will thus require some additional clean-up (which is done separately).
Rather than having a copy of this regular expression in the `test/unit/api_spec.js` file, with a comment about keeping it up-to-date with the code in the viewer (note the incorrect file reference as well), we can just import it instead to simplify all of this.
The original issue did not contain a (reduced) test case that we could
include and linked test cases are not ideal for unit tests, so the
original PR could only be verified manually.
I found this a bit unfortunate considering that the print data is
exposed through the API, so I thought about how we could have an
automated test and managed to create a reduced test case with the
OpenAction dictionary from the file in the original issue.
Therefore, this commit includes a unit test for parsing OpenAction
dictionaries without `Type` entries. I verified that this PDF file
behaves the same as the original one, i.e., no print dialog is shown for
older viewers and the print dialog is shown for the most recent viewer.
This patch makes the follow changes:
- Remove no longer necessary inline `// eslint-disable-...` comments.
- Fix `// eslint-disable-...` comments that Prettier moved down, thus causing new linting errors.
- Concatenate strings which now fit on just one line.
- Fix comments that are now too long.
- Finally, and most importantly, adjust comments that Prettier moved down, since the new positions often is confusing or outright wrong.
Note that Prettier, purposely, has only limited [configuration options](https://prettier.io/docs/en/options.html). The configuration file is based on [the one in `mozilla central`](https://searchfox.org/mozilla-central/source/.prettierrc) with just a few additions (to avoid future breakage if the defaults ever changes).
Prettier is being used for a couple of reasons:
- To be consistent with `mozilla-central`, where Prettier is already in use across the tree.
- To ensure a *consistent* coding style everywhere, which is automatically enforced during linting (since Prettier is used as an ESLint plugin). This thus ends "all" formatting disussions once and for all, removing the need for review comments on most stylistic matters.
Many ESLint options are now redundant, and I've tried my best to remove all the now unnecessary options (but I may have missed some).
Note also that since Prettier considers the `printWidth` option as a guide, rather than a hard rule, this patch resorts to a small hack in the ESLint config to ensure that *comments* won't become too long.
*Please note:* This patch is generated automatically, by appending the `--fix` argument to the ESLint call used in the `gulp lint` task. It will thus require some additional clean-up, which will be done in a *separate* commit.
(On a more personal note, I'll readily admit that some of the changes Prettier makes are *extremely* ugly. However, in the name of consistency we'll probably have to live with that.)
There's a fair number of (primarily) `Array`s/`TypedArray`s whose formatting we don't want disturb, since in many cases that would lead to the code becoming much more difficult to read and/or break existing inline comments.
*Please note:* It may be a good idea to look through these cases individually, and possibly re-write some of the them (especially the `String` ones) to reduce the need for all of these ignore commands.
I recently noticed a couple of intermittent failures on Travis, hence this patch which changes the expectation to be identical to the 'Page Request' check in the preceding test-case.
https://npmjs.com/advisories/1084
webpack-stream between 4.0.3 and 5.0.0 added official support for Webpack 4. Which is nice since we ARE using Webpack 4... Also, dropped support for Node.js 4 which shouldn't be a big deal for us since we are already using packages that are incompatible with Node.js 4 (Webpack 4.x supports Node.js 6 and up).
Apparently Ghostscript can, in some cases, generate/include `Metadata` with incorrectly encoded characters.[1] This results in the viewer title looking wrong, which we thus attempt to avoid by falling back to the `Info` entry instead.
*Please note:* Obviously it would be better if this was fixed in the `Metadata` parser instead, rather than using a viewer work-around, but I'm just not sure how or even *if* that could actually be done given that the `Metadata` stream contains no trace of the *original* character.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1605526
---
[1] The problematic characters are from the Specials Unicode block, see https://en.wikipedia.org/wiki/Specials_(Unicode_block)
With these changes we'll always set the `pdfTitle` to the `Info` dictionary entry *first* (assuming it exists and isn't empty), before attempting to override it with the `Metadata` stream entry (assuming it exists, is non-empty *and* valid).
There should be no functional changes with this patch, but it will simplify the following patch somewhat.
Since bundlers, such as Webpack, cannot be told to leave `require` statements alone we are thus forced to jump through hoops in order to prevent these warnings in third-party deployments of the PDF.js library; please see [Webpack issue 8826](https://github.com/webpack/webpack) and libraries such as [require-fool-webpack](https://github.com/sindresorhus/require-fool-webpack).
*Please note:* This is based on the assumption that code running in Node.js won't ever be affected by e.g. Content Security Policies that prevent use of `eval`. If that ever occurs, we should revert to a normal `require` statement and simply document the Webpack warnings instead.
This patch reduces some duplication, by moving *all* fake worker loader code into the `setupFakeWorkerGlobal` function. Furthermore, the functions are simplified further by using `async`/`await` where appropriate.
There's no particularily good reason, as far as I can tell, to not support a custom worker path in Node.js environments (even if workers aren't supported). This patch thus make the Node.js fake worker loader code-path consistent with the fallback code-path used with *browser* fake worker loader.
Finally, this patch also deprecates[1] the `fallbackWorkerSrc` functionality, except in Node.js, since the user should *always* provide correct worker options since the fallback is nothing more than a best-effort solution.
---
[1] Although it probably shouldn't be removed until the next major version.
For performance reasons, and to avoid hanging the browser UI, the PDF.js library should *always* be used with web workers enabled.
At this point in time all of the supported browsers should have proper worker support, and Node.js is thus the only environment where workers aren't supported. Hence it no longer seems relevant/necessary to provide, by default, fake worker loaders for various JS builders/bundlers/frameworks in the PDF.js code itself.[1]
In order to simplify things, the fake worker loader code is thus simplified to now *only* support Node.js usage respectively "normal" browser usage out-of-the-box.[2]
*Please note:* The officially intended way of using the PDF.js library is with workers enabled, which can be done by setting `GlobalWorkerOptions.workerSrc`, `GlobalWorkerOptions.workerPort`, or manually providing a `PDFWorker` instance when calling `getDocument`.
---
[1] Note that it's still possible to *manually* disable workers, simply my manually loading the built `pdf.worker.js` file into the (current) global scope, however this's mostly intended for testing/debugging purposes.
[2] Unfortunately some bundlers such as Webpack, when used with third-party deployments of the PDF.js library, will start to print `Critical dependency: ...` warnings when run against the built `pdf.js` file from this patch. The reason is that despite the `require` calls being protected by *runtime* `isNodeJS` checks, it's not possible to simply tell Webpack to just ignore the `require`; please see [Webpack issue 8826](https://github.com/webpack/webpack) and libraries such as [require-fool-webpack](https://github.com/sindresorhus/require-fool-webpack).
*Please note:* This patch does *not* prevent the 'default_preferences' task from running more than once during `npm test`, but it does ensure that the tasks won't run *concurrently* by running the relevant tests in *series*.
While it would obviously still make sense to re-factor the gulpfile to account for changes in `gulp` version 4, by at least tweaking the `npm test` command the intermittent failures on Travis should at least go away.
During initial parsing of every PDF document we're currently creating a few `1 kB` strings, in order to find certain commands needed for initialization.
This seems inefficient, not to mention completely unnecessary, since we can just as well search through the raw bytes directly instead (similar to other parts of the code-base). One small complication here is the need to support backwards search, which does add some amount of "duplication" to this function.
The main benefits here are:
- No longer necessary to allocate *temporary* `1 kB` strings during initial parsing, thus saving some memory.
- In practice, for well-formed PDF documents, the number of iterations required to find the commands are usually very low. (For the `tracemonkey.pdf` file, there's a *total* of only 30 loop iterations.)
Looking at the `parseCurrentHash` function again it's now difficult for me to understand *what* I was thinking, since having a helper function that needs to be manually passed a `linkService` reference just looks weird.
This patch addresses a couple of smaller issues with the `PDFHistory` class:
- Most, if not all, other viewer components can be reset in one way or another, and there's no good reason for the `PDFHistory` implementation to be different here.
- Currently it's (technically) possible to keep adding entries to the browser history, via the `PDFHistory` instance, even after the document has been closed. That obviously makes no sense, and is caused by the lack of a `reset` method.
- The internal `this._isPagesLoaded` property was never actually reset, which would lead to it being temporarily wrong when a new document was opened in the default viewer.
Note that most (reasonably) modern browsers have supported this for a while now, see https://developer.mozilla.org/en-US/docs/Web/API/ReadableStream#Browser_compatibility
By moving the polyfill into `src/shared/compatibility.js` we can thus get rid of the need to manually export/import `ReadableStream` and simply use it directly instead.
The only change here which *could* possibly lead to a difference in behavior is in the `isFetchSupported` function. Previously we attempted to check for the existence of a global `ReadableStream` implementation, which could now pass (assuming obviously that the preceding checks also succeeded).
However I'm not sure if that's a problem, since the previous check only confirmed the existence of a native `ReadableStream` implementation and not that it actually worked correctly. Finally it *could* just as well have been a globally registered polyfill from an application embedding the PDF.js library.
Given that the error in question is surfaced on the API-side, this patch makes the following changes:
- Updates the wording such that it'll hopefully be slightly easier for users to understand.
- Changes the plain `Error` to an `InvalidPDFException` instead, since that should work better with the existing Error handling.
- Adds a unit-test which loads an empty PDF document (and also improves a pre-existing `InvalidPDFException` message and its test-case).
Given how this method is currently used there shouldn't be any fonts loaded at the point in time where it's called, but it does seem like a bad idea to assume that that's always going to be the case. Since `PDFDocument.checkFirstPage` is already asynchronous, it's easy enough to simply await `Catalog.cleanup` here.
(The patch also makes a tiny simplification in a loop in `Catalog.cleanup`.)
The `viewer` option was *only* used for checking that a document is loaded in `PDFPresentationMode.request`, however that's just as easy to do by simply utilizing `BaseViewer.pagesCount` instead and this way we can also avoid the DOM lookup.
In the PDF document in question, there's an ASCII85Decode inline image where the '>' part of EOD (end-of-data) marker is missing; hence the PDF document is corrupt.
This was a blatant oversight in PR 10217, since there's obviously no `this.pageNumber` property anywhere in the `BaseViewer`. Luckily this shouldn't have caused any bugs, since the only call-site is also validating the `pageNumber` (but correctly that time).
*This patch is simple enough that I almost feel like I'm overlooking some trivial reason why this would be a bad idea.*
Note how in `{BaseViewer, PDFThumbnailViewer}.setDocument` we're always getting the *first* `pdfPage` in order to initialize all pages/thumbnails.
However, once that's done the first `pdfPage` is simply ignored and rendering of the first page thus requires calling `PDFDocumentProxy.getPage` yet again. (And in the `BaseViewer` case, it's even done once more after `onePageRenderedCapability` is resolved.)
All in all, I cannot see why we cannot just immediately set the first `pdfPage` and thus avoid an early round-trip to the API in the `_ensurePdfPageLoaded` method before rendering can begin.
Note that the XRef cache will only hold objects returned through `Parser.getObj`, and indirectly via `Lexer.getObj`. Since neither of those methods will ever return `undefined`, we can simply `assert` that when inserting objects into the cache and thus get rid of one function call when doing cache lookups.
Obviously this won't have a huge effect on performance, however `XRef.fetch` is usually called *a lot* in larger documents and this patch thus cannot hurt.
I'm slightly surprised that this hasn't actually caused any (known) bugs, but that may be more luck than anything else since it fortunately doesn't seem common for Streams to be defined inside of an 'ObjStm'.[1]
Note that in the `XRef.fetchUncompressed` method we're *not* caching Streams, and that for very good reasons too.
- Streams, especially the `DecodeStream` ones, can become *very* large once read. Hence caching them really isn't a good idea simply because of the (potential) memory impact of doing so.
- Attempting to read from the *same* Stream more than once won't work, unless it's `reset` in between, since using any method such as e.g. `getBytes` always starts at the current data position.
- Given that even the `src/core/` code is now fairly asynchronous, see e.g. the `PartialEvaluator`, it's generally impossible to assert that any one Stream isn't being accessed "concurrently" by e.g. different `getOperatorList` calls. Hence `reset`-ing a cached Streams isn't going to work in the general case.
All in all, I cannot understand why it'd ever be correct to cache Streams in the `XRef.fetchCompressed` method.
---
[1] One example where that happens is the `issue3115r.pdf` file in the test-suite, where the streams in question are not actually used for anything within the PDF.js code.
- Change all occurences of `var` to `let`/`const`.
- Initialize the (temporary) Arrays with the correct sizes upfront.
- Inline the `isCmd` check. Obviously this won't make a huge difference, but given that the check is only relevant for corrupt documents it cannot hurt.
Having ran the entire test-suite locally with these `Number.isInteger` checks removed, there wasn't a single test failure anywhere; see also PR 8857.
Hence everything points to this being completely unnecessary now, and by removing this code there's thus fewer function calls being made in `XRef.fetchUncompressed`.
The contents of this comment hasn't been correct for *years*, ever since the library was properly split into main/worker-threads, so it's probably high time for this to be updated.
This code was originally added to support IE10 (and below), however with those browsers *explicitly* unsupported since PDF.js version `2.0` this code is now dead.
Obviously the `_pagesRequests` functionality is *mainly* used when `disableAutoFetch` is set, but it will also be used during ranged/streamed loading of documents.
However, the `_pagesRequests` property is currently an Array which seems a bit strange:
- Arrays are zero-indexed, but the first element will never actually be set in the code.
- The `_pagesRequests` Array is never cleared, unless a new document is loaded, and once the `PDFDocumentProxy.getPage` call has resolved/rejected the element is just replaced by `null`.
- Unless the document is browsed *in order* the resulting `_pagesRequests` Array can also be arbitrarily sparse.
All in all, I don't believe that an Array is an appropriate data structure to use for this purpose.
This patch simply restores the behaviour that existed prior to PR 7697, since I cannot imagine that that was changed other than by pure accident.
As mentioned by a comment in `BaseViewer.setDocument`: "Printing is semi-broken with auto fetch disabled.", and note that since triggering of printing is a synchronous operation there's generally no easy way to load the missing data.
https://github.com/mozilla/pdf.js/pull/7697/files#diff-529d1853ee1bba753a0fcb40ea778723L1114-L1118
Considering just how small/simple this code is, it doesn't seem necessary to have a separate method for it (even more so when there's only one call-site).
For documents with a Linearization dictionary the computed `startXRef` position will be relative to the raw file, rather than the actual PDF document itself (which begins with `%PDF-`).
Hence it's necessary to subtract `stream.start` in this case, since otherwise the `XRef.readXRef` method will increment the position too far resulting in parsing errors.
*Please note:* A a similar change was attempted in PR 5005, but it was subsequently backed out (in PR 5069) since other parts of the patch caused issues.
With these changes, it's possible to replace repeated function calls within a loop with just a single function call and subsequent assignment instead.
I've always disliked the solution in PR 10461, since it required changes to the `PDFHistory` code itself to deal with a bug in IE11.
Now that IE11 support is limited, it seems reasonable to remove these `pushState`/`replaceState` hacks from the main code-base and simply use polyfills instead.
For Popup annotation trigger elements consisting of an arbitrary polyline, you need to ensure that the 'stroke-width' is always non-zero since otherwise it's impossible to actually open/close the popup.
Unfortunately I don't believe that any of the test-suites can be used to test this, hence why no tests are included in the patch.
As we've seen in numerous other cases, avoiding unnecessary function calls is never a bad thing (even if the effect is probably tiny here).
With these changes we also avoid potentially two back-to-back `isDict` checks when evaluating possible Page nodes, and can also no longer accidentally pick a dictionary with an incorrect /Type.
For certain canvas-related errors (and probably others), the browser rendering exceptions may be propagated "as-is" to the PDF.js code. In this case, the exceptions are of the somewhat cryptic `NS_ERROR_FAILURE` type.
Unfortunately these aren't actual `Error`s, which thus ends up unintentionally triggering the `assert` in `PDFPageProxy._abortOperatorList`; sorry about that!
The bug report seem to suggest that we don't support UTF-16 strings with a BOM (byte order mark), which we *actually* do as evident by both the code and a unit-test.
The issue at play here is rather that we previously only supported big-endian UTF-16 BOM, and the `Title` string in the PDF document is using a *little-endian* UTF-16 BOM instead.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1593902
Rather than having to store a `PromiseCapability` on the `ObjectLoader` instances, we can simply convert `_walk` to be `async` and thus have the same functionality with native JavaScript instead.
I happened to look at this code and the way that the link target is set seems unecessarily convoluted, since we're using `Object.values` and `Array.prototype.includes` for *every* link being parsed.
Given that the number of link targets are so few, the easist solution honestly seem to be to just use a `switch` statement to do the link target mapping.
When the end of data has already been reached for the various Streams, the `getByte` methods will return `-1` to signal that to the caller. Note however that the current position obviously won't be incremented in this case, meaning that the `peekByte` methods will in this case *incorrectly* decrement the position.
Thankfully the corresponding `peekBytes` shouldn't be affected by this bug, since they decrement the current position with the *actually* returned number of bytes.
I'm not aware of any bugs caused by this blatant oversight, but that doesn't mean this shouldn't be fixed :-)
This will allow us to attempt to recover as much as possible of a page, rather than immediately failing, when a broken/unsupported ColorSpace is encountered. This patch thus extends the framework added in PRs such as e.g. 8240 and 8922, to also cover parsing of ColorSpaces.
The code in question is *only* relevant in non-`PRODUCTION` mode, i.e. the *development* version of the viewer run with `gulp server`, and has been completely unused at least since SystemJS was added.
I really cannot see any reason to keep this, since it's code which first of all isn't shipping and secondly isn't even being used in the development viewer.
When `ReadableStream` support was added to the `MessageHandler`, the `_onComObjOnMessage` function became more complex than previously.
All of the nested `if`/`else if`/`else` branches are now, at least in my opinion, making some of this code a bit difficult to follow. Hence this patch, which attempts to help readability by making use of early `return`s and `Error`s.
The patch also changes a couple of `var`/`let` occurences to `const`.
Note that using `in` leads to unnecessary stringification of the properties, which seems completely unnecessary here. To avoid future problems from these changes the `MessageHandler.on` method will now assert, in non-`PRODUCTION`/`TESTING` builds, that it's always called with a function as expected.
This patch also renames `callbacksCapabilities` to `callbackCapabilities`, note the removed "s", since using a double plural format looks a bit strange.
Given that the `isReply` property is an internal implementation detail, changing its type shouldn't be a problem. Note that by directly indicating if either data or an Error is sent, it's no longer necessary to use `in` when handling the callback.
Currently, for data in `ChunkedStream` instances, the `getMissingChunks` method is used in a couple of places to determine if data is already available or if it needs to be loaded.
When looking at how `ChunkedStream.getMissingChunks` is being used in the `ObjectLoader` you'll notice that we don't actually care about which *specific* chunks are missing, but rather only want essentially a yes/no answer to the "Is the data available?" question.
Furthermore, when looking at how `ChunkedStream.getMissingChunks` itself is implemented you'll notice that it (somewhat expectedly) always iterates over *all* chunks.
All in all, using `ChunkedStream.getMissingChunks` in the `ObjectLoader` seems like an unnecessary "heavy" and roundabout way to obtain a boolean value. However, it turns out there already exists a `ChunkedStream.allChunksLoaded` method, consisting of a *single* simple check, which seems like a perfect fit for the `ObjectLoader` use cases.
In particular, once the *entire* PDF document has been loaded (which is usually fairly quick with streaming enabled), you'd really want the `ObjectLoader` to be as simple/quick as possible (similar to e.g. loading a local files) which this patch should help with.
Note that I wouldn't expect this patch to have a huge effect on performance, but it will nonetheless save some CPU/memory resources when the `ObjectLoader` is used. (As usual this should help larger PDF documents, w.r.t. both file size and number of pages, the most.)
I completely overlooked this in PR 11281, but you obviously need to make similar changes in `PartialEvaluator.hasBlendModes` since it will otherwise ignore valid Blend Modes.
As can be seen in `PageViewport` only multiples of 90 degrees are really supported by the code, hence the unit-test doesn't really make sense.
(Possibly this should be enforced in the API, to avoid surprises, but given that this problem has always existed I'm passing on that for now.)
This argument is a left-over from older API code, where we unconditionally initialized `StatTimer` instances for every page. For quite some time that's only been done when `pdfBug` is set, hence it seems unnecessary to keep this functionality.
Even though the currect situation only results in six unnecessary function calls per page, it nonetheless seems completely unnecessary to call dummy functions when `pdfBug` is *not* set (i.e. the default behaviour).
For *very* long/large documents fetching all pages on load may cause quite bad performance, both memory and CPU wise. In order to at least slightly alleviate this, we can let the viewer treat these kind of documents[1] as if `disableAutoFetch` were set.
---
[1] One example of a really bad case is https://bugzilla.mozilla.org/show_bug.cgi?id=1588435, which this patch should at least help somewhat. In general, for these cases, we'd probably need to implement switching between `PDFViewer`/`PDFSinglePageViewer` (as already tracked on GitHub) and use the latter for these kind of long documents.
Unfortunately I forgot to test `gulp mozcentraldiff` with PR 11256, which is really bad since it will cause the en-US `chrome.properties` l10n file to be deleted at mozilla-central; sorry about breaking this!
In order to address this we'll have to re-add (only) the en-US `chrome.properties` l10n file, which is simple enough. Furthermore, since we're not doing any sort of build-specific parsing of the l10n files, we can just copy the en-US files as-is rather than having to run `gulp locale` during `gulp mozcentral`.
As can be seen in the API, there's a number of document loading Exception handlers which are both really simple and highly similar. Hence these are changed such that all the relevant Exceptions are sent via *one* message instead.
Furthermore, the patch also avoids unnecessarily re-creating `UnknownErrorException`s at the worker side and removes an unnecessary `bind` call.
This was done automatically, using the `npm audit fix` command.
*Please note:* There's two vulnerabilities left, however they couldn't be fixed automatically and may thus benefit from being handled separately.
Obviously this won't look exactly right, but considering that the PDF file doesn't bother embedding non-standard fonts this is the best that we can do here.
Originally only `skipPages` existed, but given that `firstPage`/`lastPage` has existed for a long time now using them whenever possible looks simpler overall.
- In the `ibwa-bad` case the sixteenth page contains corrupt/incomplete commands, but given that we're suppressing `Error`s by default now skipping hardly seems warranted any more.
- In the `geothermal.pdf` case the first page contains an unsupported ColourSpace, but again we're suppressing `Error`s by default now and skipping hardly seems warranted any more.
With the removal of the (standalone) Firefox building code in PR 9566 (a year and a half ago), these files are now completely unused in the GitHub repository[1].
Hence it doesn't really seem necessary to keep fetching them with `gulp importl10n`, and the existing files in the `l10n` folder can also be removed (thanks to version control, they're easy enough to restore should the need ever arise).
The patch also allows an additional simplification, for the `gulp locale` and `gulp mozcentral` commands, since it's now possible to stop writing `l10n` files to the `extensions/firefox/` folder and instead just copy them similar to other build targets.
---
[1] They're obviously still used in `mozilla-central`, for fallback messages displayed through `PdfStreamConverter.jsm`, but that doesn't make it necessary to keep them *here* as far as I'm concerned.
With the removal of the (standalone) Firefox building code in PR 9566 (a year and a half ago), these files are now completely unused.
Hence it doesn't really make sense to keep building them as part of `gulp locale`, and the existing files in the `l10n` folder can also be removed (thanks to version control, they're easy enough to restore should the need ever arise).
The issue has been open for years now, and has even been marked with `5-good-beginner-bug` for *months*, without any movement.
Considering just how simple the suggested solution is, I'm submitting this patch just to close out a long-standing issue.
The following is printed with every build, which gets kind of annoying when looking at the logs:
```
Replace Autoprefixer browsers option to Browserslist config.
Use browserslist key in package.json or .browserslistrc file.
Using browsers option cause some error. Browserslist config
can be used for Babel, Autoprefixer, postcss-normalize and other tools.
If you really need to use option, rename it to overrideBrowserslist.
Learn more at:
https://github.com/browserslist/browserslist#readmehttps://twitter.com/browserslist
```
Given how we're using Autoprefixer, with a number of build-specific configs, simply changing the option name seems like the easiest solution here.
(I'm also adding a couple of newlines at the `autoprefixer` call-sites, to aid readability.)
Obviously userAgent checks aren't that great, since it's very easy to spoof, but it probably doesn't hurt to attempt to extend this (since it's limited to `GENERIC` builds).
Besides, anyone using the default viewer can always set the `maxCanvasPixels` option to a value of their choosing.
This patch is making me somewhat worried about future regressions, since it's certainly easy to imagine this completely breaking certain kinds of corrupt/edited PDF documents while fixing others.[1]
Obviously it passes all existing reference tests (and even improves one), however compared to many other patches there's no telling how much it could break.
The only reason that I'm even submitting this patch, is because of the number of open issues that it would address.
Generally speaking though, the best course of action would probably be if `XRef.indexObjects` was re-written to be much more robust (since it currently feels somewhat hand-wavy in parts). E.g. by actually checking/validating more of the objects before committing to them.
---
[1] Especially given that it's reverting part of PR 5910, however in the case of issue 5909 it seems that other (more recent) changes have actually made that PR redundant.
Sometimes we also used `@return`, but `@returns` is what the JSDoc
documentation recommends. Even though `@return` works as an alias, it's
good to use the recommended syntax and to be consistent within the
project.
Sometimes we also used `@return` or `@returns`, but `@type` is what
the JSDoc documentation recommends. This also improves the documentation
because before this commit the types were not shown and now they are.
Rather than specifying certain build targets manually, it seems much more appropriate (and future-proof) to use the `SKIP_BABEL` build target instead.
Also, the patch adds a missing `/* eslint no-var: error */` line since I'm touch the file anyway and no code-changes were necessary for it.
As part of attempting to fix a number issues containing PDF documents with corrupt XRef tables, I'd like to improve the reference test-coverage slightly *first*.
Obviously this will increase the runtime of the tests a bit, however I'd rather "waste" resources on the bots instead of developer time fixing regressions which could have been avoided.
For badly generated PDF documents, with issue 6961 being one example, there's well over one hundred thousand function calls being made in total for just the *two* pages.
This handles the two different ways that fonts can be loaded, either by Name (which is the common case) or by Reference.
Furthermore, this also takes the `ignoreErrors` option into account when deciding whether to fallback or Error.
Finally, by creating a minimal but valid Font dictionary, there's no special-cases necessary in any of the font parsing code.
Co-authored-by: huzjakd <huzjakd@gmail.com>
Co-Authored-By: Jonas Jenwald <jonas.jenwald@gmail.com>
With the sole exception of `meh`[1], none of the locales removed here have even been updated since the last change was made to the default `en-US` locale.
Please note that if/when these locales would start shipping in Nightly again, they will automatically be re-added in PDF.js as well with the `gulp importl10n` command.
---
[1] The current translation is also somewhat incomplete, to put it mildly.
All of these methods have been marked as `deprecated` in *three* releases now, and I'd thus like to (slowly) move towards complete removal.
However rather than just removing the methods right away, which would cause somewhat cryptic failures, this patch tries to implement a hopefully reasonable middle ground by throwing `Error`s with (essentially) the same information as the previous warnings.
While the previous `deprecated` messages could perhaps be seen as optional, with these changes API consumers will now be forced to actually migrate their code.
Rather than having the script remove locales automatically, which seems like a heavy-handed approach at least initially, listing these for manual checking seems nice though.
Rather than having to manually maintain a static list of language codes, it's much easier to simply fetch the active ones from `mozilla-central` instead.
As part of this the code in `external/importL10n/locales.js` was modernized slightly, by using Promises/async functions to get rid of a bunch of annoying callbacks (which shouldn't be a problem for reasonably modern Node.js versions).
*Please note:* I've been thinking about possible ways of addressing this issue for a while now, but all of the solutions I came up with became too complicated and thus hurt readability of the code.
However, it occured to me that we're essentially trying to add a heuristic *on top* of another heuristic, and that it shouldn't matter how efficient the code is as long as it works.
In the PDF file in the issue the Encoding contains glyphNames of the `Cdd` format, which our existing heuristics will treat as base 10 values. However, in this particular file they actually contain base 16 values, which we thus attempt to detect and fix such that text-selection works.
I've absolutely no idea why I wrote the code that way originally, since the nested `if`s are not really helping readability one bit.
Hence this patch which changes things to use early `return`s instead, to help readability.
Rather than manually clamping the `width` here, we can just `export` an already existing helper function from `ui_utils.js` instead.
(Hopefully it will eventually be possible to replace this helper function with a native `Math.clamp` function, given that there exists a "Stage 1 Proposal" for adding such a thing to the ECMAScript specification.)
By utilizing a base "class", things become significantly simpler. Unfortunately the new `BaseException` cannot be a proper ES6 class and just extend `Error`, since the SystemJS dependency doesn't seem to play well with that.
Note also that we (generally) need to keep the `name` property on the actual `...Exception` object, rather than on its prototype, since the property will otherwise be dropped during the structured cloning used with `postMessage`.
By default, i.e. with workers enabled, it's *purposely* not possible to send `Dict`s and `Stream`s from the worker-thread. This is achieved by defining a `function` on every `Dict` instance, since that ensures that [the structured clone algoritm](https://developer.mozilla.org/en-US/docs/Web/API/Web_Workers_API/Structured_clone_algorithm) will throw an Error on `postMessage`.
However, with workers *disabled* we fall-back to the `LoopbackPort` implementation which just ignores any `function`s, thus incorrectly allowing sending of data which *should* be unclonable.
The following changes were made:
- Remove unnecessary `typeof` checks in the `get`/`getAsync` methods.
- Reduce unnecessary code duplication in the `get`/`getAsync` methods.
- Inline the `Ref` checks in the `get`/`getAsync`/`getArray` methods, since it helps avoid many unnecessary functions calls. I.e. this way it's possible to directly call `XRef.{fetch, fetchAsync)` only when necessary, rather than always having to call `XRef.{fetchIfRef, fetchIfRefAsync)`.
This patch was tested using the PDF file from issue 2618, i.e. http://bugzilla-attachments.gnome.org/attachment.cgi?id=226471, using the following manifest file:
```
[
{ "id": "issue2618",
"file": "../web/pdfs/issue2618.pdf",
"md5": "",
"rounds": 250,
"type": "eq"
}
]
```
This gave the following results when comparing this patch against the `master` branch:
```
-- Grouped By browser, stat --
browser | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ------------ | ----- | ------------ | ----------- | --- | ----- | -------------
Firefox | Overall | 250 | 2821 | 2790 | -32 | -1.12 | faster
Firefox | Page Request | 250 | 2 | 2 | 0 | 6.68 |
Firefox | Rendering | 250 | 2820 | 2788 | -32 | -1.13 | faster
```
Compared to the recently replaced `URL` polyfill, the new `ReadableStream` polyfill isn't being exported globally for two reasons:
- We're currently checking for the existence of a global `ReadableStream` implementation when determining if the Fetch API will be used; please see `isFetchSupported` in the src/display/display_utils.js file.
- Given that it's much newer functionality (compared to `URL`) and that not all browsers may implement all parts of the specification yet, not exposing the `ReadableStream` globally seems safer for now.
Having these methods fallback to returning `null` in only *one* particular case seems outright wrong, since a "falsy" value will thus be handled incorrectly.
The only reason that this hasn't caused issues in practice is that there's only one call-site passing in three keys, and in that case we're trying to read a font file where falling back to `null` isn't a problem.
This only required changing the import paths. The `es` folder contains
all polyfills we need now. If we want to import everything, we need to
explicitly require the `index` file.
Hopefully this patch makes sense, and in order to reduce the regression risk the implementation ensures that only completely missing widths are being replaced.
Given that there's only a couple of call-sites, and that the helper function is really simple, it doesn't seem entirely necessary to keep it around. While fewer function calls is always a good thing, in this case the performance impact is small enough to be unmeasurable.
With *one* single exception the code in `MessageHandler` is using `reason` when passing around various Errors, hence this patch also renames an `error` key for consistency.
The `streamId` short-hand in `MessageHandler._processStreamMessage` was only used partially througout the method, which seemed kind of strange, hence that's fixed in this patch.
Furthermore, always giving the `streamController` object a constant shape in `MessageHandler.sendWithStream` cannot hurt either.
With this patch we're finally able to abort worker-thread parsing of the `OperatorList`, rather than *only* aborting the main-thread rendering itself, when the `RenderTask.cancel` method is being called.
This will help improve perceived performance in the default viewer, especially when reading longer and more complex documents, since pages that've been scrolled out-of-view (and thus evicted from the cache) will no longer compete for parsing resources on the worker-thread.
*Please note:* With the implementation in this patch we're *not* aborting worker-thread parsing immediately on `RenderTask.cancel`, since that would lead to *worse* performance in many cases. For example: When zoom/rotation occurs in the viewer, while parsing/rendering is still ongoing, a `cancel` call will usually be (almost) immediately folled by a new `PDFPageProxy.render` call. In that case you obviously don't want to abort parsing on the worker-thread, since that would risk throwing away a partially parsed `OperatorList` and thus force unnecessary re-parsing which will regress perceived performance (especially for more complex documents).
When choosing a reasonable delay, before cancelling `getOperatorList` on the worker-thread when `RenderTask.cancel` is called, two different positions need to be considered:
1. The delay needs to be short enough, since a timeout in the multiple seconds range would essentially make this entire functionality meaningless (by always allowing most/all pages enough time to finish parsing).
2. The delay cannot be *too* short, since that would actually *reduce* performance in the zoom/rotation case outlined above. Furthermore, the time between `RenderTask.cancel` and `PDFPageProxy.render` calls will obviously be affected by both general computer performance and current CPU load.
It's certainly possible that the timeout may require some further tweaks, however the value settled on in this patch was easily *one order* of magnitude larger than the delta between cancel/render in my tests.
Currently this leads to confusing behaviour with `eventBusDispatchToDOM = true` set, since then there's *two* different "localized" events being fired.
There's no good reason for calling this helper function without a `url` parameter, and this way we can prevent that from happening.
Note how the `PDFOutlineViewer` call-site was already doing the right thing here, and only the `LinkAnnotationElement` call-site needed a small adjustment to make it work.
This rule is in the process of being rolled out in mozilla-central, and it helps avoid unnecessary `async` functions together with `new Promise(...)`.
Please see https://eslint.org/docs/rules/no-async-promise-executor for additional information.
With the changes made in PR 11069, it's no longer necessary to include the `pageIndex`/`intent` parameters when sending 'GetOperatorList' data. In the previous implementation these properties were used to associate the `OperatorList` with the correct `RenderTask`, however now that `ReadableStream`s are used that's handled automatically and it's thus dead code at this point.
Besides avoiding errors during loading, this also ensures that the document will be correctly scrolled/zoomed into view once the viewer becomes visible.
This "new" behaviour was always intended, see PR 2613, however various re-factoring over the years seem to have broken this (and I'm probably at least somewhat responsible for that).
By transfering, rather than copying, `ArrayBuffer`s between the main- and worker-threads, you can avoid unnecessary allocations by only having *one* copy of the same data.
Hence manually setting `postMessageTransfers: false`, when calling `getDocument`, is a performance footgun[1] which will do nothing but waste memory.
Given that every reasonably modern browser supports `postMessage` transfers[2], I really don't see why it should be possible to force-disable this functionality.
Looking at the browser support, for `postMessage` transfers[2], it's highly unlikely that PDF.js is even usable in browsers without it. However, the feature testing of `postMessage` transfers is kept for the time being just to err on the safe side.
---
[1] This is somewhat similar to the, now removed, `disableWorker` parameter which also provided API users a much too simple way of reducing performance.
[2] See e.g. https://developer.mozilla.org/en-US/docs/Web/API/MessagePort/postMessage#Browser_compatibility and https://developer.mozilla.org/en-US/docs/Web/API/Transferable#Browser_compatibility
Note how the sent values have inconsistent types, with a boolean in one case and an object in the other (normal) case.
Furthermore, explicitly sending a `supportTypedArray: true` property seems superfluous at least to me.
This check was added in PR 2445, however it's no longer necessary since all data[1] is now loaded on the main-thread (and then transferred to the worker-thread).
Furthermore, by default the Fetch API is now (usually) used rather than `XMLHttpRequest`.
All in all, while these checks *were* necessary at one point that's no longer the case and they can thus be removed.
---
[1] This includes both the actual PDF data, as well as the CMap data.
It recently occurred to me that the CMap data should be an excellent candidate for transfering.
This will help reduce peak memory usage for PDF documents using CMaps, since transfering of data avoids duplicating it on both the main- and worker-threads.
Unfortunately it's not possible to actually transfer data when *returning* data through `sendWithPromise`, and another solution had to be used.
Initially I looked at using one message for requesting the data, and another message for returning the actual CMap data. While that should have worked, it would have meant adding a lot more complexity particularly on the worker-thread.
Hence the simplest solution, at least in my opinion, is to utilize `sendWithStream` since that makes it *really* easy to transfer the CMap data. (This required PR 11115 to land first, since otherwise CMap fetch errors won't propagate correctly to the worker-thread.)
Please note that the patch *purposely* only changes the API to Worker communication, and not the API *itself* since changing the interface of `CMapReaderFactory` would be a breaking change.
Furthermore, given the relatively small size of the `.bcmap` files (the largest one is smaller than the default range-request size) streaming doesn't really seem necessary either.
Having recently worked with this code, it struck me that most of the `postMessage` calls where `Error`s are involved have never been correctly implemented (i.e. missing `wrapReason` calls).
There's only three call-sites and one of them doesn't even need the complete functionality of `resolveCall`, hence it seems reasonable to just inline this code.
An additional benefit of this is that the `Function.prototype.apply()` instance can also be converted into "normal" function calls, which should be a tiny bit more efficient.
The patch also replaces a number of unnecessary arrow functions, in relevant parts of the `MessageHandler` code, with "normal" functions instead.
Finally, all `Promise.resolve().then(...)` calls are replaced with `new Promise(...)` instead since the latter is a tiny bit more efficient. This also explains the test failures on the Linux bot, with a prior version of the patch, since the `Promise.resolve().then(...)` format essentially creates two Promises thus causing additional delay.
At this point in time it's easy to convert the `MessageHandler.on` call-sites to use arrow functions, and thus let the JavaScript engine handle scopes for us, rather than having to manually keep references to the relevant scopes in `MessageHandler`.[1]
An additional benefit of this is that a couple of `Function.prototype.call()` instances can now be converted into "normal" function calls, which should be a tiny bit more efficient.
All in all, I don't see any compelling reason why it'd be necessary to keep supporting custom `scope`s in the `MessageHandler` implementation.
---
[1] In the event that a custom scope is ever needed, simply using `bind` on the handler function when calling `MessageHandler.on` ought to work as well.
Since `wrapReason` and `makeReasonSerializable` are essentially functionally equivalent it doesn't seem necessary to keep both of them around, especially when `makeReasonSerializable` only has a *single* call-site.
Given that the `stream` property is an internal implementation detail, changing its type shouldn't be a problem. By using Numbers instead, we can avoid unnecessary String allocations when creating/processing Streams.
With PR 11069 we're now using Streams for OperatorList parsing (in addition to just TextContent parsing), which brings the nice benefit of being able to easily abort parsing on the worker-thread thus saving resources.
However, since we're now creating many more `ReadableStream` there appears to be a tiny bit more overhead because of it (giving ~1% slower runtime of `browsertest` on the bots). In this case we're just going to have to accept such a small regression, since the benefits of using Streams clearly outweighs it.
What we *can* do here, is to try and make the Streams part of the `MessageHandler` implementation slightly more efficient by e.g. removing unnecessary function calls (which has been helpful in other parts of the code-base). To that end, this patch makes the following changes:
- Actually support `transfers` in `MessageHandler.sendWithStream`, since the parameter was being ignored.
- Inline the `sendStreamRequest`/`sendStreamResponse` helper functions at their respective call-sites. Obviously this causes some amount of code duplication, however I still think this change seems reasonable since for each call-site:
- It avoids making one unnecessary function call.
- It avoids allocating one temporary object.
- It avoids sending, and thus structure clone, various undefined object properties.
- Inline objects in the `MessageHandler.{send, sendWithPromise}` methods.
- Finally, directly call `comObj.postMessage` in various methods when `transfers` are *not* present, rather than calling `MessageHandler.postMessage`, to further reduce the amount of function calls.
One of the motivations for using `setAttribute` in the first place was to support more efficient DOM updates in the `expandTextDivs` method, since performance of the `enhanceTextSelection` mode can be somewhat bad when there's a lot of `textDivs` on the page.
With recent `TextLayer` changes/optimizations it's no longer necessary to store a complete `style`-string for every `textDiv`, and we can thus re-visit the `setAttribute` usage.
Note that with the current code, in `appendText`, there's only *one* string per `textDiv` which avoids a bunch of temporary strings. While the changes in this patch means that there's now *three* strings per `textDiv` instead, the total length of these strings are now quite a bit shorter (42 characters to be exact).
*This should obviously have been done in PR 11097, but for some reason I completely overlooked it; sorry about that.*
There's no good reason to update the font unless you're actually going to measure the width of the textContent. This can reduce unnecessary font switching a fair bit, even for documents which are somewhat simple/short (in e.g. the `tracemonkey.pdf` file this cuts the amount of font switches almost in half).
For performance reasons single-char text divs aren't being scaled, as outlined in a comment in `appendText`. Hence it doesn't seem necessary, or even a good idea, to unconditionally measuring the width of the text in `_layoutText`.
This major version bump required two changes:
- The global line in the mobile viewer example should be removed because
the `.eslintrc` file already defines these globals and with the new
`eslint` version we otherwise get an error saying "'pdfjsLib' is already
defined as a built-in global variable".
- The ECMA version for the examples must be set to 6 since we're using
modules, otherwise we get an error saying "sourceType 'module' is not
supported when ecmaVersion < 2015". It turns out that the previous
version of `eslint` already used ECMA version 6 silently even though
we set 5, see https://github.com/eslint/eslint/issues/9687#issuecomment-432413384,
so in terms of our code nothing really changes.
These functions aren't returning anything, now that they're using `ReadableStream`s, and it thus doesn't seem necessary to re-throw errors (also given the console message that's caused by it).
*Please note:* The majority of this patch was written by Yury, and it's simply been rebased and slightly extended to prevent issues when dealing with `RenderingCancelledException`.
By leveraging streams this (finally) provides a simple way in which parsing can be aborted on the worker-thread, which will ultimately help save resources.
With this patch worker-thread parsing will *only* be aborted when the document is destroyed, and not when rendering is cancelled. There's a couple of reasons for this:
- The API currently expects the *entire* OperatorList to be extracted, or an Error to occur, once it's been started. Hence additional re-factoring/re-writing of the API code will be necessary to properly support cancelling and re-starting of OperatorList parsing in cases where the `lastChunk` hasn't yet been seen.
- Even with the above addressed, immediately cancelling when encountering a `RenderingCancelledException` will lead to worse performance in e.g. the default viewer. When zooming and/or rotation of the document occurs it's very likely that `cancel` will be (almost) immediately followed by a new `render` call. In that case you'd obviously *not* want to abort parsing on the worker-thread, since then you'd risk throwing away a partially parsed Page and thus be forced to re-parse it again which will regress perceived performance.
- This patch is already *somewhat* risky, given that it touches fundamentally important/critical code, and trying to keep it somewhat small should hopefully reduce the risk of regressions (and simplify reviewing as well).
Time permitting, once this has landed and been in Nightly for awhile, I'll try to work on the remaining points outlined above.
Co-Authored-By: Yury Delendik <ydelendik@mozilla.com>
Co-Authored-By: Jonas Jenwald <jonas.jenwald@gmail.com>
Furthermore, it's possible to re-use the same Array for all `textDiv`s on the page and the resulting padding string also becomes a lot more compact.
Please note that the `paddingLeft` branch was moved, since the padding values need to be ordered as `top, right, bottom, left`.
Finally, with this re-factoring it's no longer necessary to cache the original `style` string for every `textDiv` when `enhanceTextSelection` is enabled.
Given that browsers will reject padding values smaller than zero (which may be caused by limited numerical precision during calculations in the `expand` code), it makes no sense to include those when expanding the `textDiv`s.
Given that the different types of `Stream`s will never be cached, this thus implies that the `XRef.cache` Array will *always* be more-or-less sparse.
Generally speaking, the longer the document the more sparse the `XRef.cache` will thus become. For example, looking at the `pdf.pdf` file from the test-suite: The length of the `XRef.cache` Array will be a few hundred thousand elements, with approximately 95% of them being empty.
Hence it seems pretty clear that an Array isn't really the best data-structure for this kind of cache, and this patch thus changes it to a Map instead.
This patch-series was tested using the PDF file from issue 2618, i.e. http://bugzilla-attachments.gnome.org/attachment.cgi?id=226471, with the following manifest file:
```
[
{ "id": "issue2618",
"file": "../web/pdfs/issue2618.pdf",
"md5": "",
"rounds": 200,
"type": "eq"
}
]
```
which gave the following results when comparing this patch-series against the `master` branch:
```
-- Grouped By browser, stat --
browser | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ------------ | ----- | ------------ | ----------- | --- | ----- | -------------
Firefox | Overall | 200 | 2736 | 2736 | 1 | 0.02 |
Firefox | Page Request | 200 | 2 | 2 | 0 | -8.26 | faster
Firefox | Rendering | 200 | 2733 | 2734 | 1 | 0.03 |
```
The relevant methods are usually not hot enough for these changes to have an easily measurable effect, however there's been a lot of other cases where similiar inlining has helped performance. (And these changes may help offset the changes made in the next patch.)
For very large and complex PDF files this will help performance *slightly*, since `Parser.getObj` is called *a lot* during parsing in the worker.
This patch was tested using the PDF file from issue 2618, i.e. http://bugzilla-attachments.gnome.org/attachment.cgi?id=226471, with the following manifest file:
```
[
{ "id": "issue2618",
"file": "../web/pdfs/issue2618.pdf",
"md5": "",
"rounds": 200,
"type": "eq"
}
]
```
which gave the following results when comparing this patch against the `master` branch:
```
-- Grouped By browser, stat --
browser | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ------------ | ----- | ------------ | ----------- | --- | ----- | -------------
Firefox | Overall | 200 | 2847 | 2830 | -17 | -0.60 | faster
Firefox | Page Request | 200 | 2 | 2 | 0 | -7.14 |
Firefox | Rendering | 200 | 2844 | 2827 | -17 | -0.60 | faster
```
Looking at this again, it struck me that added functionality in `Util.intersect` is probably more confusing than helpful in general; sorry about the churn in this code!
Based on the parameter name you'd probably expect it to only match when the intersection is `[0, 0, 0, 0]` and not when only one component is zero, hence the `skipEmpty` parameter thus feels too tightly coupled to the `Page.view` getter.
This was added in PR 4470, but doesn't appear to have been used since.
While it's certainly easy to understand how this was helpful during development of that PR, actually providing this hash parameter isn't going to work anymore given that the original CMap files were also removed from the repository.
I suppose that the hash parameter *could* be useful if you'd attempt to update the BCMap files, however that hasn't been attempted even once in over *five* years time. Furthermore, at this point using the `AppOptions` directly in that situation should also work fine.
All in all, this seems like a piece of old and unused code which we can simply remove now.
This is based on a real-world PDF file I encountered very recently[1], although I'm currently unable to recall where I saw it.
Note that different PDF viewers handle these sort of errors differently, with Adobe Reader outright failing to render the attached PDF file whereas PDFium mostly handles it "correctly".
The patch makes the following notable changes:
- Refactor the `cropBox` and `mediaBox` getters, on the `Page`, to reduce unnecessary duplication. (This will also help in the future, if support for extracting additional page bounding boxes are added to the API.)
- Ensure that the page bounding boxes, i.e. `cropBox` and `mediaBox`, are never empty to prevent issues/weirdness in the viewer.
- Ensure that the `view` getter on the `Page` will never return an empty intersection of the `cropBox` and `mediaBox`.
- Add an *optional* parameter to `Util.intersect`, to allow checking that the computed intersection isn't actually empty.
- Change `Util.intersect` to have consistent return types, since Arrays are of type `Object` and falling back to returning a `Boolean` thus seem strange.
---
[1] In that case I believe that only the `cropBox` was empty, but it seemed like a good idea to attempt to fix a bunch of related cases all at once.
The current code will only consider the `cropBox` and `mediaBox` as equal when they both point to the *same* underlying Array. In the case where a PDF file actually specifies both boxes independently, with the exact same values in each, the comparison will currently fail and lead to an unneeded intersection computation.
By using the same internal formatting here as in the `Ref.toString` method, in `src/core/primitives.js`, all cache-keys will become at least two bytes shorter (and most three bytes shorter).
Obviously this won't have a huge effect on memory since there's only one cache entry per page, but it nonetheless seems wasteful to use longer keys than strictly required.
With the changes to the `StreamType`/`FontType` "enums" in PR 11029, one unfortunate result is that `getStats` now *always* returns empty Arrays. Something that everyone, myself included, apparently missed is that you obviously cannot index an Array with Strings :-)
I wrongly assumed that the unit-tests would catch any bugs, but they apparently suffered from the same issue as the code in `src/core/`.
Another possible option could perhaps be to use `Set`s, rather than objects, but that will require larger changes since `LoopbackPort` (in `src/display/api.js`) doesn't support them.
Firefox telemetry supports using string labels now. Convert our integers
that we used for categories to just use strings.
The upstream work will happen in:
https://bugzilla.mozilla.org/show_bug.cgi?id=1566882
There's a number of spots in the current code, and tests, where `cancel` methods are not called with appropriate arguments (leading to Promises not being rejected with Errors as intended).
In some cases the cancel `reason` is implicitly set to `undefined`, and in others the cancel `reason` is just a plain String. To address this inconsistency, the patch changes things such that cancelling is done with `AbortException`s everywhere instead.
Currently the indicator may remain visible even after the document has been closed, which seems weird given that no page is either visible nor rendering :-)
Add a work-around, in `glyphlist.js`, for bad PDF generators which use a non-standard `/f_f` string in the `Encoding` dictionary when referring to the ff ligature (issue 11016)
This patch will not incur any (measurable) overhead, since the glyphlist is already quite long and one more entry won't really matter, which is important given that this sort of PDF corruption ought to be very rare.
Furthermore, this patch purposely does *not* add a bunch of similarly modified ligature names on pure speculation. Any similar additions, for other ligatures, should only be made if there's real-world examples of PDF files where that's actually necessary.
For very large and complex PDF files this will help performance slightly, since `EvaluatorPreprocessor.read` is called a lot during parsing in the worker.
This patch was tested using the PDF file from issue 2618, i.e. http://bugzilla-attachments.gnome.org/attachment.cgi?id=226471, using the following manifest file:
```
[
{ "id": "issue2618",
"file": "../web/pdfs/issue2618.pdf",
"md5": "",
"rounds": 200,
"type": "eq"
}
]
```
This gave the following results when comparing this patch against the `master` branch:
```
-- Grouped By browser, stat --
browser | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ------------ | ----- | ------------ | ----------- | --- | ----- | -------------
Firefox | Overall | 200 | 3402 | 3358 | -43 | -1.28 | faster
Firefox | Page Request | 200 | 1 | 1 | 0 | 26.71 |
Firefox | Rendering | 200 | 3401 | 3357 | -44 | -1.28 | faster
```
Ensure that setting the `zoomDisabledTimeout` isn't skipped, regardless of the supported zoom keys, when handling mouse wheel events (PR 7097 follow-up)
*Possible follow-up:* It probably wouldn't hurt to try and shorten the `supportedMouseWheelZoomModifierKeys` name a bit, but I'm not attempting that here since it'd also require updating `PdfStreamConverter.jsm` in mozilla-central in order to be consistent.
For very large and complex PDF files this will help performance slightly, since `Parser.shift` is called *a lot* during parsing.
This patch was tested using the PDF file from issue 2618, i.e. http://bugzilla-attachments.gnome.org/attachment.cgi?id=226471 (with well over *four million* `Parser.shift` calls for just the one page), using the following manifest file:
```
[
{ "id": "issue2618",
"file": "../web/pdfs/issue2618.pdf",
"md5": "",
"rounds": 100,
"type": "eq"
}
]
```
This gave the following results when comparing this patch against the `master` branch:
```
-- Grouped By browser, stat --
browser | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ------------ | ----- | ------------ | ----------- | --- | ----- | -------------
Firefox | Overall | 100 | 3386 | 3322 | -65 | -1.92 | faster
Firefox | Page Request | 100 | 1 | 1 | 0 | -8.08 |
Firefox | Rendering | 100 | 3385 | 3321 | -65 | -1.92 | faster
```
Remove an unnecessary `PDFDocumentProperties.setFileSize` call, relevant for the Firefox built-in viewer, and use the "normal" code-path in `PDFViewerApplication.open` instead
Attempt to significantly reduce the number of `ChunkedStream.{ensureByte, ensureRange}` calls by inlining the `this.progressiveDataLength` checks at the call-sites
This functionality is very old, and pre-dates e.g. the introduction of the `EventBus` by a number of years. Rather than attaching two callback functions to every single `PDFPageView` instance, it's thus now possible to utilize the `EventBus` such that you only need a grand total of two listeners to achieve the same result.
For the `onAfterDraw` callback the replacement is particularly simple, given that a 'pagerendered' event is already being dispatched in the appropriate spot. An added benefit here is the ability to remove the event listener, since we only ever care about *one* (arbitrary) page being rendered for the `BaseViewer.onePageRendered` promise.
For the `onBeforeDraw` callback, a new 'pagerender' event was thus added to replace the callback.
Given that this special-case only matters for the Firefox PDF viewer, it's probably better to just move it into `firefoxcom.js` instead to reduce unnecessary confusion.
The way that this method handles documents without an `ID` entry in the Trailer dictionary feels overly complicated to me. Hence this patch adds `getByteRange` methods to the various Stream implementations[1], and utilize that rather than manually calling `ensureRange` when computing a fallback `fingerprint`.
---
[1] Note that `PDFDocument` is only ever initialized with either a `Stream` or a `ChunkedStream`, hence why the `DecodeStream.getByteRange` method isn't implemented.
The `finalize` helper function has only a *single* call-site, and furthermore it's just a one-liner too. Furthermore it's only ever called with a `Promise` as its argument, meaning that it's unnecessarily convoluted as well (i.e. the `Promise.resolve()` part shouldn't be necessary).
Hence this code can be both simplified *and* inlined at its only call-site instead.
Currently `wrapReason` is manually called at *every* `resolveOrReject` call-site, despite it being completely unnecessary unless there's an actual error being handled. This is obviously inefficient, and it's easy enough to avoid by having `resolveOrReject` handle this only when actually needed.
Similar to the `zoomReset` method we need to ensure that this code won't run for zoom events originating within the browser UI itself, since checks in e.g. the `keydown` event handler won't help in that case.
Note that, in the old code, there was a code-path which could prevent this from happening thus affecting future cleanup.
Furthermore, ensure that we'll always attempt to cleanup when handling the 'PageError' message, similar to the code in e.g. the `PDFPageProxy._renderPageChunk` method.
The `receivingOperatorList` property is currently tracked *twice* in the rendering code, both directly and inversely through the `intentState.operatorList.lastChunk` boolean. This type of double bookkeeping is never a good idea, since it's just too easy for the properties to accidentally fall out of sync.
In this case there's even a `cleanup`-related bug caused by this, which means that `PDFPageProxy._tryCleanup` will never be able to discard any data if there's an error on the worker-thread (as handled through the 'PageError' message).
Hence the simplest solution seems, at least to me, to update `PDFPageProxy._tryCleanup` to replace the `intentState.receivingOperatorList` check with a `!intentState.operatorList.lastChunk` check and completely remove the former property.
When searching occurs for the first time in a document, the `textContent` of all pages will be fetched from the API. If there's a pending search operation when the document loads that will thus lead to a lot of `getTextContent` calls very early on, which may unnecessarily delay rendering of the first page. Generally, in the viewer, a number of non-essential API calls[1] will be deferred until the first page has been rendered, and there's no good reason as far as I can tell to handle searching differently.
---
[1] Such as e.g. `getOutline` and `getAttachments`.
*Please note:* A a similar change was attempted in PR 5005, but it was subsequently backed out in PR 5069.
Unfortunately I don't think anyone ever tried to debug *exactly* why it didn't work, since it ought to have worked, and having re-tested this now I'm not able to reproduce the problem any more. However, given just how inefficient the current code is, with thousands of strictly unnecessary function calls for each `find` invocation, I'd really like to try fixing this again.
This reduces the total number of function calls, when reading the XRef table respectively when fetching uncompressed XRef entries.
Note in particular the `XRef.readXRefTable` method, where there're *two* back-to-back `isCmd` checks rather than just one.
A lot of the `new Parser()` call-sites look quite unwieldy/ugly as-is, with a bunch of somewhat randomly ordered arguments, which we can avoid by changing the constructor to accept an object instead. As an added bonus, this provides better documentation without having to add inline argument comments in the code.
See https://github.com/mozilla/eslint-plugin-no-unsanitized
Since we've generally never allowed e.g. `innerHTML`, which is enforced during review, there's only one linting failure with this patch. (Which is white-listed, according to the existing comment and the fact that it's test-only code.)
Since all other `IPDFStream` implementations live in their own files, it seems reasonable for these to do so as well.
Furthermore, converts all of the relevant code to ES6 classes and updates the interface definitions to mark a couple of methods `async`.
Unless the `PDFLinkService` instance contains all of the expected methods, a lot of things will break in various places in the default viewer. Hence there's not much value in having this check, and outright falling seems more appropriate.
Finally, this also makes the return value explicit in this case, since that's consistent with the rest of the `PDFFindController._shouldDirtyMatch` method.
As have already been stated multiple times, simply increasing the printing resolution may have undesirable effects on both memory usage *and* general performance. Hence why PR 10854 did *not* add a preference, and only exposed AppOption by default in `GENERIC` builds for now.
However, considering how differently printing works in the built-in Firefox version (with `mozPrintCallback`) compared to the general default viewer, any testing done in the latter case might not be completely relevant to the first (and most important) case of the Firefox PDF Viewer.
Note that considering the implementation of `AppOptions.get`, this patch will be safe and should allow experimenting with `printResolution` in all builds of the default viewer[1]. By not, however, having `printResolution` appear in AppOptions for either the `MOZCENTRAL` or `CHROMIUM` build targets, there should be no indication of official support for now.
Furthermore, it shouldn't be a preference at this point in time (or even at all), since that makes it too easy for users to change it permanently[2] and possible "break" printing.
---
[1] By running `PDFViewerApplicationOptions.get('printResolution', /* value here */);` in the console after the viewer loads.
[2] I've seen Firefox bugs, filed in Bugzilla, where users modified e.g. preferences manually in `about:config` and then some time later (maybe months) wondered why something was suddenly broken. In those cases, trying to work out that a preference change was the culprit can take time/effort.
Given that `cleanupAfterRender` is already set for large images, when handling 'obj' messages, this patch *should* thus be safe in general (since otherwise there ought be existing bugs related to cleanup and printing).
The border `width` will instead fallback to the default value of `1`, rather than ignoring it altoghether, to also ensure that e.g. `LinkAnnotation`s become clickable as intended.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1552113
Usually when the worker is terminated it will also be completely destroyed/removed, which means that any global caches (such as the ones in `src/core/primitive.js`) should be automatically cleared in the process.
However, for certain ways of loading the `pdf.worker.js` file, e.g. passing in a re-usable worker to `getDocument`, using the `workerPort` functionality, or even disabling workers completely (even though this is never a good idea), the worker file may be kept in memory and these caches will not be cleared as expected.
Calling `someArray = []` will create a new Array, which seems completely unnecessary when it's sufficient to just call `someArray.length = 0` to achieve the same effect.
Even though I cannot imagine these particular cases having any noticeable performance impact, similar changes were made in `core/` code years ago since it's apparently more efficient memory wise.
The purpose of these caches is to reduce peak memory usage, by only ever having *a single* instance of a particular object.
However, as-is these caches are never cleared and they will thus remain until the worker is destroyed. This could very well have a negative effect on total memory usage, particularly for large/long documents, hence it seems to make sense to clear out these caches together with various other ones.
This is similar to the existing caching used to reduced the number of `Cmd` and `Name` objects.
With the `tracemonkey.pdf` file, this patch changes the number of `Ref` objects as follows (in the default viewer):
| | Loading the first page | Loading *all* the pages |
|----------|------------------------|-------------------------|
| `master` | 332 | 3265 |
| `patch` | 163 | 996 |
The specification states that `CreationDate` is only available for
markup annotations instead of for all annotation types.
Moreover, popup annotations are not markup annotations according to the
specification, so the creation date inheritance from the parent
annotation is also removed there (note that only the modification date
is used in e.g., the viewer).
This includes the information in the core and display layers. The
date parsing logic from the document properties is rewritten according
to the specification and now includes unit tests.
Moreover, missing unit tests for the color of a popup annotation have
been added.
Finally the styling of the popup is changed slightly to make the text a
bit smaller (it's currently quite large in comparison to other viewers)
and to make the drop shadow a bit more subtle. The former is done to be
able to easily include the modification date in the popup similar to how
other viewers do this.
Currently `handleColorN` will fallback to add a completely unparsed/unvalidated operator when no valid pattern was found. This is unfortunate, since it could very easily lead to a couple of different errors:
- `DataCloneError`s when attempting to send the data to the main-thread, e.g. when `args` is `Dict`/`Stream`.
- Errors in `getShadingPatternFromIR` on the main-thread, unless `args` just happens to have the expected format.
- Errors when actually attempting to render the pattern on the main-thread, since the `args` will most likely not have the expected format.
Hence it probably makes sense to error in `PartialEvaluator.handleColorN`, and having invalid patterns fail gracefully via the existing `ignoreErrors` code-paths instead.
It appears that this has been broken ever since PR 9089, which also introduced this code, since the `QueueOptimizer`/`NullOptimizer` choice was made based on the still undefined `this.intent` property.
Furthermore, fixing this also uncovered the fact that the `NullOptimizer.reset` method was missing.
First of all, while this simple approach appears to work OK in practice I'm not sure if it's the best way of addressing the problem (assuming that you even want to).
Second of all, while the solution implemented here only requires tracking/checking one new boolean in order for this to work, I'm nonetheless not entirely happy about this since it will add additional overhead (albeit *very* small) to the parsing of path operators in PDF documents just for a handful of *corrupt* ones.
This way we can avoid manually building a "document id" in multiple places in `evaluator.js`, and it also let's us avoid passing in an otherwise unnecessary `PDFManager` instance when creating a `PartialEvaluator`.
While PR 10714 did address the `disableRange=true` case, it also managed to "break" the `disableStream=true` case instead since the indeterminate loadingBar is now displayed when it shouldn't; sorry about that!
The solution is simple enough though, don't attempt to fallback to `_fullRequestReader.onProgress` when handling "incomplete" loading information.
Please see the specification, https://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#M11.9.12864.1Heading.71.Viewer.Preferences
Furthermore, note that this patch *only* adds API support and unit-tests but does not attempt to integrate e.g. the `ViewerPreferences -> Direction` property into the viewer (which would be necessary to address issue 10736).
The reason for this is that it's not entirely clear to me exactly if/how that could be implemented; e.g. would it be as simple as setting the `dir` attribute on the `viewerContainer` DOM element, or will it be more complicated?
There's also the question of how the `ViewerPreferences -> Direction` value interacts with the `PageMode`, and this will generally require a fair bit of manual testing. Since the direction of the *entire* viewer depends on the browser locale, there's also a somewhat open question regarding what default value to use for different locales.
Finally, if the viewer supports `ViewerPreferences -> Direction` then I'm assuming that it will be necessary to allow users to override the default value, which will require (most likely) new `SecondaryToolbar` buttons and icons for those etc.
Hence this patch only lays the necessary foundation for eventually addressing issue 10736, but defers the actual implementation until later. (Time permitting, I'll try to look into the viewer part later.)
*Please note:* This patch purposely ignores `src/display/network.js`, since its support for progressive reading depends on the non-standard `moz-chunked-arraybuffer` responseType which is currently in the process of being removed.
The file `test/pdfs/annotation-caret-ink.pdf` is already available in
the repository as a reference test for this since I supplied it for
another patch that implemented ink annotations.
This mirrors the canvas implementation where we ignore these operators.
This avoids console spam regarding unimplemented operators we're not
interested in.
For the Tracemonkey paper, we're now down to one warning about tiling
patterns which is in fact a valid one.
In particular, this should reduce intermediate string creation by using
template strings and reduce variable lookup times by removing unneeded
variables and caching `this.current` in more places.
With PR 10675 having fixed the completely broken `disableRange=true` setting in the Firefox version of PDF.js, I couldn't help but noticing that loading progress is never reported properly in that case.
Currently loading progress is only reported for the `rangeProgress` chrome-event, which obviously isn't dispatched with `disableRange=true` set. However, the `progressiveRead` chrome-event includes loading progress as well, but this information isn't being used in any way.
Furthermore, the `PDFDataRangeTransport.onDataProgress` method wasn't able to handle "complete" loading information, and neither was `PDFDataTransportStream._onProgress` since that method would only ever attempt to report it through a RangeReader (which won't exist when `disableRange=true` is set).
Note how at https://mozilla.github.io/pdf.js/api/ it's being described as API docs, however `src/core/annotation.js` is not part of the public API.
Furthermore, given that the code residing in the `src/core/` folder is run in a worker-thread, it's not even accessible on the main-thread (since `postMessage` is being used to transfer the data).
Hence the different API methods simply returns a "proxy" to the underlying data, but not actually the same objects and data structures as in the worker-thread itself; thus it doesn't make a whole lot of sense to expose this in API docs as far as I'm concerned.
Finally, the patch fixes a small JSDoc related typo in `src/display/api.js` when referring to the `TextStyle` typedef.
The `moz-chunked-arraybuffer` responseType is a non-standard property, which has been subsumed by the Fetch API, and it's in the process of being removed from Firefox; please see https://bugzilla.mozilla.org/show_bug.cgi?id=1120171 and https://bugzilla.mozilla.org/show_bug.cgi?id=1411865
*Please note:* Rather than waiting for both `Fetch` *and* `ReadableStream` to be available in e.g. a Firefox ESR version (which is probably going to be 68 at the earliest), let's just decide that PDF.js release `2.1.266` will be the last one with `moz-chunked-arraybuffer` support and land this patch (since nothing should outright break without it anyway).
Note how `XRef.fetchUncompressed`, which is used *a lot* for most PDF documents, is calling the `makeSubStream` method without providing a `length` argument.
In practice this results in the `makeSubStream` method, on the `ChunkedStream` instance, calling the `ensureRange` method with `NaN` as the end position, thus resulting in no data being requested despite it possibly being necessary.
This may be quite bad, since in this particular case it will lead to a new `ChunkedStream` being created *and* also a new `Parser`/`Lexer` instance. Given that it's quite possible that even the very first `Parser.getObj` call could throw `MissingDataException`, this could thus lead to wasted time/resources (since re-parsing is necessary once the data finally arrives).
You obviously need to be very careful to not have `ChunkedStream.makeSubStream` accidentally requesting the *entire* file, hence its `this.end` property is of no use here, but it should be possible to at least check that the `start` of the data is present before any potentially expensive parsing occurs.
This transform resulted in an incorrectly positioned object when the
bounding box's upper-left corner did not start at (0,0), because
the translation was not reverted. This patch adds the missing transform.
The test file (tiling-pattern-box.pdf) is based on the PDF from #2825.
All but the first cube (including the PDF data) have been removed.
To trigger the bug that is fixed by this commit, I changed the BBox of
the first pattern from "[ 0 0 596 842]" to "[90 0 596 842]". Without
this patch, the dashed vertical line that intersects the corners at A
and E would disappear.
The new test file (tiling-pattern-large-steps.pdf) was manually created,
to have the following characteristics:
- Large xstep and ystep (90000)
- Page width is 4000 (which is larger than MAX_PATTERN_SIZE)
- Visually, the page consists of a red rectangle with a black border,
surrounded by a 50 unit white padding.
- Before patch: blurry; After patch: sharp
Fixes#6496Fixes#5698Fixes#1434Fixes#2825
Without this some fonts may incorrectly end up with matching `hash`es, thus breaking rendering since we'll not actually try to load/parse some of the fonts.
Note that `PartialEvaluator.preEvaluateFont` will return an empty string when no hash was computed. This will complete short-circuit the `fontAlias` comparison in `PartialEvaluator.loadFont`, since fonts which are totally different will then match if their `hash`es are empty.
Currently if trying to set `disableRange=true` in the built-in PDF Viewer in Firefox, either through `about:config` or via the URL hash, the PDF document will never load. It appears that this has been broken for a couple of years, without anyone noticing.
Obviously it's not a good idea to set `disableRange=true`, however it seems that this bug affects the PDF Viewer in Firefox even with default settings:
- In the case where `initialData` already contains the *entire* file, we're forced to dispatch a range request to re-fetch already available data just so that file loading may complete.
- (In the case where the data arrives, via streaming, before being specifically requested through `requestDataRange`, we're also forced to re-fetch data unnecessarily.) *This part was removed, to reduce the scope/risk of the patch somewhat.*
In the cases outlined above, we're having to re-fetch already available data thus potentially delaying loading/rendering of PDF files in Firefox (and wasting resources in the process).
Support for the non-standard `moz-chunked-arraybuffer` response type is in the process of being removed from Firefox; see e.g. https://bugzilla.mozilla.org/show_bug.cgi?id=1411865
For the time being, you probably want to keep support for this in the general PDF.js library given that feature detection is used. However, removing the unit-test immediately seems reasonable, since it will otherwise start failing once the platform support for `moz-chunked-arraybuffer` is gone.
Fixes 8851; please note that if unit-tests for the code in `fetch_stream.js` are wanted, which I'm assuming they are, those should live in their own file rather than being lumped into `network_spec.js` anyway.
This lays the necessary foundation for handling zoom events originating within the browser itself, rather than in the viewer. Please note that this will also require a follow-up patch to `mozilla-central`, such that the viewer is actually notified when zooming occurs.
Moreover, group the lexer unit tests per method. This matches what we
do for other classes and makes it more easily visible which methods
we don't or insufficiently unit test.
The parser itself is not unit tested yet, so this patch provides a start
for doing so. The `inlineStreamSkipEI` method is used in other end
marker detection methods, so it's important that its functionality is
correct for proper parsing.
This function is currently called with the `OperatorList` instance as its argument, hence I cannot think of any good reason for not just moving it into the `OperatorList` properly. (This will also help with other planned changes regarding the `ImageCache` functionality.)
This will further help reduce the amount of image data that's currently being held alive, by explicitly removing the `src` attribute.
Please note that this is mostly relevant for browsers which do not support `URL.createObjectURL`, or where `disableCreateObjectURL` was manually set by the user, since `blob:` URLs will be revoked (see the previous patch).
However, using `about:memory` (in Firefox) it does seem that this may also be generally helpful, given that calling `URL.revokeObjectURL` won't invalidate the image data itself (as far as I can tell).
Natively supported JPEG images are sent as-is, using a `blob:` or possibly a `data` URL, to the main-thread for loading/decoding.
However there's currently no attempt at releasing these resources, which are held alive by `blob:` URLs, which seems unfortunately given that images can be arbitrarily large.
As mentioned in https://developer.mozilla.org/en-US/docs/Web/API/URL/createObjectURL the lifetime of these URLs are tied to the document, hence they are not being removed when a page is cleaned-up/destroyed (e.g. when being removed from the `PDFPageViewBuffer` in the viewer).
This is easy to test with the help of `about:memory` (in Firefox), which clearly shows the number of `blob:` URLs becomming arbitrarily large *without* this patch. With this patch however the `blob:` URLs are immediately release upon clean-up as expected, and the memory consumption should thus be considerably reduced for long documents with (simple) JPEG images.
Note how `PDFDocumentProxy.destroy` is a nothing more than an alias for `PDFDocumentLoadingTask.destroy`. While removing the latter method would be a breaking API change, there's still room for at least some clean-up here.
The main changes in this patch are:
- Stop providing a `PDFDocumentLoadingTask` instance *separately* when creating a `PDFDocumentProxy`, since the loadingTask is already available through the `WorkerTransport` instance.
- Stop tracking the `PDFDocumentProxy` instance on the `WorkerTransport`, since that property is completely unused.
- Simplify the 'Multiple `getDocument` instances' unit-tests by only destroying *once*, rather than twice, for each document.
For Type3 fonts text-selection is often not that great, and there's a couple of heuristics used to try and improve things. This patch simple extends those heuristics a bit, and fixes a pre-existing "naive" array comparison, but this all feels a bit brittle to say the least.
The existing Type3 test-coverage isn't that great in general, and in particular Type3 `text` tests are few and far between, hence why this patch adds *two* different new `text` tests.
Notable changes:
- Remove the `return this;` from the `MurmurHash3_64.update` method, since it's completely unused and doesn't make a lot of sense.
- Remove the loop(s) from the `MurmurHash3_64.hexdigest` method, since creating a temporary array and then looping over it is wasteful given how simple this can be written with modern JavaScript.
The `DOMCanvasFactory` class is now fully covered. Moreover, missing
cases for the `getFilenameFromUrl` function have been included.
Finally, `var` usage has been removed.
The generated `default_preferences.json` file is necessary when initializing the Firefox preferences, which only supports certain types, hence this patch adds additional validation to help prevent run-time errors in Firefox.
Given that these changes add a code-path to `AppOptions.getAll` which could throw, the `OptionKind.PREFERENCE` branch is also modified to require *exact* matching to prevent (future) errors in the viewer.
Finally the conditionally defined `defaultOptions` will no longer (potentially) be considered during the `gulp default_preferences` task, to make it more difficult for them to be accidentally included.
Given that the function is (purposely) independent of the verbosity level and that its message is worded to only apply on the main-thread, there's no reason to duplicate this across the built `pdf.js`/`pdf.worker.js` files.
Currently for every single parsed/rendered page there's no less than *four* `Date.now()` calls being made on the worker-side. This seems totally unnecessary, since the result of these calls are, by default, not used for anything *unless* the verbosity level is set to `INFO`.
The default size of these canvases seem to be `300 x 150` (two orders of magnitude larger than the ones in PR 10597), which probably is sufficient enough to matter since there's one such canvas for each textLayer that's rendered in the viewer.
Also fixes the incorrect rejection reason, i.e. one using a string rather than an `Error`, in the `TextLayerRenderTask.cancel` method.
While this particular canvas may be small, there can still be an arbitrarily large number of them (one per page rendered), which can/will eventually add up memory wise. This can be easily avoided by using the `cachedCanvases` abstraction instead, which will ensure that the `isFontSubpixelAAEnabled` canvas is removed together with other temporary canvases in `CanvasGraphics.endDrawing`.
The `src/shared/util.js` file is being bundled into both the `pdf.js` and `pdf.worker.js` files, meaning that its code is by definition duplicated.
Some main-thread only utility functions have already been moved to a separate `src/display/display_utils.js` file, and this patch simply extends that concept to utility functions which are used *only* on the worker-thread.
Note in particular the `getInheritableProperty` function, which expects a `Dict` as input and thus *cannot* possibly ever be used on the main-thread.
In order to, quickly, unblock future Webpack updates this patch implements a *temporary* work-around for the regression; please refer to issue 10177 for additional details.
Please note that this patch doesn't require *any* changes to the source code, but only (small) build-time changes to `gulpfile.js` and as such will be easy to revert.
This file (currently) contains not only DOM-specific helper functions/classes, but is used generally for various helper code relevant for main-thread functionality.
*Hopefully this patch makes sense, since I cannot claim to fully understand this function.*
With the changes made in PR 3354 *some* Type3 glyph outlines are no longer rendering correctly, since the final paths were being accidentally ignored.
The fact that Type3 fonts are not very common in PDF documents, and that most Type3 glyphs are unaffected by this regression, probably explains why this has gone unnoticed since 2013.
There doesn't appear to be any particular reason for only running these unit-tests in browsers, since the `PDFDataRangeTransport` functionality itself should be back-end agnostic.
This allows simplification of the 'creates pdf doc from URL and aborts loading after worker initialized' API unit-test.
Note that the `DOMFileReaderFactory` uses the Fetch API, for simplicity, since it should be available in all browsers where we're running tests.
Currently these methods are only used from the respective `reset` methods, and from `{BaseViewer, PDFThumbnailViewer}._cancelRendering` which only runs when the active document is closed.
This patch changes `{PDFPageView, PDFThumbnailView}.cancelRendering` to *only* cancel any pending rendering operations, and doesn't attempt to reset e.g. the `renderingState`, since that causes visual glitches (duplicated canvases in the viewer) when called directly.
Furthermore, unless you "know" what you're doing, the `{PDFPageView, PDFThumbnailView}.reset` methods are what *should* normally be used instead.
This implements the nice suggestion from https://github.com/mozilla/pdf.js/pull/10099#discussion_r219698000, which I at the time didn't think would work.
I'll probably have to plead temporary insanity here, since it *should* have been totally obvious to me how this could be implemented. By simply not registering the event until the `textLayer` is actually rendered, removing the event on `cancel` works just fine.
This patch also removes the `pagecancelled` event, given that it's no longer used anywhere in the code-base and that its implemention was flawed since it wasn't guaranteed to be called in *every* situation where rendering was actually cancelled.
Currently any editing of the preferences require updates in *three* separate files, which isn't a great developer experience to say the least.
This has annoyed me sufficiently to write this patch, which moves the definition of all preferences into `AppOptions` and adds a new `gulp` task to generate the `default_preferences.json` file for the builds where it's needed.
After PR 9340 all glyphs are now re-mapped to a Private Use Area (PUA) which means that if a font fails to load, for whatever reason[1], all glyphs in the font will now render as Unicode glyph outlines.
This obviously doesn't look good, to say the least, and might be seen as a "regression" since previously many glyphs were left in their original positions which provided a slightly better fallback[2].
Hence this patch, which implements a *general* fallback to the PDF.js built-in font renderer for fonts that fail to load (i.e. are rejected by the sanitizer). One caveat here is that this only works for the Font Loading API, since it's easy to handle errors in that case[3].
The solution implemented in this patch does *not* in any way delay the loading of valid fonts, which was the problem with my previous attempt at a solution, and will only require a bit of extra work/waiting for those fonts that actually fail to load.
*Please note:* This patch doesn't fix any of the underlying PDF.js font conversion bugs that's responsible for creating corrupt font files, however it does *improve* rendering in a number of cases; refer to this possibly incomplete list:
[Bug 1524888](https://bugzilla.mozilla.org/show_bug.cgi?id=1524888)
Issue 10175
Issue 10232
---
[1] Usually because the PDF.js font conversion code wasn't able to parse the font file correctly.
[2] Glyphs fell back to some default font, which while not accurate was more useful than the current state.
[3] Furthermore I'm not sure how to implement this generally, assuming that's even possible, and don't really have time/interest to look into it either.
- The only existing call-site, of this method, is never passing more than *one* font at a time anyway.
- As far as I can remember, this functionality has never actually been used (caveat: I didn't check the git history).
- This allows simplification of the method, especially by making use of the fact that it's now asynchronous.
- It should be just as easy to call `BaseFontLoader.bind` from within a loop, rather than having the loop in the method itself.
Currently all fonts are using the `_queueLoadingCallback` method to determine when they have been loaded[1]. However in most cases this is just adding unnecessary overhead, especially with `BaseFontLoader.bind` now being asynchronous, given how fonts are loaded:
- For fonts loaded using the Font Loading API, it's already possible to easily tell when a font has been loaded simply by checking the `loaded` promise on the FontFace object itself.
- For browsers, e.g. Firefox, which support synchronous font loading it's already assumed that fonts are immediately available.
Hence the `_queueLoadingCallback` method is moved into the `GenericFontLoader`, such that it's only utilized for fonts which are loaded using CSS.
---
[1] In the "fonts loaded using CSS" case, this is already a hack anyway as outlined in the comments.
pdf.js had a problem when copying characters on supplementary planes
(0xPPXXXX where PP is nonzero). This is because certain methods of
PartialEvaluator use classic String.fromCharCode instead of ES6's
String.fromCodePoint.
Despite the fact that readToUnicode method *tried* to parse out-of-UCS2
code points by parsing UTF-16BE, it was inadequate because
String.fromCharCode only supports UCS-2 range of Unicode.
Unsurprisingly IE11 doesn't support this, so a polyfill is needed since otherwise the sidebar can no longer be opened.
Also, simplifies the existing `classList.toggle` polyfill.
There's a bunch of code, in the viewer, which for historical reasons use `switch` statements to add and remove CSS classes.
This code can be simplified, and unnecessary duplication avoided, by using `classList.toggle` instead.
This polyfill is currently used in only *one* file, i.e. `src/display/api.js`, and only when trying to build a *fallback* `workerSrc` path.
Given that the global `workerSrc` should *always* be set[1] when using the PDF.js library[2], and that the fallback `workerSrc` should only be regarded as a best-effort solution anyway, there isn't a particularily strong reason to keep the compatibility code in my opinion.
---
[1] Other supported options include setting the global `workerPort`, or passing in a `PDFWorker` instance as part of the `getDocument` call.
[2] Which is clearly mentioned in the JSDocs in `src/display/worker_options.js`.
This avoids having the initialization code "spread out", and will become even simpler once the `TODO` is addressed (which I'm planning on fixing as soon as possible).
This patch ignores the recently added `disableOpenActionDestination` preference, since the latest PDF.js version found on the "Chrome Web Store" doesn't include it.
The intention with preferences such as `sidebarViewOnLoad`/`scrollModeOnLoad`/`spreadModeOnLoad` were always that they should be able to *unconditionally* override their view history counterparts.
Due to the way that these preferences were initially implemented[1], trying to e.g. force the sidebar to remain hidden on load cannot be guaranteed[2]. The reason for this is the use of "enumeration values" containing zero, which in hindsight was an unfortunate choice on my part.
At this point it's also not as simple as just re-numbering the affected structures, since that would wreak havoc on existing (modified) preferences. The only reasonable solution that I was able to come up with was to change the *default* values of the preferences themselves, but not their actual values or the meaning thereof.
As part of the refactoring, the `disablePageMode` preference was combined with the *adjusted* `sidebarViewOnLoad` one, to hopefully reduce confusion by not tracking related state separately.
Additionally, the `showPreviousViewOnLoad` and `disableOpenActionDestination` preferences were combined into a *new* `viewOnLoad` enumeration preference, to further avoid tracking related state separately.
The new "private" method will return a boolean, indicating if the `sidebarviewchanged` event was dispatched, thus allowing some simplification of the `PDFSidebar.setInitialView` method.
When `firstVisibleElementInd === 0`, regardless of the number of views, there's no reason to attempt to backtrack at all since it's never possible to find an element before the *first* one anyway.
This piggybacks of the existing `cancel` functionality, to ensure that any pending operations are closed *and* that any temporary canvases are actually being removed.
Also simplifies `finishPaintTask` in `PDFPageView.draw` slightly, by converting it to an async function.
All objects in the PDF document follow this pattern:
```
0000000001 0 obj
<<
% Some content here...
>>
endobj
0000000002 0 obj
<<
% More content here...
endobj
```
This attempts to provide more "default" methods in the base class, in order to reduce unnecessary duplication and to improve self-documentation of the `BaseViewer` class slightly.
The following changes are made (in no particular order):
- Have `BaseViewer` implement the `_scrollIntoView` method, and *extend* it as necessary in `PDFViewer`/`PDFSinglePageViewer`.
- Simply inline the `BaseViewer._resizeBuffer` method, in `BaseViewer.update`, since there's only one call-site at this point.
- Provide a default implementation of `_isScrollModeHorizontal` in `BaseViewer`, and have `PDFSinglePageViewer` override it.
- Provide a default implementation of `_getVisiblePages`, and have `PDFViewer` extend it and `PDFSinglePageViewer` override it.
Based on the discussion in https://bugzilla.mozilla.org/show_bug.cgi?id=1521413, this patch simply removes the `ReadableStream` polyfill completely from MOZCENTRAL builds.
With this patch, the size of the `gulp mozcentral` build target is thus further reduced (building on PR 10470):
| | `build/mozcentral`
|-------|-------------------
|master | 3 339 666
|patch | 3 209 572
The `FontInspector` was completely broken by PR 10197, and trying to select a particular font (using the checkboxes) or clicking on a piece of text currently does nothing.
With https://bugzilla.mozilla.org/show_bug.cgi?id=1505122 landing in Firefox 65, the native `ReadableStream` implementation is now enabled by default in Firefox.
Obviously it would be nice to simply stop bundling the polyfill in MOZCENTRAL builds altogether, however given that it's still possible to disable[1] `ReadableStream` this is probably not a good idea just yet.
Nonetheless, now that native support is available, it seems unnecessary (and wasteful) to keep bundling the polyfill twice[2] in MOZCENTRAL builds. Hence this patch, which contains a suggest approach for packing the polyfill in a *separate* file which is then *only* loaded if/when needed.
With this patch, the size of the `gulp mozcentral` build target is thus reduced accordingly:
| | `build/mozcentral`
|-------|-------------------
|master | 3 461 089
|patch | 3 340 268
Besides the PDF.js files taking up less space in Firefox this way, the additional benefit is that there's (by default) less code that needs to be loaded and parsed when the PDF Viewer is used which also cannot hurt.
---
[1] In `about:config`, by toggling the `javascript.options.streams` preference.
[2] Once in the `build/pdf.js` file, and once in the `build/pdf.worker.js` file.
In many cases in the code you don't actually care about the index itself, but rather just want to know if something exists in a String/Array or if a String starts in a particular way. With modern JavaScript functionality, it's thus possible to remove a number of existing `indexOf` cases.
Since most of the important rendering code is already (almost) identical between `PDFViewer.update` and `PDFSinglePageViewer.update`, it's possible to further reduce duplication by moving the code into `BaseViewer.update` instead.
Apparently in IE 11 when `history.{pushState, replaceState}` is called, there's actually a difference between not providing the *third* argument vs providing it set implicitly to `undefined`. It appears that in IE 11 it's actually being stringified, rather than ignored, which seems completely wrong (obviously other browsers aren't affected, so no surprises there).
This is yet another reason why I think the feature itself was a really bad idea, since it now requires extra/duplicated code just to prevent weird/incorrect URL behaviour in crappy browsers.
This will allow the Metadata to be successfully extracted from the PDF file in issue 10395.
Furthermore, this patch also fixes a bug in `Metadata.get` which causes the method to return `null` rather than an empty string or zero (since either ought to be allowed).
Previously a couple of different attempts at fixing this problem has been rejected, given how *crucial* this code is for the correct function of the viewer, since no one has thus far provided any evidence that the problem actually affects the default viewer[1] nor an example using the viewer components directly (without another library on top).
The fact that none of the prior patches contained even a *simple* unit-test probably contributed to the unwillingness of a reviewer to sign off on the suggested changes.
However, it turns out that it's possible to create a reduced test-case, using the default viewer, that demonstrates the error[2]. Since this utilizes a hidden `<iframe>`, please note that this error will thus affect Firefox as well.
Note that while errors are thrown when the hidden `<iframe>` loads, the default viewer doesn't break completely since rendering does start working once the `<iframe>` becomes visible (although the errors do break the initial Toolbar state).
Before making any changes here, I carefully read through not just the immediately relevant code but also the rendering code in the viewer (given it's dependence on `getVisibleElements`). After concluding that the changes should be safe in general, the default viewer was tested without any issues found. (The above being much easier with significant prior experience of working with the viewer code.)
Finally the patch also adds new unit-tests, one of which explicitly triggers the relevant code-path and will thus fail with the current `master` branch.
This patch also makes `PDFViewerApplication` slightly more robust against errors during document opening, to ensure that viewer/document initialization always completes as expected.
Please keep in mind that even though this patch prevents an error in `getVisibleElements`, it's still not possible to set the initial position/zoom level/sidebar view etc. when the viewer is hidden since rendering and scrolling is completely dependent[3] on being able to actually access the DOM elements.
---
[1] And hence the PDF Viewer that's built-in to Firefox.
[2] Copy the HTML code below and save it as `iframe.html`, and place the file in the `web/` folder. Then start the server, with `gulp server`, and navigate to http://localhost:8888/web/iframe.html
```html
<!DOCTYPE html>
<html>
<head>
<title>Iframe test</title>
<script>
window.onload = function() {
const button = document.getElementById('button1');
const frame = document.getElementById('frame1');
button.addEventListener('click', function(evt) {
frame.hidden = !frame.hidden;
});
};
</script>
</head>
<body>
<button id="button1">Toggle iframe</button>
<br>
<iframe id="frame1" width="800" height="600" src="http://localhost:8888/web/viewer.html" hidden="true"></iframe>
</body>
</html>
```
[3] This is an old, pre-exisiting, issue that's not relevant to this patch as such (and it's already being tracked elsewhere).
With modern JavaScript supporting block-scoped variables, it's no longer necessary to have large numbers of top-level variable definitions (especially when all variables are, implicitly, set to `undefined`).
Besides moving variable definintions, a number of them are also converted to `const` to help ensure that they cannot be *accidentally* modified in the code.
Finally, the `views.length` calculation is now cached *once* rather than being re-computed multiple times.
Handle more cases of corrupt PDF files with missing 'endobj' operators, where the "obj" string is immediately followed by the dictionary (PR 9288 follow-up)
This is *really* the best that we can do here, since other proposed solutions would interfere with (and break) the painstakingly implemented browsing history that's present in the default viewer.
I'm still not convinced that this is a good idea in general, but this patch implements it in a way where it is possible to toggle[1] for users that wish to have this feature. In particular, there's a couple of reasons why I'm not finding this feature necessary/great:
- It's already possible to easily obtain the current hash, by simply clicking on the `viewBookmark` button at any time.
- Hash changes requires a bit of special handling[2], i.e. extra code, to prevent issues when the browser history is traversed (see `PDFHistory._popState`). Currently this is only necessary when the user has manually changed the hash, with this patch it will always be the case (assuming the feature is active).
- It's not always possible to change the URL when updating the browser history. For example: In the Firefox built-in viewer, the URL cannot be modified for local files (i.e. those using the `file://` protocol).
This leads to inconsistent behaviour, and may in some cases even result in errors being thrown and the history thus not updating, if the browser prevents changes to the URL during `pushState`/`replaceState` calls.
---
[1] Using the `historyUpdateUrl` viewer preference.
[2] This depends, to a great extent, on browsers always firing `popstate` events *before* `hashchange` events, which may or may not actually be guaranteed.
This should hopefully be sufficient to address issue 6847, and given the limited impact of the code changes I'm not completely sure if this would need to be controlled by a preference!?
Initially my intention was to try and provide some (slightly more detailed) implementation suggestions in the issue, but having looked briefly at doing that it would essentially have amounted to actually writing the code anyway. (Especially considering that the recent questions seemed to more-or-less ignore the information already provided in the first post.)
Finally, note that since `performance.navigation.type` is marked as deprecated, a slightly different approach was choosen instead.
The error was triggered for a particular set of metadata, where an end tag was encountered without the corresponding begin tag being present in the data.
(The patch also fixes a minor oversight, from a recent PR, in the `SimpleDOMNode.nextSibling` method.)
Given that the issue, as filed, is incomplete since no PDF file was provided for debugging, this patch is really the best that we can do here. *Please note:* This patch will *not* enable the Metadata to be successfully parsed, but it should at least prevent the errors.
This method creates quite a few intermediate strings on each call and
it's called often, even for smaller documents like the Tracemonkey
document. Scrolling from top to bottom in that document resulted in
14126 strings being created in this method. With this commit applied,
this is reduced to 2018 strings.
This method creates quite a few intermediate strings on each call and
it's called often, even for smaller documents like the Tracemonkey
document. Scrolling from top to bottom in that document resulted in
12936 strings being created in this method. With this commit applied,
this is reduced to 3610 strings.
The `toString` method always creates two string objects (for the 'R'
character and for the `num` concatenation) and in the worst case
creates three string objects (one more for the `gen` concatenation).
For the Tracemonkey paper alone, this resulted in 12000 string
objects when scrolling from the top to the bottom of the document.
Since this is a hot function, it's worth minimizing the number of string
objects, especially for large documents, to reduce peak memory usage.
This commit refactors the `toString` method to always create only one
string object.
For PDF documents with sufficiently broken XRef tables, it's usually quite obvious when you need to fallback to indexing the entire file. However, for certain kinds of corrupted PDF documents the XRef table will, for all intents and purposes, appear to be valid. It's not until you actually try to fetch various objects that things will start to break, which is the case in the referenced issues[1].
Since there's generally a real effort being in made PDF.js to load even corrupt PDF documents, this patch contains a suggested approach to attempt to do a bit more validation of the XRef table during the initial document loading phase.
Here the choice is made to attempt to load the *first* page, as a basic sanity check of the validity of the XRef table. Please note that attempting to load a more-or-less arbitrarily chosen object without any context of what it's supposed to be isn't a very useful, which is why this particular choice was made.
Obviously, just because the first page can be loaded successfully that doesn't guarantee that the *entire* XRef table is valid, however if even the first page fails to load you can be reasonably sure that the document is *not* valid[2].
Even though this patch won't cause any significant increase in the amount of parsing required during initial loading of the document[3], it will require loading of more data upfront which thus delays the initial `getDocument` call.
Whether or not this is a problem depends very much on what you actually measure, please consider the following examples:
```javascript
console.time('first');
getDocument(...).promise.then((pdfDocument) => {
console.timeEnd('first');
});
console.time('second');
getDocument(...).promise.then((pdfDocument) => {
pdfDocument.getPage(1).then((pdfPage) => { // Note: the API uses `pageNumber >= 1`, the Worker uses `pageIndex >= 0`.
console.timeEnd('second');
});
});
```
The first case is pretty much guaranteed to show a small regression, however the second case won't be affected at all since the Worker caches the result of `getPage` calls. Again, please remember that the second case is what matters for the standard PDF.js use-case which is why I'm hoping that this patch is deemed acceptable.
---
[1] In issue 7496, the problem is that the document is edited without the XRef table being correctly updated.
In issue 10326, the generator was sorting the XRef table according to the offsets rather than the objects.
[2] The idea of checking the first page in particular came from the "standard" use-case for the PDF.js library, i.e. the default viewer, where a failure to load the first page basically means that nothing will work; note how `{BaseViewer, PDFThumbnailViewer}.setDocument` depends completely on being able to fetch the *first* page.
[3] The only extra parsing is caused by, potentially, having to traverse *part* of the `Pages` tree to find the first page.
If, as PR 10368 suggests, more parameters should be added to `getViewport` I think that it would be a mistake to not change the signature *first* to avoid needlessly unwieldy call-sites.
To not break any existing code and third-party use-cases, this is obviously implemented with a deprecation warning *and* with a working fallback[1] for the old method signature.
---
[1] This is limited to `GENERIC` builds, which should be sufficient.
Given that it's really not clear to me if this is actually desired functionality in the default viewer, and considering that it doesn't fit in *great* with the way that `PDFHistory` is initialized, this feature is currently off by default[1].
---
[1] It's controlled with the `disableOpenActionDestination` Preference/AppOption.
Note that the OpenAction dictionary may contain other information besides just a destination array, e.g. instructions for auto-printing[1].
Given first of all that an arbitrary `Dict` cannot be sent from the Worker (since cloning would fail), and second of all that the data obviously needs to be validated, this patch purposely only adds support for fetching a destination from the OpenAction entry[2].
---
[1] This information is, currently in PDF.js, being included through the `getJavaScript` API method.
[2] This significantly reduces the complexity of the implementation, which seems fine for now. If there's ever need for other kinds of OpenAction to be fetched, additional API methods could/should be implemented as necessary (could e.g. follow the `getOpenActionWhatever` naming scheme).
The custom entries, provided that they exist *and* that their types are safe to include, are exposed through a new `Custom` infoDict entry to clearly separate them from the standard ones.
Fixes 5970.
Fixes 10344.
This required the following changes in the Gulpfile:
- Defining a series of tasks is no longer done with arrays, but with the
`gulp.series` function. The `web` target is refactored to use a
smaller number of tasks to prevent tasks from running multiple times.
- Getting all tasks must now be done through the task registry.
- Tasks that don't return anything must call `done` upon completion.
Moreover, this upgrade allows us to use the latest Node.js on Travis CI
again.
Given that Signature (Widget) annotations are currently not supported, since they cannot be validated, simply ignoring the `fieldValue` seems OK for now considering that attempting to blindly include unparsed/unvalidated data isn't very useful.
Fixes 10347.
This patch re-factors, and extends, the already existing `zoomDisabledTimeout` used during mouse wheel zooming.
Unfortunately I haven't got the required hardware to actually test this patch, but there's a decent chance that it will fix, or at least reduce, the problems reported in https://bugzilla.mozilla.org/show_bug.cgi?id=1503412.
Given that a larger number of pages may now be visible at once, and importantly that their layout may be non-vertical, one of the conditions should be tweaked to not accidentally miss cases where a page is still visible.
Please note: This patch is based on code-inspection, and the only ill effect occurring without it would be a couple of (near) duplicate history entries in some *rare* edge-cases.
This patch does three things:
- Updates the `gulp unittestcli` command, using `gulp lint` as a guide, such that it can be run locally on Windows without any modifications.
- Updates the `gulp lib` command to support disabling of Babel through the `SKIP_BABEL` environment variable. Note that all other build targets support this mode, and there's no good reason for `lib` to be any different here.
- Updates the `npm test` command, used in Node.js/Travis, to test the code as-is test. Since modern Node.js versions seem to have no problems with ES6 compatible code in general, we should just test the source code as-is instead (similar to the tests running on the regular bots).
With the removal of the global `PDFJS` object, in PDF.js version `2.0`, the viewer options are no longer as easily accessible as they previously were (and issues have been filed about this).
In particular, since the viewer files aren't necessarily loaded *immediately*, this means that `PDFViewerApplication`/`PDFViewerApplicationOptions` aren't necessarily available directly. By dispatching an event once all viewer files are loaded but *before* the viewer initialization has run, setting `AppOptions` during load (in custom implementations of the default viewer) should hopefully become a little bit easier[1].
---
[1] In hindsight, this should probably have been implemented when the global `PDFJS` object was removed...
Rather than having a (somewhat) randomly choosen list of Preferences which `AppOptions` are allowed to override, it makes much more sense to simply add an AppOption to allow custom implementations to ignore Preferences altogether (it's also inline with the AppOption that allows the `ViewHistory` to be bypassed on load).
This should have been part of the previous commit that updated the
Wintersmith dependency. Markdown support is no longer included in Pug
itself and should be done by a transformer instead.
The intent of the code, based on existing comments, is to perform a binary search. However, because of what appears to be a typo in the code responsible for computing the current search index, this code is always checking *every* entry (albeit only at the "final" node) starting from the last one.
Update configuration to work with Babel 7
Explicitly require globals - eslint-plugin-mozilla needs it, but doesn't require it on its own.
Fix Regexp to match Babel 7's inlined _interopRequireDefault
According to the specification, see https://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#G6.2384179, the keys of NameTree/NumberTree should be ordered.
For corrupt PDF files, which violate this assumption, we thus need to fallback to an exhaustive search in order to e.g. find all destinations.
*Please note:* Given that this only implements a fallback for the "final" node of the Tree, there's obviously a risk that the patch isn't sufficient for dealing with all kinds of out-of-order corruption. However, this kind of problem should be rare in practice, and without a real-world test-case it's difficult to implement a completely general solution (and there's obviously a question if you'd even want to).
This changes all occurrences of `var` to `let`/`const` in this code, and updates the signature of the constructor to use object destructuring for better readability (and self documentation).
Also, `useRequestAnimationFrame` is changed to a parameter and the `typeof window` check is now done *once* rather than at every `_scheduleNext` call.
This changes all occurrences of `var` to `let`/`const` in this code, and updates the signatures of a couple of methods to use object destructuring.
Finally, when creating `InternalRenderTask` instances *only* the necessary parameter are now provided, since passing through the `RenderParameters` as-is seems completely unnecessary.
Jasmine recommends to use the `configure` method on the environment
during boot. This commit makes the code correspond to how it's done in
Jasmine's default boot file. The options dropdown in the HTML reporter
now works again after these changes, because this broke in the upgrade
to Jasmine 3, and the unit tests are executed in a random order by
default, which is important to make sure the unit tests are
self-contained and don't depend on the result of another unit test.
Rather than closing [bug 1505824](https://bugzilla.mozilla.org/show_bug.cgi?id=1505824) as WONTFIX (which is my preferred solution), given how *minor* this "problem" is, it's still possible to adjust the error messages a bit.
The main point here, which is relevant even if the changes in `BaseViewer` are ultimately rejected during review, is that we'll no longer attempt to call `BaseViewer.currentPageLabel` with an empty string from `webViewerPageNumberChanged` in `app.js`.
The other changes are:
- Stop printing an error in `BaseViewer._setCurrentPageNumber`, and have it return a boolean indicating if the page is within bounds.
- Have the `BaseViewer.{currentPageNumber, currentPageLabel}` setters print their own errors for invalid pages.
- Have the `BaseViewer.currentPageLabel` setter no longer depend, indirectly, on the `BaseViewer.currentPageNumber` setter.
- Improve a couple of other error messages.
When `highlightAll` is *not* set, there's only going to be a single match per page and unconditionally calling `PDFFindController.scrollMatchIntoView` doesn't really matter.
However, when `highlightAll` is set the current code may result in a large number of unnecessary `PDFFindController.scrollMatchIntoView` calls. Since `TextLayerBuilder._renderMatches` already checks if a particular match is the selected one, for highlighting purposes, it's simple enough to also skip scrolling completely for non-selected matches.
Only scroll search results into view as a result of an actual find operation, and not when the user scrolls/zooms/rotates the document (bug 1237076, issue 6746)
Most of the code in `PDFFindController` assumes that a valid `state` always exits, hence it cannot hurt to add a simple check to avoid errors being thrown.
Currently searching, and particularily highlighting of search results, may interfere with subsequent user-interactions such as scrolling/zooming/rotating which can result in a somewhat jarring UX where the document suddenly "jumps" to a previous position.
This is especially annoying in cases where the highlighted search result isn't even visible when a user initiated scrolling/zooming/rotating happens, and there exists a couple of bugs/issues about this behaviour.
It seems reasonable, as far as I'm concerned, to treat searching as one operation and any subsequent non-search user interactions with the viewer as separate and thus not scroll the current search result into view *unless* the user is actually doing another search.
This also seems consistent with general searching in e.g. Firefox and Adobe Reader:
- Compare with "regular" searching of e.g. HTML files in Firefox, where the user scrolling and/or zooming the document will not force a currently highlighted search result to become re-scrolled into view.
- Compare also with Adobe Reader, where the user scrolling, zooming, and/or rotating the document will not force the currently highlighted search result to become re-scrolled into view.
The question is then why search highlighting was implemented this way in PDF.js to begin with. It might be that this wasn't really intended behaviour, but more a consequence of the asynchronous nature of the API. Considering that most operations, such as fetching the page, rendering it and extracting its text-content are all asynchronous; searching and highlighting of matches thus becomes asynchronous too.
However, it should be possible to track when search results have been scrolled into view and highlighted, and thus prevent these wierd "jumps" when the user interacts with the document.
*Please note:* Unfortunately this required moving the scrolling of matches back into `PDFFindController`, since I simply couldn't see any other (reasonable) way of implementing the functionality without tracking the `_shouldScroll` property in only *one* spot.
However, given that the new `PDFFindController.scrollMatchIntoView` method follows a similar pattern as `BaseViewer.scrollPageIntoView` and `PDFThumbnailViewer.scrollThumbnailIntoView`, this is hopefully deemed OK.
Given that the `_updateScrollMode`/`_updateSpreadMode` methods are "private", there's no particular reason to not just directly call `_setCurrentPageNumber`.
Note that when e.g. presentation mode is active, we fail[1] to ensure that the `pageNumber` parameter is actually an integer before calling `_setCurrentPageNumber` (that method expects the argument be an integer).
Also changes the method signature, of `scrollPageIntoView`, to use object destructuring instead.
---
[1] Most likely, this is actually *my* oversight :-)
These interfaces are already used in different files, in both the `src/core/` and `src/display/` folders, and having them reside in their own file seems a lot clearer and is also similar to the existing viewer interfaces.
As part of moving the `interface` definitions, they're also converted to ES6 classes.
First of all, note how there's currently *two* methods for checking if a certain object exists, which seems completely unwarranted.
Furthermore, the rarely used `getData` method was removed and its only callsite changed to use a combination of `PDFObjects.{has, get}` instead.
Finally, the methods were rearranged slightly, to bring the most important ones (for an API user) to the top of the class.
Currently we'll only attempt to start from the current page when a new search is done, however for 'findagain' operations we'll always continue from the last match position.
This could easily lead to confusing behaviour if the user has scrolled to a completely different part of the document. In an attempt to improve this somewhat, for repeated 'findagain' operations, we'll instead reset the position to the current page when it's *absolutely* certain that the user has scrolled.
Note that this required adding a new `BaseViewer` method, and exposing that through `PDFLinkService`, in order to check if a given page is visible.
In an attempt to avoid issues, in custom implementations of `PDFFindController`, the code checks for the existence of the `PDFLinkService.isPageVisible` method *before* using it.
Unfortunately the `PDFFindController.executeCommand` method has now become a bit more complicated than one would like, but hopefully this small change will improve the structure somewhat (especially for subsequent patches).
Remove the `attachDOMEventsToEventBus` functionality, since `EventBus` instances are able to re-dispatch events to the DOM (PR 10019, bug 1492849 follow-up)
With only *one* event now being dispatched when the scale changes, in combination with there only being two call-sites, it doesn't seem necessary to keep the helper method for dispatching the 'scalechanging' event.
Note how nowhere in the code `canvasInRendering.get()` is ever called, and that this structure is really only used to store references to `<canvas>` DOM elements.
The reason for this being a `WeakMap` is probably because at the time we weren't using `core-js` polyfills yet, and since there already existed a manually implemented `WeakMap` polyfill it was probably simpler to use that.
Please note that, given the lack of a runnable example, I'm not totally sure if this first of all is enough to *completely* address the issue as filed and second of all if we actually want this new behaviour.
Currently `PDFFindController._calculateMatch` is (indirectly) dispatching an `updatetextlayermatches` event for every *single* page of the document. For short documents, such as the `tracemonkey` file, this probably doesn't matter too much, but for documents with a couple of thousand pages it seems unfortunate.
It shouldn't be necessary, in general, to dispatch `updatetextlayermatches` events here, since that's already being taken care of in `PDFFindController._updateMatch` which is always called when a match has been found.
However, when `highlightAll` is set we still need to ensure that pages which finished rendered *before* searching begun are updated correctly.
[Regression] Ensure that "Highlight All" is propagated to all pages for 'findagain' events where the findbar was previously closed (PR 10100 follow-up)
**STR:**
1. Open the default viewer, with the `tracemonkey` file.
2. Open the findbar, and search for "trace".
3. Enable the "Highlight All" option.
4. Close the findbar.
5. Re-open the findbar, and click on the "findNext" button.
6. Scroll down to the *second* page of the document.
**ER:**
Since "Highlight All" is active, all matches on the *second* page should be highlighted.
**AR:**
No matches are highlighted on the *second* page.
This is relevant for e.g. `PDFSinglePageViewer`, and `PDFViewer` with Presentation Mode active.
By moving this code to a helper method in `BaseViewer`, it's thus possible to reduce the amount of duplicate code that currently needed in `PDFViewer` and `PDFSinglePageViewer`.
For a short document, such as e.g. the `tracemonkey` file, this repeated normalization won't matter much, but for documents with a couple of thousand pages it seems completely unnecessary (and wasteful) to keep repeating the normalization whenever for every single page.
Currently the text-content is normalized every time that a new search operation is started, which seems completely useless considering that the "raw" text-content is never used for anything.
For a short document, such as e.g. the `tracemonkey` file, this repeated normalization won't matter much, but for documents with a couple of thousand pages it seems completely unnecessary (and wasteful) to keep repeating the normalization whenever e.g. a new search operation starts.
In the event that multiple instances of `PDFFindController` ever exists simultaneously, they will all be able to share just one `normalize` function in this way. Furthermore, the regular expression is now created lazily rather than at class construction time.
Despite all highlighted matches being removed in response to the 'findbarclose' event, there's a risk that a match could still be scrolled into view *after* the findbar has been closed[1].
Hence we need to ensure that long running searches, particularily those happening in large and/or slow loading documents[2], are ignored as well.
---
[1] The match is hidden, as expected, but the document could still scroll unexpectedly.
[2] Large documents loaded with `disableAutoFetch = true` and `disableStream = true` set are particularily susceptible to this issue.
Given that dispatching the 'updatetextlayermatches' event with `pageIndex = -1` set is now used to target the textLayers of *all* pages, there's no need to send individual events to every single page during `_nextMatch`. Since there can be an arbitrary number of pages in a document, this small/simple optimization seems too easy to ignore.
This patch does four things:
- Change the search related methods in `TextLayerBuilder` to be "private", since there're only called from within the class itself now.
- Use `const` for local variables not intended to change in the search related methods in `TextLayerBuilder`.
- Finally, removes most `this.findController` checks since they are redundant. Note how both `this._convertMatches` and `this._renderMatches` are *only* ever called, from `this._updateMatches`, when `this.findController` is actually defined. Hence there's really no need to repeat those checks all over the place, especially with all the relevant methods now being marked as "private".
- Always initialize the `this._pageMatchesLength` property with an empty array, to simplify the code in `TextLayerBuilder`.
This line in the annotation tests subtracts an array from a number. This is because `splice(-1, 1)` returns a one-element array, while `pop()` returns only the element itself.
`CMapCompressionType` makes a lot of sense to export, for anyone attempting to implement a custom `CMapReaderFactory`; fixes 10148.
`PermissionFlag` likewise needs to be exported, since otherwise the result of the `getPermissions` API method becomes difficult to interpret; follow-up to 10033.
*Please note:* I'm totally fine with this patch being rejected, and the issue closed as WONTFIX; however these changes should address the issue if that's desired.
From a conceptual point of view, reporting loading progress doesn't really make a lot of sense for PDF files opened by passing raw binary data directly to `getDocument` (since obviously *all* data was loaded).
This is compared to PDF files loaded via e.g. `XMLHttpRequest` or the Fetch API, where the entire PDF file isn't available from the start and knowing the loading progress makes total sense.
However I can certainly see why the current API could be considered inconsistent, which isn't great, since a registered `onProgress` callback will never be called for certain `getDocument` calls.
The simplest solution to this inconsistency thus seem to be to ensure that `onProgress` is always called when handling the `DataLoaded` message, since that will *always* be dispatched[1] from the worker-thread.
---
[1] Note that this isn't guaranteed to happen, since setting `disableAutoFetch = true` often prevents the *entire* file from ever loading. However, this isn't relevant for the issue at hand, and is a well-known consequence of using `disableAutoFetch = true`; note how the default viewer even has a specialized code-path for hiding the loadingBar.
*This should have been part of PR 10139.*
In the event that a user has attempted to manually load the worker file on the main-thread, but somehow failed to do that correctly, there's a possibility that `getMainThreadWorkerMessageHandler` could throw. Considering how/where that helper function is being called, an error could still prevent `PDFDocumentLoadingTask` from completing (regardless if it's being resolved/rejected).
With the way that the `getWorkerSrc()` helper function is implemented now, there's no longer a particularly strong reason for keeping the global `pdfjsFilePath` variable around.
With this patch the fallback `workerSrc` will thus, assuming is wasn't already set, be set to the "pdfjsFilePath" which simplifies the `getWorkerSrc()` function and reduces the amount of global state.
Finally, the global `workerSrc` variable was renamed to prevent shadowing.
This should, hopefully, cover all the possible ways[1] in which "fake workers" are loaded. Given the different code-paths, adding unit-tests might not be that simple.
Note that in order to make this work, the various `fakeWorkerFilesLoader` functions were converted to return `Promises`.
---
[1] Unfortunately there's lots of them, for various build targets and configurations.
Rather than having every invocation of `Toolbar._updateUIState` compute a valid `pageScaleValue`, it seems easier to simply ensure that it happens when the value is actually updated.
Looking at the history of this code, this parameter has never been used.
I'm guessing that most likely the code in `web/toolbar.js` began life as a copy of `web/secondary_toolbar.js`, which would probably explain why that parameter exists.
*This patch is based on something that I noticed while working on PR 10126.*
The recent re-factoring of `PDFFindController` brought many improvements, among those the fact that access to `BaseViewer` is no longer required. However, with these changes there's one thing which now strikes me as not particularly user-friendly[1]: The fact that in order for searching to actually work, `PDFFindController.setDocument` must be called *and* a 'pagesinit' event must be dispatched (from somewhere).
For all other viewer components, calling the `setDocument` method[2] is enough in order for the component to actually be usable.
The `PDFFindController` thus stands out quite a bit, and it also becomes difficult to work with in any sort of custom implementation. For example: Imagine someone trying to use `PDFFindController` separately from the viewer[3], which *should* now be relatively simple given the re-factoring, and thus having to (somehow) figure out that they'll also need to manually dispatch a 'pagesinit' event for searching to work.
Note that the above even affects the unit-tests, where an out-of-place 'pagesinit' event is being used.
To attempt to address these problems, I'm thus suggesting that *only* `setDocument` should be used to indicate that searching may start. For the default viewer and/or the viewer components, `BaseViewer.setDocument` will now call `PDFFindController.setDocument` when the document is ready, thus requiring no outside configuration anymore[4]. For custom implementation, and the unit-tests, it's now as simple as just calling `PDFFindController.setDocument` to allow searching to start.
---
[1] I should have caught this during review of PR 10099, but unfortunately it's sometimes not until you actually work with the code in question that things like these become clear.
[2] Assuming, obviously, that the viewer component in question actually implements such a method :-)
[3] There's even a very recent issue, filed by someone trying to do just that.
[4] Short of providing a `PDFFindController` instance when creating a `BaseViewer` instance, of course.
[Regression] Restore the ability to start searching before a document has loaded, and ignore searches for previously opened documents (PR 10099 follow-up)
Since searching itself is an asynchronous operation, removal of highlights needs to be asynchronous too since otherwise there's a risk that the events happen in the wrong order and find highlights thus remain visible.
Also, this patch will now ensure that only 'findbarclose' events for the *current* document is handled since other ones doesn't really matter. Note in particular that when no document is loaded text-layers are, obviously, not present and subsequently it's unnecessary to attempt to hide non-existent find highlights.
For many years it's been possible to enter a search term into the findbar(s) before the document has finised loading, such that searching starts immediately once it has loaded.
PR 10099 accidentally broke that, which I unfortunately missed during reviewing.
Since searching is asynchronous you cannot directly check in `executeCommand` if the document is loaded/current, but need to wait until searching is actually enabled first.
Furthermore this patch also ensures that the `_findTimeout` is always correctly cleared given that it adds further asynchronous behaviour to searching, since you obviously only want to deal with searches relevant to the current document.
This is similar to the format used by a number of other viewer components, and should simplify the `PDFSidebar` initialization slightly.
Furthermore, by using the `eventBus` it's no longer necessary for `PDFSidebar` to have a direct dependency on `PDFOutlineViewer`.
There's still room for improvement here, but this patch is at least a start (since it's not clear to me how best to handle the viewers).
This attempts to reduced the level of indirection, and the amount of code, when dispatching `fileattachmentannotation` events, by removing the `PDFLinkService.onFileAttachmentAnnotation` method and just accessing `PDFLinkService.eventBus` directly in the `FileAttachmentAnnotationElement` constructor.
Given that other properties, such as `externalLinkTarget`/`externalLinkRel`, are already being accessed directly this pattern seems fine here as well.
This commit shows that we can now unit test the find controller and
that executing regular queries works. Note that this is only a first
step and not a complete suite of unit tests for all possible options
of the find controller.
While writing this unit test, I found two smaller issues that I
addressed directly. The first one is that in the previous find
controller refactoring I forgot to rename some occurrences of a now
private member variable. Fortunately this did not cause any bugs since
we did have a public getter and the fetched value may be changed by
reference, but it's nevertheless good to fix. The second issue is that
some entries in the `test/unit/clitests.json` file were not correct,
resulting in these tests not being executed on e.g., Travis CI.
With `PDFFindController` instances no longer (directly) depending on
`BaseViewer` instances, we can pass a single `findController` when
initializing a viewer, similar to other components.
This removes the dependency on a `PDFViewer` instance from the find
controller, which makes it more similar to other components and makes it
easier to unit test with a mock link service.
Finally, we remove the search capabilities from the SVG example since it
doesn't work there because there is no separate text layer.
Now it follows the same pattern as e.g., the document properties
component, which allows us to have one instance of the find controller
and set a new document to search upon switching documents.
Moreover, this allows us to get rid of the dependency on `pdfViewer` in
order to fetch the text content for a page. This is working towards
getting rid of the `pdfViewer` dependency upon initializing the
component entirely in future commits.
Finally, we make the `reset` method private since it's not supposed to
be used from the outside anymore now that `setDocument` takes care of
this, similar to other components.
This way the resetting of `PDFLinkService`/`PDFDocumentProperties` instances, as is done in `PDFViewerApplication.close`, only requires passing in *one* `null` argument instead of two.
The `started` timestamp is completely usused, and the `end` timestamp is currently[1] being used essentially like a boolean value.
Hence this code can be simplified to use an actual boolean value instead, which avoids potentially hundreds (or even thousands) of unnecessary `Date.now()` calls.
---
[1] Looking briefly at the history of this code, I cannot tell if the timestamps themselves were ever used for anything (except for tracking "boolean" state).
The `Font.loading` property is only ever used *once* in the code, whereas `Font.missingFile` is more widely used. Furthermore the name `loading` feels, at least to me, slight less clear than `missingFile`. Finally, note that these two properties are the inverse of each other.
Ensure that all event properties are included, even if no (internal) listeners are registered, when re-dispatching events to the DOM (PR 10019 follow-up)
The built-in PDF Viewer (in Firefox) cannot use the browser findbar when PDF files are embedded in e.g. iframe/object tags, and the PDF.js findbar (i.e. `PDFFindBar`) will thus be used instead in those cases.
This is slightly problematic, since the `MOZCENTRAL` version of the viewer uses a special, slimmed down, version of the `l10n.js` file that doesn't (currently) support plural forms. To prevent the matchesCounter from breaking completely in this edge-case, temporarily hard-code the plural form to use the default `[other]` version of the locale strings.
This prevents the findbar from intermittently displaying `0 of {number} matches`, which *could* theoretically happen for large and/or slow loading documents.
The find controller should only coordinate finding a string in the
document and should not be responsible for presenting the matches to the
user. The text layer builder already contains the logic to render the
matches in the viewer, so it should also take care of scrolling the
selected match into view.
Currently there's only a single spot in the code-base where `JpegImage.getData` is called, however it nonetheless seem like a good idea to ensure during tests that the `isSourcePDF` parameter is correctly set. (Especially considering that the PDF use-cases will break without it.)
Additionally, in `JpegImage._getLinearizedBlockData`, the code can be made a tiny bit more efficient by checking the value of `isSourcePDF` *first* to avoid useless checks (for the default PDF use-cases).
The find controller already has quite a lot of state to maintain. We can
avoid keeping track of this member variable because when the find
controller is reset, so is the extract text promises array. Therefore,
we can just check if that array contains items or not to determine if
text extraction already started.
Moreover, there is no need to reset the `pageContents` array since the
`reset` method already takes care of that.
Implement unit tests for the `isSameOrigin` and `createValidAbsoluteUrl` utility functions and use the `const` keyword for constants in `src/shared/util.js`
This is an unfortunate oversight on my part, which I stumbled upon when (locally) testing the `mozilla-central` follow-up patch necessary to enable the matches counter in the built-in PDF viewer.
The property is intended to contain a reference to a DOM element, which not only is nowhere to be found *now* but appears to never have existed in the first place.
As outlined in https://bugzilla.mozilla.org/show_bug.cgi?id=1282759 the internal Firefox name for the feature is `entireWord`, hence that name is used here as well for consistency (with "Whole words" being limited to the UI).
Given existing limitations of the PDF.js search functionality, e.g. the existing problems of searching across "new lines", there's some edge-cases where "Whole words" searching will ignore (valid) results.
However, considering that this is a pre-existing issue related to the way that the find controller joins text-content together, that shouldn't have to block this new feature in my opionion.
*Please note:* In order to enable this feature in the `MOZCENTRAL` version, a small follow-up patch for [PdfjsChromeUtils.jsm](https://hg.mozilla.org/mozilla-central/file/tip/browser/extensions/pdfjs/content/PdfjsChromeUtils.jsm) will be required once this has landed in `mozilla-central`.
For the `PDFFindBar` implementation, similar to the native Firefox findbar, the matches count displayed is now limited to a (hopefully) reasonable value.
*Please note:* In order to enable this feature in the `MOZCENTRAL` version, a follow-up patch will be required once this has landed in `mozilla-central`.
There have been lots of problems with trying to map glyphs to their unicode
values. It's more reliable to just use the private use areas so the browser's
font renderer doesn't mess with the glyphs.
Using the private use area for all glyphs did highlight other issues that this
patch also had to fix:
* small private use area - Previously, only the BMP private use area was used
which can't map many glyphs. Now, the (much bigger) PUP 16 area can also be
used.
* glyph zero not shown - Browsers will not use the glyph from a font if it is
glyph id = 0. This issue was less prevalent when we mapped to unicode values
since the fallback font would be used. However, when using the private use
area, the glyph would not be drawn at all. This is illustrated in one of the
current test cases (issue #8234) where there's an "ä" glyph at position
zero. The PDF looked like it rendered correctly, but it was actually not
using the glyph from the font. To properly show the first glyph it is always
duplicated and appended to the glyphs and the maps are adjusted.
* supplementary characters - The private use area PUP 16 is 4 bytes, so
String.fromCodePoint must be used where we previously used
String.fromCharCode. This is actually an issue that should have been fixed
regardless of this patch.
* charset - Freetype fails to load fonts when the charset size doesn't match
number of glyphs in the font. We now write out a fake charset with the
correct length. This also brought up the issue that glyphs with seac/endchar
should only ever write a standard charset, but we now write a custom one.
To get around this the seac analysis is permanently enabled so those glyphs
are instead always drawn as two glyphs.
It's no longer necessary since https://bugzilla.mozilla.org/show_bug.cgi?id=1305963 is fixed quite some time ago.
While we're here, mark the `cachedGetSinglePixelWidth` member as being
private and use ES6 syntax in the `getSinglePixelWidth` method.
The purpose of this patch is to provide a better default behaviour when `JpegImage` is used to parse standalone JPEG images with CMYK colour spaces.
Since the issue that the patch concerns is somewhat of a special-case, the implementation utilizes the already existing decode support in an attempt to minimize the impact w.r.t. code size.
*Please note:* It's always possible for the user of `JpegImage` to control image inversion, and thus override the new behaviour, by simply passing a custom `decodeTransform` array upon initialization.
This patch is the first step to be able to eventually get rid of the `attachDOMEventsToEventBus` function, by allowing `EventBus` instances to simply re-dispatch most[1] events to the DOM.
Note that the re-dispatching is purposely implemented to occur *after* all registered `EventBus` listeners have been serviced, to prevent the ordering issues that necessitated the duplicated page/scale-change events.
The DOM events are currently necessary for the `mozilla-central` tests, see https://hg.mozilla.org/mozilla-central/file/tip/browser/extensions/pdfjs/test, and perhaps also for custom deployments of the PDF.js default viewer.
Once this have landed, and been successfully uplifted to `mozilla-central`, I intent to submit a patch to update the test-code to utilize the new preference. This will thus, eventually, make it possible to remove the `attachDOMEventsToEventBus` functionality.
*Please note:* I've successfully ran all `mozilla-central` tests locally, with these patches applied.
---
[1] The exception being events that originated on the `window` or `document`, since those are already globally available anyway.
The new events follow the same naming pattern as the 'pagesinit'/'pagesloaded' events dispatched on `BaseViewer` instances, and the intention is to allow the eventual removal of 'documentload'.
Ensure that the built `PdfJsDefaultPreferences.jsm` file won't be affected/touched during tree-wide ESLint rule changes in `mozilla-central` (PR 9571 follow-up)
Apparently there's some PDF generators, in this case the culprit is "Nooog Pdf Library / Nooog PStoPDF v1.5", that manage to mess up PDF creation enough that endstream[1] commands actually become truncated.
*Please note:* The solution implemented here isn't perfect, since it won't be able to cope with PDF files that contains a *mixture* of correct and truncated endstream commands.
However, considering that this particular mode of corruption *fortunately* doesn't seem very common[2], a slightly less complex solution ought to suffice for now.
Fixes 10004.
---
[1] Scanning through the PDF data to find endstream commands becomes necessary, in order to determine the stream length in cases where the `Length` entry of the (stream) dictionary is missing/incorrect.
[2] I cannot recall having seen any (previous) issues/bugs with "Missing endstream" errors.
Reduces the amount of boilerplate code when defining the the sub-classes.
Please note that a couple of the closures were kept, since it's not (yet) possible to include helper functions inside of `class`es.
This property is not only completely unused now, it never actually appears to have been used. Even though the memory savings, from not initializing these extra typed arrays, won't be significant in the grand scheme of things it still seems completely unnecessary to keep allocating this data.
As far as I can tell, the main reason for the existence of `defaultColor` seem to be for documentation purposes. Hence the code is changed into comments instead, to keep the information around (but without the unnecessary allocations).
For proof-of-concept, this patch converts a couple of `Promise` returning methods to use `async` instead.
Please note that the `generic` build, based on this patch, has been successfully testing in IE11 (i.e. the viewer loads and nothing is obviously broken).
Being able to use modern JavaScript features like `async`/`await` is a huge plus, but there's one (obvious) side-effect: The size of the built files will increase slightly (unless `SKIP_BABEL == true`). That's unavoidable, but seems like a small price to pay in the grand scheme of things.
Finally, note that the `chromium` build target was changed to no longer skip Babel translation, since the Chrome extension still supports version `49` of the browser (where native `async` support isn't available).
Not only is this method completely unused *now*, looking through the history of the code it never appears to have been used for anything either.
Years ago `mainXRefEntriesOffset` was included when creating `XRef` instances, however it wasn't actually used for anything (the parameter was never checked, nor assigned to a property on `XRef`).
If this method ever becomes useful (again) it's easy enough to restore it thanks to version control, but including dead code in the builds just seems wasteful.
Please note that while this *improves* issue 9984 slightly (and likely others too), it's not a complete solution.
The remaining issues are related to the, more general, problems with the existing heuristics related to attempting to combine separate text items.
One of the `QueueOptimizer` cases wasn't updated to use `Uint8ClampedArray`s, which leads to inconsistent image data on the API side (but no actual rendering bugs, as far as I can tell).
To prevent future errors, a non-production/test-only `assert` was added to ensure that the relevant image data only uses `Uint8ClampedArray`s.
This commit is the first step towards implementing parsing for the
appearance streams of annotations.
Co-authored-by: Jonas Jenwald <jonas.jenwald@gmail.com>
Co-authored-by: Tim van der Meij <timvandermeij@gmail.com>
The font tests use Jasmine too, so while they are technically unit
tests, it's a bit confusing to see `Started unit tests` when the font
tests are run on the bots.
The current font type/subtype detection code is quite inconsistent/unwieldy. In some cases it will simply assume that the font dictionary is correct, in others it will somewhat "arbitrarily" check the actual font file (more of these cases have been added over the years to fix specific bugs).
As is evident from e.g. issue 9949, the font type/subtype detection code is continuing to cause issues. In an attempt to get rid of these hacks once and for all, this patch instead re-factors the type/subtype detection to *always* parse the font file.
Please note that, as far as I can tell, we still appear to need to rely on the composite font detection based on the font dictionary. However, even if the composite/non-composite detection would get it wrong, that shouldn't really matter too much given that there's basically only two different code-paths (for "TrueType-like" vs "Type1-like" fonts).
The font in the PDF is marked as a CIDFontType0, but the font file is
actually a true type font. To fully address this issue we should really
peek into the font file and try to determine what it is. However, this
is the first case of this issue, so I think this solution is acceptable for
now.
Fixes a stupid oversight on my part, since /Filter may (obviously) contain an Array, which resulted in unnecessary console warning spam in perfectly valid PDF files.
Note that it still makes sense to check that /Filter is actually a Name, before attempting to access its `name` property, but the warning should definitely be removed.
This should really have been included in PR 9868, since it will help ensure that the `URL` constructor is correctly imported/exported by `src/shared/util.js`.
According to the PDF specification, see https://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#page=45
> When using the JPXDecode filter with image XObjects, the following changes to and constraints on some entries in the image dictionary shall apply (see 8.9.5, "Image Dictionaries" for details on these entries):
>
> - Width and Height shall match the corresponding width and height values in the JPEG2000 data.
>
> - . . .
Hence it seems reasonable to use the Width/Height of the image data *itself*, rather than the image dictionary when there's a mismatch. Given that JPEG 2000 images are already being parsed, in order to obtain basic parameters, the actual Width/Height is readily available in the `PDFImage` constructor.
Given that the code is currently assuming that the /Filter entry is a `Name`, it cannot hurt to actually ensure that's the case.
Also fixes an error message, for JPEG 2000 images with unsupported ColorSpaces, since `this.numComps` hasn't been initialized when it's accessed during the `throw new Error()` invocation.
[api-minor] Add an `IsLinearized` property to the `PDFDocument.documentInfo` getter, to allow accessing the linearization status through the API (via `PDFDocumentProxy.getMetadata`)
Since PDF.js already supports range requests and streaming, not to mention chunked rendering, attempting to use the `Linearization` dictionary in `PDFDocument.getPage` probably isn't going to improve performance in any noticeable way.
Nonetheless, when `Linearization` data is available, it will allow looking up the first Page *directly* without having to descend into the `Pages` tree to find the correct object.
With the `builtInCMapCache` being a simple Object, it unfortunately means that the `Catalog.cleanup` method isn't resetting it as intended.
By just replacing the `builtInCMapCache` with an empty Object, existing references to it will not actually be updated. The result is that e.g. `Page` instances still keeps references to, what should have been removed, CMap data.
To fix these problems, the `builtInCMapCache` is converted into a `Map` instead (since it can be easily reset).
When updating Preferences using the `set` method, the input is carefully validated. However, no validation is (currently) done when a `BasePreferences` instance is created, which probably isn't that great. Hence this patch that simply ignores, to not unnecessarily break loading of the viewer itself, any invalid Preferences.
Given that the various Preferences are currently, and have been for quite some time, only used when initializing `PDFViewerApplication` re-loading them when a new PDF file is opened in the viewer is essentially a no-op.
Furthermore, with the only usage of `BasePreferences.reload` now gone, the value of that method seems questionable at best. In the event that the functionality is actually needed again, similar to the `ViewHistory`, it'd probably make more sense to simply replace `PDFViewerApplication.preferences` with a new `BasePreferences` instance instead (using e.g. `DefaultExternalServices.createPreferences`).
The `git` protocol is unencrypted which means other parties could potentially eavesdrop your traffic. `https` or `ssh` is often encouraged because of this. (For example, the Ruby package manager `bundler` prints a warning when `git` sources are being used.)
There was a (somewhat) recent question on IRC about accessing the linearization status of a PDF document, and this patch contains a simple way to expose that through already existing API methods.
Please note that during setup/parsing in `PDFDocument` the linearization data is already being fetched and parsed, provided of course that it exists. Hence this patch will *not* cause any additional data to be loaded.
With this file now being a proper (ES6) module, it's no longer (technically) necessary for this structure to be lazily initialized. Considering its size, and simplicity, I therefore cannot see the harm in letting `DocumentInfoValidators` just be simple Object instead.
While I'm not aware of any bugs caused by the current code, it cannot hurt to add an `isDict` check in `PDFDocument.documentInfo` (since the current code assumes that `infoDict` being defined implies it also being a Dictionary).
Finally, the patch also converts a couple of `var` to `let`/`const`.
Note first of all that `PDFDocument` will be initialized with either a `Stream` or a `ChunkedStream`, and that both of these have `length` getters. Secondly, the `PDFDocument` constructor will assert that the `stream` has a non-zero (and positive) length. Hence there's no point in checking `stream.length` in the `linearization` getter.
For most other `DecodeStream` based streams, we'll attempt to estimate the minimum `buffer` length based on the raw stream data. The purpose of this is to avoid having to unnecessarily re-size the `buffer`, thus reducing the number of *intermediate* allocations necessary when decoding the stream data.
However, currently no such optimization is attempted for `StreamsSequenceStream`, and given that they can often be quite large that seems unfortunate. To improve this, at least somewhat, this patch utilizes the raw sizes of the `StreamsSequenceStream` sub-streams to estimate the minimum required `buffer` length.
Most likely this patch won't have a huge effect on memory consumption, however for pathological cases it should help reduce peak memory usage slightly.
One example is the PDF file in issue 2813, where currently the `StreamsSequenceStream` instances would grow their `buffer`s as `2 MiB -> 4 MiB -> 8 MiB -> 16 MiB -> 32 MiB`. With this patch, the same stream `buffers`s grow as `8 MiB -> 16 MiB -> 32 MiB`, thus avoiding a total of `12 MiB` of *intermediate* allocations (since there's two `StreamsSequenceStream` used, for rendering/text-extraction).
Add a `BasePreferences.getAll` method and use it to fetch all Preferences at once in `PDFViewerApplication._readPreferences`; ensure that the `externalLinkTarget` option is correctly set when the viewer is embedded
Given that *all* Preferences are already fetched in `PDFViewerApplication._readPreferences`, the amount of boilerplate/duplication can be considerably reduced with the addition of a `BasePreferences.getAll` method.
The only reason that this check ever existed in the first place, is that originally there was a global `PDFJS.openExternalLinkInNewWindow` option which was then subsumed by the (more generic) `PDFJS.externalLinkTarget` option. (The `externalLinkTarget` has since been moved into a `PDFLinkService` option, as part of PDF.js version `2.0`.)
Hence, during the period where both `PDFJS.openExternalLinkInNewWindow` and `PDFJS.externalLinkTarget` existed side-by-side, there was a need to allow the former one to override the latter one (for backward compatibility purposes). However, that's no longer the case, and this extra `externalLinkTarget` check can now be removed.
*This was a stupid error on my part; sorry about breaking this!*
With the current code, the value of the `externalLinkTarget` option is now (potentially) updated *after* the viewer components have been initialized. For the "viewer in iframe/object tag" case, the result is that the value of the `externalLinkTarget` option isn't adjusted as intended any more.
Without providing useful (custom) error messages for the `no-restricted-globals` rule, see https://eslint.org/docs/rules/no-restricted-globals, it's quite likely that the rule will be incorrectly disabled rather than the required globals being imported as intended.
To reduced duplication of the `no-restricted-globals` rule in multiple `.eslintrc` files, it's instead moved to the top-level `.eslintrc` file and disabled as needed on a folder/file basis outside of `/src` and `/web`.
With the new XML parser, see PR 9573, the referenced PDF file now causes `getMetadata` to fail when incomplete XML tags are encountered. This provides a simple, and hopefully generally useful, work-around that may also help prevent future bugs.
(Without being able to reproduce nor even understand the other (non XML) errors mentioned in issue 8884, I'd say that this patch is enough to close that one as fixed.)
*Another small piece of clean-up of code I've previously written; follow-up to PR 8775.*
Importing `createPromiseCapability`, and then using it in just *one* spot, seems unnecessary since the `waitOnEventOrTimeout` function may just as well return a regular `Promise` directly.
Compared to all the other (static) methods in `Util`, the `toRoman` one looks slightly out of place. Even more so considering that `Util` is being exposed through `pdfjsLib`, where access to a Roman numerals conversion method doesn't make much sense.
*I was feeling bored; so this is a very quick, and somewhat naive, attempt at fixing the bug.*
The breaking error, i.e. `Error during font loading: invalid array length`, was thrown when attempting to re-size the `stack` to a *negative* length when parsing the CALL functions.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1473809.
PR 6141 changed `CompiledFont.compileGlyph` to, in the general case, return an Array. However, that PR apparenly forgot to update the no-glyph, empty-glyph, and endchar-glyph code-path and a String was still being (incorrectly) returned.
Given the way that `FontFaceObject.getPathGenerator` (on the API side) is implemented, this shouldn't have caused any bugs despite the Worker possible returning unexpected data.
It gives warnings when running `gulp locale`, in a previous patch it was
explicitly asked to remove it and there is quite a bit of discussion
about the validity of the locale at
https://bugzilla.mozilla.org/show_bug.cgi?id=418485. In short, let's
remove it to prevent these issues.
Given that the non-default Spread modes (currently) doesn't affect the page layout when horizontal scrolling is enabled, having the Spread buttons appear active when clicking them appears to do *nothing* is probably confusing rather than helpful to users.
If the current viewer is a `PDFSinglePageViewer` instance the Scroll/Spread modes are no-ops, hence displaying buttons that do *nothing* when clicked will probably do very little besides confuse users.
The names 'resetscrollmode'/'resetspreadmode' were probably *not* great choices, given that the only thing being reset are toolbar buttons and not the actual Scroll/Spread modes. Furthermore, there's really no need for two separate events here.
The patch also adds a comment that ought to have been included in PR 9040, to prevent future refactoring/removing of what may appear to be an unnecessary `Promise.resolve` call.
This moves/exposes the `URL` polyfill similarily to the existing `ReadableStream` polyfill, rather than exposing it globally, to avoid interfering with any "outside" code.
Both the `URL` and `ReadableStream` polyfills are now exposed on the `pdfjsLib` object, such that they are accessible to the viewer components.
Furthermore, the `no-restricted-globals` ESLint rule is also enabled to prevent accidental usage of the native `URL`/`ReadableStream` implementations directly in the `src/` and `web/` folders; see also https://eslint.org/docs/rules/no-restricted-globals
Addresses the remaining TODO in https://github.com/mozilla/pdf.js/projects/6
Rather than having to manually call a method on `PDFFindController` instances from `BaseViewer.setDocument`, thus essentially having to resolve the private `_firstPagePromise` from the "outside", this can be done easily with the 'pagesinit' event dispatched on the `eventBus` instead.
Please note this particular `PDFFindController` code pre-dates the `eventBus` by almost three years, which should explain why the code looks the way it does.
Since the Scroll/Spread modes are now document specific, as all other properties such as page/scale/rotation, ensure that the toolbar is always correctly reset.
Since other viewer state, such as the current page/scale/rotation[1], are not available as `BaseViewer` constructor options, this makes the Scroll/Spread modes stand out quite a bit. Hence it probably makes sense to remove/deprecate this, to avoid inconsistent and possibly confusing state in this code.
---
[1] These properties are *purposely* not available in the constructor, since attempting to set them before a document is loaded has number of issues; please refer to https://github.com/mozilla/pdf.js/pull/8539#issuecomment-309706629 for additional details.
Since all the other viewer methods use the getter/setter pattern, e.g. for setting page/scale/rotation, the way that the Scroll/Spread modes are set thus stands out. For consistency, this really ought to use the same pattern as the rest of the `BaseViewer`. (To avoid breaking third-party implementations, the old methods are kept around as aliases.)
Note how the other "...OnLoad" preferences will allow a *non-default* value to always override a history entry. To improve overall consistency for the viewer options, and to reduce possibly confusing behaviour, this patch changes the `scrollModeOnLoad`/`spreadModeOnLoad` preferences to behave as all the other ones.
For some reason, these weren't added to `AppOptions` despite actually being set and read from `web/app.js`. Not adding them creates inconsistencies, since all other options *are* present in `web/app_options.js`.
This property isn't accessed anywhere in the `web/app.js` file, and is also not being reset in `PDFViewerApplication.close`. Hence it seems that it can simply be removed, especially since the fingerprint is already synchronously available through `PDFViewerApplication.pdfDocument.fingerprint` (provided that a document is loaded).
Currently if `RenderTask.cancel` is called *immediately* after rendering was started, then by the time that `InternalRenderTask.initializeGraphics` is called rendering will already have been cancelled.
However, we're still inserting the canvas into the `canvasInRendering` map, thus breaking any future attempts at re-rendering using the same canvas. Considering that `InternalRenderTask.cancel` always removes the canvas from the map, I cannot imagine that we'd ever want to re-add it *after* rendering was cancelled (it was likely just a simple oversight in PR 8519).
Fixes 9456.
With the current code, the location in the viewer could change *well* after the user has started to interact with the viewer (for very large and/or slow loading documents). Attempt to reduce the likelyhood of that happening, by adding an upper bound to the time spent waiting before attempting to re-apply the initial position.
Rather than using a "special" property to check if a `ViewHistory` database entry existed on load, it seems that you could just as well check for the existence of one of the actually needed properties instead (here 'page' is used).
This way we can avoid storing what, more of less, amounts to useless state, which will help reduce the size of the `ViewHistory` database. Given that we don't directly, nor need to, validate the `ViewHistory` data but rely on other parts of the code-base to do so, the existance of an 'exists' property doesn't in and of itself really add much utility.
Finally, to simplify the implementation the 'exists' property won't be actively removed from the `ViewHistory` data. Instead we'll simply stop adding 'exists' when writing `ViewHistory` data, and rely on the existing pruning of old entries to eventually remove any remnants of 'exists' from storage.
Note how, in the current code, only *one* old history entry would ever be removed. That would mean that if e.g. the `cacheSize` is reduced, it would potentially require loading of multiple files before the database would be correctly pruned.
Furthermore, in the case where the database was empty on load there's no need to attempt to shrink it, since trying to reduce the size of an *empty* array won't do much :-)
This wasn't included in PR 9245, since all the API options were still global at that time.
Writing the unit-tests also uncovered an issue with `getOperatorList` not starting the "Page Request" timer.
Incomplete path operators, in particular, can result in fairly chaotic rendering artifacts, as can be observed on page four of the referenced PDF file.
The initial (naive) solution that was attempted, was to simply throw a `FormatError` as soon as any invalid (i.e. too short) operator was found and rely on the existing `ignoreErrors` code-paths. However, doing so would have caused regressions in some files; see the existing `issue2391-1` test-case, which was promoted to an `eq` test to help prevent future bugs.
Hence this patch, which adds special handling for invalid path operators since those may cause quite bad rendering artifacts.
You could, in all fairness, argue that the patch is a handwavy solution and I wouldn't object. However, given that this only concerns *corrupt* PDF files, the way that PDF viewers (PDF.js included) try to gracefully deal with those could probably be described as a best-effort solution anyway.
This patch also adjusts the existing `warn`/`info` messages to print the command name according to the PDF specification, rather than an internal PDF.js enumeration value. The former should be much more useful for debugging purposes.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1443140.
Since the current page will be explicitly scrolled into view *directly* afterwards anyway (compare with e.g. the `pagesRotation` code), trying to maintain the current position when re-applying the zoom level during changing of Scroll modes is redundant.
Note how in `BaseViewer.forceRendering` the Scroll mode is used to determine how pre-rendering will work. Currently this is broken in Presentation Mode, if horizontal scrolling was enabled prior to entering fullscreen.
Furthermore, there's a few additional cases where the `this.scrollMode === ScrollMode.HORIZONTAL` check is pointless either in Presentation Mode or when a `PDFSinglePageViewer` instance is used.
Since the Scroll/Spread modes doesn't make (any) sense in `PDFSinglePageViewer` instances, the general structure of these methods can be improved to reflect that.
Given that this method is a no-op in `PDFSinglePageViewer`, similar to `_regroupSpreads`, let's improve the general code structure by simply moving the method.
There's no good reason to iterate through an arbitrary number of DOM elements this way, since a document could contain thousands of pages, when everything can be easily removed at once; compare with e.g. `BaseViewer._resetView` and `PDFThumbnailViewer._resetView`.
Furthermore given that it's a `PDFViewer` instance, the `this.viewer` property can be accessed directly. Besides, `_setDocumentViewerElement` only exists as a helper method for `setDocument` in the base class and none of this code applies for `PDFSinglePageViewer` instances either.
The Secondary Toolbar buttons for, not to mention the actual toggling of, Scroll/Spread modes are currently completely broken in older browsers (such as IE11).
As a follow-up, it'd probably be a good idea to try and find a *feature complete* `classList` polyfill that could be used instead, but this patch at least addresses the immediate regression.
Please refer to the compatibility information in https://developer.mozilla.org/en-US/docs/Web/API/Element/classList#Browser_compatibility
This patch avoids choosing a (possible) 'trailer' dictionary that `XRef.parse` and/or the `Catalog` constructor/methods will reject anyway.
Since `XRef.indexObjects` is already parsing the entire PDF file, the extra dictionary look-ups added here shouldn't matter much. Besides, this is a fallback code-path that only applies to corrupt PDF files anyway.
Note that the `Catalog` constructor, and some of its methods, are already enforcing that the 'Root' dictionary is valid/well-formed. However, by doing additional validation already in `XRef.parse` there's a slightly larger chance that corrupt PDF files could be successfully parsed/rendered.
With the current code line-breaks are accepted not just after an operator, but after a decimal point as well. When looking at this again, the latter case seems prone to cause false positives and might also interfere with subsequent patches.
Hence this is code is adjusted to actually do what the original commit message says, and nothing more.
The `onOpenWithURL` method, in `PDFViewerApplication.initPassiveLoading`, accepts a `originalURL` parameter which is then passed on to `PDFViewerApplication.open` as is. However, the latter method expects the name of the parameter to be `originalUrl` (note the casing), meaning that `getDocument` will fail in this case.
For consistency, and to avoid confusion, the renaming is done in `web/chromecom.js` as well.
Please note that the standalone `pdf.image_decoders.js` file will be including the complete `src/shared/util.js` file, despite only using parts of it.[1] This was done *purposely*, to not negatively impact the readability/maintainability of the core PDF.js code.
Furthermore, to ensure that the compatibility is the same in the regular PDF.js library *and* in the the standalone image decoders, `src/shared/compatibility.js` was included as well.
To (hopefully) prevent future complaints about the size of the built `pdf.image_decoders.js` file, a few existing async-related polyfills are being skipped (since all of the image decoders are completely synchronous).
Obviously this required adding a couple of pre-processor statements, but given that these are all limited to "compatibility" code, I think this might be OK!?
---
[1] However, please note that previous commits moved `PageViewport` and `MessageHandler` out of `src/shared/util.js` which reduced its size.
The purpose of this patch is to hopefully provide *slightly* better user ergonomics, if/when the PDF.js image decoders are used standalone.
This implementation is (basically) reverting the changes in PR 9386, in conjunction with code from the `parse` method found at https://github.com/notmasteryet/jpgjs/blob/master/src/pdfjs.js
*The danger of fixing one bug is that it can, sometimes too easily, cause another one in the process; sorry for not catching this when testing PR 9816 locally.*
When the *entire* viewer becomes narrow enough, as controlled by the `@media all and (max-width: 840px)` media query, the sidebar should (semi-transparently) overlay the `viewerContainer` instead of moving it horizontally. Unfortunately the changes made in PR 9816 caused the relevant CSS rules to be skipped, because of how the inheritance model works in CSS.
I'm well aware that `!important` is usually advised against, since it "breaks" the CSS inheritance model. However in this case it seemed reasonable to use it, to not only fix the bug at hand but to also prevent similar bugs from occurring in the future.
This is a regression from PR 8993; it causes the pages to be offset horizontally in Presentation Mode, if and only if the sidebar is currently open when the user triggers Presentation Mode.
Please note that while this doesn't seem to affect Firefox, both Chrome and IE are however affected.
Interestingly enough, despite the Chrome extension being affected as well, I cannot find any issue filed about this. (Either Presentation Mode isn't used much at all, or users don't open the sidebar before entering Presentation Mode.)
It appears that Microsoft silently fixed the problem that required disabling of fullscreen mode, in e.g. `iframe`s, in IE 11; please see issue 4711 and PR 5525 for historical context.
Unfortunately my Google-fu isn't strong enough to find any *official* information regarding the fixing of the browser bug in IE. However testing of the default viewer in IE 11, with this patch applied, it now appears that Presentation Mode is working correctly even in an `iframe` in IE 11.
Further anecdotal evidence that the bug is in fact fixed, is for example that jQuery previously contained a work-around for the IE bug. However, that's removed over two years ago now; see https://github.com/jquery/jquery/commit/ff1a0822f72d2b39fac691dfcceab6ede5623b90 and the issues referenced there.
Given that the default viewer isn't intended to be used as-is anyway (in custom deployments), it didn't seem necessary to keep the `disableFullscreen` option around since it was *only* ever added for compatibility purposes.
Fixes 9585.
Obviously it's still not possible to render non-embedded fonts as paths, but in this way the rest of the page will at least be allowed to continue rendering.
*Please note:* Including the 14 standard fonts in PDF.js probably wouldn't be *that* difficult to implement. (I'm not a lawyer, but the fonts from PDFium could probably be used given their BSD license.)
However, the main blocker ought to be the total size of the necessary font data, since I cannot imagine people being OK with shipping ~5 MB of (additional) font data with Firefox. (Based on the reactions when the CMap files were added, and those are only ~1 MB in size.)
Since `ColorSpace` now depends on the native clamping of `Uint8ClampedArray`, this patch adds non-production/test-only `assert`s to enforce that the expected TypedArray is used for the output.
These `assert`s are purposely *not* included in PRODUCTION builds since that would break rendering completely, as opposed to "only" displaying some weird colours, when a `Uint8Array` was used. Furthermore, these are mostly added to help catch explicit developer errors when working with the `ColorSpace` and `PDFImage` code.
Since the tests (currently) run with the `pdf.worker.js` file built, i.e. with `PRODUCTION = true` set, there's no simple way to add e.g. `assert` calls for both non-production *and* test-only builds without also affecting PRODUCTION builds.
The built-in image decoders are already using `Uint8ClampedArray` when returning data, and this patch simply extends that to the rest of the image/colorspace code.
As far as I can tell, the only reason for using manual clamping/rounding in the first place was because TypedArrays used to be polyfilled (using regular arrays). And trying to polyfill the native clamping/rounding would probably have been had too much overhead, but given that TypedArray support is required in PDF.js version `2.0` that's no longer a concern.
*Please note:* Because of different rounding behaviour, basically `Math.round` in `Uint8ClampedArray` respectively `Math.floor` in the old code, there will be very slight movement in quite a few existing test-cases. However, the changes should be imperceivable to the naked eye, given that the absolute difference is *at most* `1` for each RGB component when comparing `master` and this patch (see also the updated expectation values in the unit-tests).
The built-in image decoders are already returning data as `Uint8ClampedArray`, and subsequently the JPEG/JBIG2/JPX streams are as well. However, for general streams we obviously don't want to force the use of `Uint8ClampedArray` unless an "Image" is actually being decoded.
Hence this patch, which adds a parameter that allows the caller of the `getBytes`/`peekBytes` methods to force a `Uint8ClampedArray` (rather than a `Uint8Array`) to be returned.
Not only is the `Util.loadScript` helper function unused on the Worker side, even trying to use it there would throw an Error (since `document` isn't defined/available in Workers).
Hence this helper function is moved, and its code modernized slightly by having it return a Promise rather than needing a callback function.
Finally, to reduced code duplication, the "new" loadScript function is exported and used in the viewer.
The various classes have `this._errored` and `this._reason` properties, where the first one is a boolean indicating if an error was encountered and the second one contains the actual `Error` (or `null` initially).
In practice this means that errors are basically tracked *twice*, rather than just once. This kind of double-bookkeeping is generally a bad idea, since it's quite easy for the properties to (accidentally) get into an inconsistent state whenever the relevant code is modified.
Rather than using a separate boolean, we can just as well check the "error" property directly (since `null` is falsy).
---
Somewhat unrelated to this patch, but `src/display/node_stream.js` is currently *not* handling errors in a consistent or even correct way; compared with `src/display/network.js` and `src/display/fetch_stream.js`.
Obviously using the `createResponseStatusError` utility function, from `src/display/network_utils.js`, might not make much sense in a Node.js environment. However at the *very* least it seems that `MissingPDFException`, or `UnknownErrorException` when one cannot tell that the PDF file is "missing", should be manually thrown.
As is, the API (i.e. `getDocument`) is not returning the *expected* errors when loading fails in Node.js environments (as evident from the `pending` API unit-test).
Since the old comment mentions a now unsupported browser, let's update it such that someone won't accidentally conclude that the code in question can be removed.
This special handling was added in PR 8567, but was made redundant in PR 8721 which stopped sending everything but the kitchen sink to the Worker side.
Since `PDFPageProxy` already provide getters for all the data returned by `GetPage` (in the Worker), there isn't any compelling reason for accessing the `pageInfo` directly on `PDFPageProxy`.
The patch also changes the `GetPage` handler, in `src/core/worker.js`, to use modern JavaScript features.
Since `PDFDocumentProxy` already provide getters for all the data returned by `GetDoc` (in the Worker), there isn't any compelling reason for accessing the `pdfInfo` directly on `PDFDocumentProxy`.
After PR 8617 the `PDFManagerReady` message handler function, in `src/display/api.js`, is now a no-op. Hence it seems completely unnecessary to keep sending this message from `src/core/worker.js`.
With native typed array support now being mandatory in PDF.js, since version 2.0, this probably isn't a huge problem even though the current code seems wrong (it was changed in PR 6571).
Note how in the `!(data instanceof Uint8Array)` case we're currently attempting to send `handler.send('test', 'main', false);` to the main-thread, which doesn't really make any sense since the signature of the method reads `send(actionName, data, transfers) {`.
Hence the data that's *actually* being sent here is `'main'`, with `false` as the transferList, which just seems weird. On the main-thread, this means that we're in this case checking `data && data.supportTypedArray`, where `data` contains the string `'main'` rather than being falsy. Since a string doesn't have a `supportTypedArray` property, that check still fails as expected but it doesn't seem great nonetheless.
Since all the built-in PDF.js image decoders now return their data as `Uint8ClampedArray`, for consistency `JpegDecode` on the main-thread should be doing the same thing; follow-up to PR 8778.
The signature of the `PDFWorker.fromPort` method, in addition to the `PDFWorker` constructor, was changed in PR 9480.
Hence it's probably a good idea to add a bit more validation to `PDFWorker.fromPort`, to ensure that it won't fail silently for an API consumer that updates to version 2.0 of the PDF.js library.
With version 2.0, native support for typed arrays is now a requirement for using the PDF.js library; see PR 9094 where the old polyfills were removed.
Hence the `isTypedArraysPresent` check, when setting up fake workers, no longer serves any purpose here and can thus be removed.
There's no good reason, as far as I can tell, to duplicate the functionality of the `LoopbackPort` in the unit-tests. The only difference between the implementations is that `LoopbackPort` mimics the (native) structured cloning, however that shouldn't matter here since the tests are only sending "simple" data (strings respectively arrays with numbers).
Furthermore the patch also changes `LoopbackPort` to default to using "structured cloning" and deferred invocation of the listeners, since native typed array support is now a requirement for using the PDF.js library.
The `MessageHandler` itself, and its assorted helper functions, are currently the single largest[1] piece of code in the `src/shared/util.js` file. By moving this code into its own file, `src/shared/util.js` thus becomes smaller and more manageable.
The `fontScale` property was added in PR 1531, see commit https://github.com/mozilla/pdf.js/commit/b312719d7e0cf1a45dda0b4ad871a45162ed6f5b in particular, apparently for the sole purpose of supporting the "acroforms" example.
However, the `fontScale` property was never used anywhere else in the code-base, and after the modernization of the "acroforms" example in PR 8030 it's been completely unused.
Finally, note that there's also a (more suitably named) `scale` property on `PageViewport` instances, which contains the exact same information as the property being removed here.
To avoid having to manually update the "shared files" list in `gulpfile.js`, whenever any changes are made to the file structure in the `src/shared/` folder, let's simply list the files we do *not* want instead.
Given that the excluded files have been consistent ever since the `lib` build target was added, this patch should help avoid unnecessary churn in `gulpfile.js` in the future.
I made some mistakes when trying to make the content_disposition.js
compatible with non-modern browsers (IE/Edge).
Notably, text decoding was usually skipped because of the inverted
logical check at the top of `textdecode`.
I verified that this new version works as expected, as follows:
1. Visit https://github.com/Rob--W/open-in-browser/tree/55c71eb44e0ad71a3bb443457666fd48421612ac/test/
and get test-content-disposition.js
also get test-content-disposition.node.js if using Node.js,
or get test-content-disposition.html if you use a browser.
2. Modify `test-content-disposition.node.js` (or the HTML file) and
change `../extension/content-disposition.js` to `PDFJS-content_disposition.js`
3. Copy the `getFilenameFromContentDispositionHeader` function from
`content_disposition.js` (i.e. the file without the trailing exports)
and save it as `PDFJS-content_disposition.js`.
4. Run the tests (`node test-content-disposition.node.js` or by opening
`test-content-disposition.html` in a browser).
5. Confirm that there are no failures: "Finished all tests (0 failures)"
The code has a best-efforts fallback for Microsoft Edge, which lacks the
TextDecoder API. The fallback only supports the common UTF-8 encoding.
To simulate this in a test, modify `PDFJS-content_disposition.js` and
deliberately throw an error before `new TextDecoder`. There will be two
failures because we don't want to include too much code to support text
decoding for non-UTF-8 encodings in Edge
```
test-content-disposition.js:265 Assertion failed: Input: attachment; filename*=ISO-8859-1''%c3%a4
Expected: "ä"
Actual : "ä"
test-content-disposition.js:268 Assertion failed: Input: attachment; filename*=ISO-8859-1''%e2%82%ac
Expected: "€"
Actual : "€"
```
Compared to running the unit-tests in "regular" browsers, where any console output won't get mixed up with test output, in Node.js/Travis the test output looks quite noisy.
By ignoring `info`/`warn` calls, when running unit-tests in Node.js/Travis, the test output is a lot smaller not to mention that any *actual* failures are more easily spotted.
Please note that while the current code works, both in the viewer and the unit-tests, it can leave the `WorkerTransport._passwordCapability` Promise in a pending state.
In the `PasswordRequest` handler, in src/display/api.js, we're returning the Promise from a `capability` object (rather than just a "plain" Promise). While an error thrown anywhere within this handler was fortunately enough to propagate it to the Worker side, it won't cause the Promise (in `WorkerTransport._passwordCapability`) to actually be rejected.
Finally note that while we're now catching errors in the `PasswordRequest` handler, those errors are still propagated to the Worker side via the (now) rejected Promise and the existing `return this._passwordCapability.promise;` line.
This prevents warnings about uncaught Promises, with messages such as "Error: Worker was destroyed during onPassword callback", when running the unit-tests both in browsers *and* in Node.js/Travis.
This should provide a better out-of-the-box experience when using PDF.js in a Node.js environment, since it's missing native support for both `@font-face` and `Image`.
Please note that this change *only* affects the default values, hence it's still possible for an API consumer to override those values when calling `getDocument`.
Also, prevents "ReferenceError: document is not defined" errors, when running the unit-tests in Node.js/Travis.
This avoids the initialization of, potentially thousands of, unnecessary `Stream` objects, by getting the required number of bytes directly instead.
Given the special behaviour, when `length === 0`, of the `getBytes`/`skip` methods, it's also necessary to handle that particular case to prevent errors when encountering empty CharStrings.
This commit adds `scrollModeOnLoad` and `spreadModeOnLoad` preferences
that control the default viewer state when opening a new document for
the first time.
This commit also contains a minor refactoring of some of the option UI
rendering code in extensions/chromium/options/options.js, as I couldn't
bear creating two more functions nearly identical to the four that
already existed.
This builds on the scrolling mode work to add three buttons for joining
page spreads together: one for the default view, with no page spreads,
and two for spreads starting on odd-numbered or even-numbered pages.
Prior to this commit, if the vertical scroll bar is absent and the horizontal
scroll bar is present, a link to a particular point on the page which should
induce a horizontal scroll did not do so, because the absence of a vertical
scroll bar meant that the viewer was not recognized as the nearest scrolling
ancestor. This commit adds a check for horizontal scroll bars when searching
for the scrolling ancestor.
In `rasterizeAnnotationLayer` we load the source CSS files directly, so
these are not processed by Autoprefixer. Since the prefixed rules have
now been removed from the source CSS files, we must manually provide one
prefixed rule that Chrome needs in the overrides CSS file for checkbox
and radio button rendering to work in the reference tests.
SystemJS version 0.21.0 works, but version 0.21.1 breaks with various
`Unable to load transpiler to transpile` errors. Pin the version to
0.21.0 for now to avoid this issue.
We need to pass `disableFontFace` and `nativeImageDecoderSupport`
because Node.js has no native support for `@font-face` and `Image`.
Doing so makes it possible to render e.g., the Tracemonkey paper, which
failed before. I made this PDF file the default because it's also the
default in other examples/demos and because it showcases the
possibilities better than the very simple hello world PDF file.
Building the library with `gulp dist-install` is easier and is already
recommended in the other examples.
*This is a final piece of clean-up of code that I recently wrote, after which I'm done :-)*
When the `getMainThreadWorkerMessageHandler` function was added, in PR 9385, it did so by basically introducing a `web/app.js` dependency in `src/display/api.js` through the `window.pdfjsNonProductionPdfWorker` property[1]. Even though this is limited to non-`PRODUCTION` mode, i.e. `gulp server`, it still seems unfortunate to have that sort of viewer dependency in the API code itself.
With the new, much nicer and shorter, names introduced in PR 9565 we can remove this non-`PRODUCTION` hack and just use `window.pdfjsWorker` in both the viewer and the API regardless of the build mode.
---
[1] It didn't seem correct to piggy-back on the `window.pdfjsDistBuildPdfWorker` property in non-`PRODUCTION` mode.
Please note that this patch *purposely* doesn't add every standard (or semi-standard) page name in existence, but rather only a few common ones. This is done to lessen the burden on localizers, since it's quite possible that all of the page names could need translation (depending on locale).
It's easy to add more standard page sizes in the future, but we should take care to *only* add those that are very commonly used in actual PDF files.
This uses a whitelist, based on the locale, to determine where non-metric units should be used.
Note that the behaviour implemented here seem consistent with desktop PDF viewers (e.g. Adobe Reader), where the pageSizes are *always* displayed with locale dependent units rather than pageSize dependent ones (since the latter would probably be quite confusing).
A couple of basic unit-tests are added, and a manual `isLandscape` check (in `web/base_viewer.js`) is also converted to use the helper function instead.
Jasmine had a major version bump and required a few minor changes in our
booting code. Most notably, using `pending` in a `describe` block is no
longer supported, so we can only return early there. On the positive
side, the unit tests now run in a random order by default, which
eliminates any dependencies between unit tests.
Note that upgrading to Webpack 4 is out of scope for this patch since
the bots cannot work well with the newly generated bundles (both
browsers on both bots do not react within 120 seconds). Webpack 4 is not
faster for us than Webpack 3, so for now there is no need to upgrade.
Test case to exercise the different encodings:
1. Create a file "some file#@%M<br>%25 .pdf"
2. Build the extension with `gulp chromium` and load it in Chrome.
3. Go to `chrome://extensions/` and ensure that the
"Allow access to file URLs" is disabled.
4. Try to open the file from step 1 in Chrome (maybe reload once).
5. PDF.js should be showing a file chooser button.
6. Click on that button and select a different file.
Test: Check that a confirmation dialog pops up that warns about
a different file name. Cancel the dialog.
7. Click on the button again and select the original file.
Test: Check that the file opens as expected.
The current PageLabel dictionary validation code won't catch some (unlikely) forms of corruption. For example: a `Type`/`S` entry being `null`/`0`/empty string, a `P`/`St` entry being `null`/`0`.
Please note: I'm not aware of any bugs caused by the old code, but I've had this patch sitting locally for some time and figured it couldn't hurt to submit it.
The units are currently repeated after each dimension, which seems unnecessary and is also not done in other PDF viewers (such as e.g. Adobe Reader).
Furthermore, the name of the l10n arguments can be simplified slightly, since the name of the strings themselves should be enough information.
Finally, the `width`/`height` should be formatted according to the current locale, as is already done for other strings in the document properties dialog.
The `getPageSizeInches` method was implemented on `PDFDocumentProxy`, which seems conceptually wrong since the size property isn't global to the document but rather specific to each page. Hence the method is moved into `PDFPageProxy`, as `get pageSizeInches` instead to address this.
Despite the fact that new API functionality was implemented, no unit-tests were added. To prevent issues later on, we should *always* ensure that new functionality has at least some test-coverage; something that this patch also takes care of.
The new `PDFDocumentProperties._parsePageSize` method seemed unnecessary convoluted. Furthermore, in the "no data provided"-case it even returned incorrect data (an array, rather than the expected object).
Finally, the fallback strings didn't actually agree with the `en-US` locale. This inconsistency doesn't look too great, and it's thus addressed here as well.
This required changing the import script in two ways:
- we should use the `default` branch and not the `tip` tag since the
latter may refer to another branch than `default` (this is the case for
the `vi` locale, which caused in the files to be overwritten with
incorrect contents since `tip` referred to the
`THUNDERBIRD600b1_2018031614_RELBRANCH` branch);
- we should check if the response code is indeed 200 because recently a
script removed all empty localization files upstream (refer to
https://bugzilla.mozilla.org/show_bug.cgi?id=1443175).
PR #9493 moved from `appConfig.defaultUrl` to `AppOptions.get('defaultUrl')`.
However, it forgot to replace `appConfig.defaultUrl` in chromecom.js,
and as a result the extension is not able to open any PDF file.
This patch fixes that issue.
Chrome 60 and earlier does not include credentials (cookies) in requests
made with fetch, regardless of extension permissions. This was fixed in
61.0.3138.0 by
https://chromium.googlesource.com/chromium/src/+/2e231cf052ca5e68e22baf0008ac9e5e29121707
This patch disables the fetch backend in all affected Chrome versions.
The browser detection is done by checking for a change that coincides
with the release of Chrome 61.
Test case:
1. Copy the `isChromeWithFetchCredentials` function from the patch.
2. Run it in the JS console of Chrome and verify the return value.
Verified results:
- 49.0.2623.75 - false (earliest supported version by us)
- 60.0.3112.90 - false (last major version affected by bug)
- 61.0.3163.100 - true (first major version without bug)
- 65.0.3325.146 - true (current stable)
Test case 2:
1. Build the extension (`gulp chromium`) and load it in Chrome.
2. Open the developer tools, and open any PDF file.
3. In the "Network tab" of the developer tools, look at "request type".
In Chrome 60-: Should be "xhr"
In Chrome 61+: Should be "fetch"
This function combines the logic of two separate methods into one.
The loop limit is also a good thing to have for the calls in
`src/core/annotation.js`.
Moreover, since this is important functionality, a set of unit tests and
documentation is added.
It's only used in two places in the class and those callsites can
directly get the information from the dictionary, which is more readable
and avoids an additional method call.
With options being moved from the global `PDFJS` object and into `getDocument`, a side-effect is that we're now passing in a fair number of useless parameters to the various transport/network streams.
Even though this doesn't *currently* cause any problems, it nonetheless seem like a good idea to explicitly provide the parameters that are actually necessary.
One additional complication with removing this option from the global `PDFJS` object, is that the viewer currently needs to check `disableAutoFetch` in a couple of places. To address this I'm thus proposing adding a getter in `PDFDocumentProxy`, to allow checking the *actually* used values for a particular `getDocument` invocation.
The `appConfig` contains (mostly) references to various DOM elements, used when initializing the viewer components.
Hence `defaultUrl` seem like a slightly better fit for the new `AppOptions` abstraction, not to mention that it should thus be easier to set/modify it for custom deployments of the default viewer.
The way that various options are handled in the default viewer is currently a bit of a mess (to say the least). Some viewer options reside in the global `PDFJS` object, while others reside in `Preferences`. To make matters worse, some options even exist in both of the two.
Since the goal, with PDF.js version `2.0`, is to reduce our usage of the global `PDFJS` object, we'll instead want pass in the options when initializing the viewer components and when calling API methods (such as `getDocument`).
However given the current state of things in the default viewer, this wouldn't be exactly easy to implement. Hence this patch, which attempts to consolidate the way that viewer (and later API) options are handled by introducing a `AppOptions` singleton that provides *one* centralized way of interacting with the various options in the default viewer.
Unfortunately, as far as I can tell, we still need the ability to adjust certain API options depending on the browser environment in PDF.js version `2.0`. However, we should be able to separate this from the general compatibility code in the `src/shared/compatibility.js` file.
I don't understand why the previous way importing the polyfills didn't work, and I don't have time to try and figure it out, however this patch seems to fix things.
Fixes 9514.
Fixes 9516.
In a1cfa5f4d7, the textLayerMode
preference was introduced, to replace the disableTextLayer and
enhanceTextSelection preferences.
As a result, the text selection preference was no longer visible
in Chrome (because preferences are only rendered by default for
boolean preferences, not for enumerations).
This commit adds the necessary bits to
extensions/chromium/options/options.{html,js}
so that the textLayerMode preference can be changed again.
Also, migration logic has been added to move over preferences
from the old to the new names:
- In web/chromecom.js, the logic is added to translate
preferences that were set by an administrator (it is read-only,
so this layer is unavoidable).
- In extensions/chromium/options/migration.js, similar logic is
added, except in this case the preference storage is writable,
so this migration logic happens only once.
The "enhanced text selection" mode is still experimental, so it
has been marked as experimental to signal that there may be bugs.
The list of tasks that block promotion to stable is at #7584.
This partially reverts df0836b9b8.
The entry in preferences_schema.json is restored because that is
required to make managed preferences visible to the extension code.
The default key is still removed from default_preferences.json,
because this change only concerns the Chrome extension, not the
other parts of PDF.js. To account for the missing key, the
deprecated key was added back in chromecom.js
The key needs to be restored in preferences_schema.json too,
because that's the only way to make managed preferences visible.
I'm using `Object.assign`, which was introduced in Chrome 45,
so the preference module will break in Chrome 45 and earlier.
This is fine, because we do not support Chrome before 49.
Compared to most other options currently/previously residing on the global `PDFJS` object, some of the Worker specific ones (e.g. `workerPort`/`workerSrc`) probably cannot be moved into options provided directly when initializing e.g. `PDFWorker`.
The reason is that in some cases, e.g. the Webpack examples, we try to provide Worker auto-configuration and I cannot see a good solution for that use-case if we completely remove the globally available Worker configuration.
However inline with previous patches for PDF.js version `2.0`, it does seem like a worthwhile goal to move away from storing options directly on the global `PDFJS` object, since that is a pattern we should avoid going forward. Especially since one of the (eventual) goals is to attempt to *completely* remove the global `PDFJS` object, and rely solely on exporting/importing the needed functionality.
By introducing the `GlobalWorkerOptions` we thus have larger flexibility in the future, if/when the global `PDFJS` object will be removed.
With PDF.js version `2.0` we'll only support browsers with built-in `TypedArray` functionality, hence there doesn't seem to be any good reason not to implement this now.
Fixes 4888.
In order to simplify things, the undocumented `enableStats` option was removed and `pdfBug` is now instead used to enabled general debugging *and* page request/rendering stats.
Considering that in the default viewer the `stats` was only used when debugging was also enabled, this simplification (code wise) definitely seem worthwhile to me.
Rather than having two different (but connected) options for the textLayer, I think that it makes sense to try and unify this. For example: currently if `disableTextLayer === true`, then the value of `enhanceTextSelection` is simply ignored.
Since PDF.js version `2.0` already won't be backwards compatible in lots of ways, I don't think that we need to worry about migrating existing preferences here.
This removes the `PDFJS.externalLinkTarget`/`PDFJS.externalLinkRel` dependency from the viewer components, but please note that as a *temporary* solution the default viewer still uses it.
This removes the `PDFJS.imageResourcesPath` dependency from the viewer components and the test-suite, but please note that as a *temporary* solution the default viewer still uses it.
This removes the `PDFJS.maxCanvasPixels` dependency from the viewer components, but please note that as a *temporary* solution the default viewer still uses it.
This removes the `PDFJS.useOnlyCssZoom` dependency from the viewer components, but please note that as a *temporary* solution the default viewer still uses it.
Unfortunately, as far as I can tell, we still need the ability to adjust certain viewer options depending on the browser environment in PDF.js version `2.0`. However, we should be able to separate this from the general compatibility code in the `src/shared/compatibility.js` file.
The reason for the bug is that we're only computing a checksum of the image data itself, but completely ignore the inline dictionary. The latter is important, since in practice it's not uncommon for inline images to be identical but use e.g. different ColourSpaces.
There's obviously a couple of different ways that we could compute a hash/checksum of the dictionary.
Initially I tried using `MurmurHash3_64` to compute a hash of the keys/values in the dictionary. Unfortunately this approach turned out to be *way* too slow in practice, especially for PDF files with a huge number of inline images; in particular issue 2618 would regresses quite badly with this solution.
The solution that is instead implemented in this patch, is to compute a checksum of the dictionary contents. While this is a much simpler, not to mention a lot more efficient, solution there's one drawback associated with it:
If the contents of inline image dictionaries are ordered differently, they will not be considered equal with this approach which could thus lead to failures to cache repeated inline images. In practice this doesn't seem to be a problem in any of the PDF files I've tested, and generally I'd rather err on the side of *not* caching given that too aggressive caching can easily lead to rendering bugs.
One small, but somewhat annoying, complication is that by the time `Parser.makeInlineImage` is called, we no longer know the *exact* stream position where the inline image dictionary starts. Having access to that information is crucial here, and the easiest solution I could come up with is to track this in the current `Lexer` instance.[1]
With the patch, we're thus able to fix the referenced issues without incurring large regressions in problematic cases such as issue 2618.
Fixes 9398; also improves/fixes the `issue8823` reference test.
---
[1] Obviously I'd have preferred if this patch could be limited to `Parser.makeInlineImage`, without the need for this "hack", but I'm not sure what that'd look like here.
This rule is available from https://www.npmjs.com/package/eslint-plugin-mozilla, and is enforced in mozilla-central. Note that we have the necessary `Array`/`String` polyfills and that most cases have already been fixed, see PRs 9032 and 9434.
This rule is available from https://www.npmjs.com/package/eslint-plugin-mozilla, and is enforced in mozilla-central. Note that we have a polyfill for `ChildNode.remove()` and that most cases have already been fixed, see PRs 8056 and 8138.
Fallback to the built-in JPEG decoder when browser decoding fails, and attempt to handle JPEG images with DNL (Define Number of Lines) markers (issue 8614)
The code responsible for highlighting/scrolling thumbnails is quite old, and parts of it hasn't really changed much over time.
In particular, the `scrollThumbnailIntoView` method is currently using `document.querySelector` in order to find both the new/current thumbnail element. This seems totally unnessary, since we can easily keep track of the current thumbnail (similar to the "regular" viewer) and thus avoid unnecessary DOM look-ups.
Furthermore, note how the `PDFThumbnailView` constructor is currently highlighting the *first* thumbnail. This is yet another leftover from older viewer code, which we can now remove and instead take care of in `PDFThumbnailViewer.setDocument`.
Given that `PDFThumbnailView` does not, nor should it, have any concept of which thumbnail is currently selected, this change also improves the general structure of this code a tiny bit.
First of all, note how in both `fetch_stream.js` and `node_stream.js` we always overwrite the `this._contentLength` property even when the response headers doesn't actually contain any (valid) length information. This could thus result in the `length` parameter, as passed to the network stream, being completely ignored despite having no better information available.
Secondly, in `node_stream.js` the `this._isRangeSupported` property wasn't always updated correctly based on the response headers.
Please refer to the specification, found at https://www.w3.org/Graphics/JPEG/itu-t81.pdf#page=49
Given how the JPEG decoder is currently implemented, we need to know the value of the scanLines parameter (among others) *before* parsing of the SOS (Start of Scan) data begins.
Hence the best solution I could come up with here, is to re-parse the image in the *hopefully* rare case of JPEG images that include a DNL (Define Number of Lines) marker.
Fixes 8614.
This works by making `PartialEvaluator.buildPaintImageXObject` wait for the success/failure of `loadJpegStream` on the API side *before* parsing continues.
Please note that in practice, it should be quite rare for the browser to fail loading/decoding of a JPEG image. In the general case, it should thus not be completely surprising if even `src/core/jpg.js` will fail to decode the image.
Since `loadJpegStream` is only used at a *single* spot in the code-base, and given that it's very heavily tailored to the calling code (since it relies on the data structure of `PDFObjects`), this patch simply inlines the code in `src/display/api.js` instead.
This method currently requires a fair number of parameters, which creates quite unwieldy call-sites. When invoking `buildPaintImageXObject`, you have to remember not only which arguments to supply, but also the correct order, to prevent run-time errors.
After the latest update of the minimum supported Firefox version, the development addon is no longer being signed. Hence this patch attempts to address that, by pinning the `minVersion` to a specific version number instead.
*Please note:* The version numbers were taken from https://addons.mozilla.org/en-US/firefox/pages/appversions/, so it ought to have worked.
At this point in time, trying to keep the development addon compatible with prior Firefox versions is already quite difficult and will become even harder very soon.
Please keep in mind that since Firefox 57, only WebExtensions are allowed/possible to install. The only exceptions are Firefox Nightly, with the `xpinstall.signatures.required` preference[1] set to `false`, and the (as of this writing) current ESR release.[2]
With the current compatibility situation, we thus need to effectively support both Nightly *and* ESR in the addon, while trying to keep up with current/upcoming changes in `mozilla-central`. With old addons no longer being officially supported, the amount of old code being removed/refactored is now increasing quite quickly.
*Please note:* The changes proposed here was to a large extent prompted by bugs such as:
- https://bugzilla.mozilla.org/show_bug.cgi?id=1431533 (already landed)
- https://bugzilla.mozilla.org/show_bug.cgi?id=767640 (already landed)
- https://bugzilla.mozilla.org/show_bug.cgi?id=1432992 (still open)
Upstreaming all of those changes for the `MOZCENTRAL` version of PDF.js, while also keeping the Firefox addon running in older versions of the browser, would quickly become non-trivial.
Since we're using the ruleset from `eslint-plugin-mozilla` for the addon files, which is enforced in `mozilla-central`, we must ensure that the neccessary changes are upstreamed to the PDF.js repo such that the `mozilla-central` version of PDF.js can still be updated without failures.
Trying to feature detect, for the `FIREFOX` build target, some of the changes in the referenced bugs would probably become both quite messy and (not to mention) difficult. E.g. with the upcoming and automatically defined `Cc, Ci, Cu, Cr` variables, any sort of feature detection might be tricky since those need to be defined in the global scope of the files in question.
Finally, given the amount of effort that we'd now need to spend to even attempt to keep the Firefox addon compatible, I just don't think it's worth the effort any more. Especially since the number of people that have, thus far, been doing this work is *very* low and those resources would be better spend elsewhere.
Unfortunately, this probably means that the development addon will no longer be compatible with release versions of Seamonkey. However the README has already mentioned, for quite some time, that support isn't guaranteed.
*In closing:* For all of the reasons mentioned above, I thus propose that we reduce the maintenance burden of the Firefox addon by only supporting the current Firefox Nightly.
---
[1] While the preference exists, and can be toggled, its value is ignored in non-Nightly versions of Firefox.
[2] There's unbranded builds of e.g. the release/beta versions of Firefox, where old and non-WebExtensions addons can be installed. However those builds can probably be assumed to be officially unsupported, and thus not recommended for users.
This commit is the first step for extracting a base class for the
`AES128Cipher` and the `AES256Cipher` classes. The objective here is to
make code changes (not altering the logic) to make the implementations
as similar as possible as found by creating a diff of both classes.
In particular, we extract the key size and cycles of repetitions
constants since they are different for AES-128 and AES-256. Moreover, we
rename functions to be similar.
In the `AES256Cipher` class, there was an additional assignment to
`this` in the decryption function. However, this was unnecessary because
the assignment would also be done when the loop was exited.
In the JPEG images in the referenced PDF file, the DHT (Define Huffman Tables) segments contain more data than expected based on the length parameter.
Fixes 9425.
Despite this patch removing the `disableWorker` option itself, please note that we'll still fallback to loading the worker file(s) on the main-thread when running in environments without proper Web Worker support.
Furthermore it's still possible, even with this patch, to force the use of fake workers by manually loading the necessary file using a `<script>` tag on the main-thread.[1]
That way, the functionality of the now removed `SINGLE_FILE` build target and the resulting `build/pdf.combined.js` file can still be achieved simply by adding e.g. `<script src="build/pdf.worker.js"></script>` to the HTML (obviously with the path adjusted as needed).
Finally note that the `disableWorker` option is a performance footgun, and unfortunately many existing third-party examples actually use it without providing any sort of warning/justification.
---
[1] This approach is used in the default viewer, since certain kind of debugging may be easier if the code is running directly on the main-thread.
This method returns the currently used `workerSrc`, which thus allows obtaining the fallback `workerSrc` value (e.g. when the option wasn't set by the user).
Please note that this build target, and the resulting `build/pdf.combined.js` file, is equivalent to setting the `PDFJS.disableWorker` option to `true` which is a performance footgun.
https://crbug.com/362061 was fixed in Chrome 36, and the lowest
supported Chrome version in the extension is Chrome 49, so the
work-around for a filesystem:-bug in chromecom can be removed.
Merge ftp and file handler now their implementations are identical.
Remove redundant comment (the referenced Chrome bug has been fixed
a long time ago - https://crbug.com/302548 ).
This was introduced in #3582 to work around https://crbug.com/274024 .
The bug in Chrome has been fixed a long time ago (at least 33), so let's
simplify the code.
Add comments with supported browser versions where missing.
Method:
- Use MDN compat tables if available.
- Otherwise test in Chrome (31+) otherwise.
(the Chrome Web Store does not update older versions of
Chrome, so probably nobody is interested in even older
versions, even though there is an existing comment for
Chrome<29 at `document.currentScript`).
It turns out that PR 9245 unfortunately broke benchmarking completely, sorry about that!
The bug is that we were attempting to reset the current instance of `StatTimer`, instead of creating a new one as was previously done. By resetting the current instance, the `StatTimer` data fetched in `test/driver.js` is now wiped out since it points to the *same* underlying object.
This re-use of a `StatTimer` instance was asked for during review, and unfortunately I didn't test this thoroughly enough before submitting the final version of the PR.[1]
---
[1] Note that while I did test the benchmarking scripts with that PR *before* initially submitting it, I did however forget to do that after addressing the review comments which might explain why this problem went unnoticed.
Remove "returns null when content disposition is form-data".
The name of the test is already misleading: It suggests that
the return value is null if the Content-Disposition starts with
"form-data". This is not the case, anything with the "filename"
parameter is accepted.
So, to correct this, one would have to rephrase the test description to
"returns null when content disposition has no filename".
But this is already tested by the test called
"gets the filename from the response header".
So, remove the test.
This patch updates the `IPDFStreamReader` interface and ensures that the interface/implementation of `network.js`, `fetch_stream.js`, `node_stream.js`, and `transport_stream.js` all match properly.
The unit-tests are also adjusted, to more closely replicate the actual behaviour of the various actual `IPDFStreamReader` implementations.
Finally, this patch adjusts the use of the Content-Disposition filename when setting the title in the viewer, and adds `PDFDocumentProperties` support as well.
The `fetch` API is only supported for http(s), even in Chrome extensions.
Because of this limitation, we should use the XMLHttpRequest API when the
requested URL is not a http(s) URL.
Fixes#9361
These were removed in PR 9170, since they were unused in the browsers that we'll support in PDF.js version `2.0`.
However looking at the output of Travis, where a subset of the unit-tests are run using Node.js, there's warnings about `btoa` being undefined. This doesn't appear to cause any errors, which probably explains why we didn't notice this before (despite PR 9201).
When navigating away from the viewer, there's no good reason for disallowing replacement of a *temporary* history entry (in addition to empty ones).
Given that the current position is temporarily added to the browser history using a (short) timeout, the history entry will most likely already be correct when the 'pagehide' event fires. However, if the user is quick enough that might not always be the case, in which case the adjusted logic may help.
Please refer to the PDF specification, in particular http://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#G7.3801570
> A colour space shall be specified in one of two ways:
> - Within a content stream, the CS or cs operator establishes the current colour space parameter in the graphics state. The operand shall always be name object, which either identifies one of the colour spaces that need no additional parameters (DeviceGray, DeviceRGB, DeviceCMYK, or some cases of Pattern) or shall be used as a key in the ColorSpace subdictionary of the current resource dictionary (see 7.8.3, "Resource Dictionaries"). In the latter case, the value of the dictionary entry in turn shall be a colour space array or name. A colour space array shall never be inline within a content stream.
>
> - Outside a content stream, certain objects, such as image XObjects, shall specify a colour space as an explicit parameter, often associated with the key ColorSpace. In this case, the colour space array or name shall always be defined directly as a PDF object, not by an entry in the ColorSpace resource subdictionary. This convention also applies when colour spaces are defined in terms of other colour spaces.
This is something that I noticed while attempting to debug https://bugzilla.mozilla.org/show_bug.cgi?id=1374945.
Just looking at the code, the `YRsiz` parameter seemed immediately wrong and the fact that every component used the *same* data also looked strange.
Comparing with the specification, see https://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-REC-T.800-200208-S!!PDF-E&type=items#page=37, confirmed that this is indeed incorrect.
Note that I haven't got any example of a PDF file that is fixed by this patch, but that might be more luck than anything else. Manually checking a couple of files with included JPEG 2000 images, the `Csiz`/`XRsiz`/`YRsiz` parameters were `1` which could explain why this hasn't been an issue before.
Obviously we shouldn't generally make changes to `core` code without adding tests, but in this case I'm simply not sure how to obtain/create one. However, since the existing code doesn't make sense this patch could hopefully be deemed acceptable anyway.
Since multiple empty lines is virtually unused in the code-base, and the few cases that do exist look like "typos", let's enforce greater consistency here; please see https://eslint.org/docs/rules/no-multiple-empty-lines.
It is only used in a few places to handle prefixing style properties if
necessary. However, we used it only for `transform`, `transformOrigin`
and `borderRadius`, which according to Can I Use are supported natively
(unprefixed) in the browsers that PDF.js 2.0 supports. Therefore, we can
remove this class, which should help performance too since this avoids
extra function calls in parts of the code that are called often.
The `gulp-util` module is now deprecated and authors are asked to stop
using it (refer to https://medium.com/gulpjs/gulp-util-ca3b1f9f9ac5 for
more information).
PDF.js does not rely on it that much, fortunately, so it's relatively
easy for us to remove the dependency. This patch does that by making the
following changes:
- Require `gulp-zip` version 4.1.0 or higher since they already removed
their `gulp-util` dependency in that version.
- Replace `gulp-util.log` with the `fancylog` module as recommended in
the article above.
- Replace `gulp-util.File` with the `Vinyl` module as recommended in the
article above.
The only change I had to make for Vinyl is removing the `base` and `cwd`
lines since they may not be empty strings anymore. This way we fall back
to the defaults Vinyl provides, which for us doesn't matter since we
move the file afterwards anyway. Moreover, I used `vfs` for `vinyl-fs`
in the `Gulpfile` to avoid confusion with `vinyl` (which is also how the
documentation names the variable).
This is all we can do on our side; the other modules that still use
`gulp-util` must be updated upstream.
I've been looking into the remaining point in 8637 about blurry images, to see if we could perhaps improve the rendering quality slightly there. After quite a bit of debugging, it seems that the issue is limited to certain progressive JPEG images.
As mentioned previously, I've got no detailed knowledge of the JPEG format, but this patch does seem to improve things quite a bit for the images in question.
Squinting at https://searchfox.org/mozilla-central/rev/6c33dde6ca02b389c52e8db3d22494df8b916f33/media/libjpeg/jdphuff.c#492-639, it seems reasonable that we should take the sign of the data into account. Furthermore, looking at the specification in https://www.w3.org/Graphics/JPEG/itu-t81.pdf#page=118, the "F.2.4.3 Decoding the binary decision sequence for non-zero DC differences and AC coefficients" section even contains a description of this (even though I cannot claim to really understand the details).
The bug that this patch fixes is limited to the built-in JPEG decoder, and was unearthed by PR 9260. The underlying issue has existed since PR 6984, where the contents of this patch ought to have been included (if it weren't for the fact that we had no *easy* way to test `src/core/jpg.js` back then).
*Please note:* The slight movement in the reference test is a result of using the `src/core/jpg.js` decoder, rather than the native browser one.
This was an oversight in PR 9095, which unfortunately breaks rendering in some PDF files (e.g. the one from issue 6737).
It thus appears that we don't have any test-coverage for this code-path, and given the relative complexity of the PDF files affected by this bug I wasn't able to easily create a reduced test-case.
*Please note:* The linked test-case included in this patch is currently *not* rendered correctly (that'd be the PR 6606), but it at least gives us some test-coverage here.
Initially I just implemented the unit tests, but quickly found that they
were failing my expectation of having a size of 256 items. Some of them
did contain 256 items and some did not. I looked up various resources
and figured that they indeed all need to have 256 items. One of the good
resources is https://github.com/davidben/poppler/blob/master/poppler/FontEncodingTables.cc
Aside from some missing `notdef` (empty string) entries at the end of
the arrays, which I assume causes issues since it may cause
out-of-bounds array access which in JavaScript gives `undefined`, there
was a `notdef` entry missing in the `MacExpertEncoding`, causing the
entries after that to be shifted. This fix for this is similar to the
one in #8589.
The unit tests verify that, for known encoding names, the return value
is not only an array, but that it is also of the right length and
contains only strings.
The PDF file in the issue uses a number of *embedded* versions of Lucida fonts, but for some reason does *not* embed the LucidaSans-Demi font. According to https://en.wikipedia.org/wiki/Lucida#Usages that one should be bold, so we can at least improve rendering here (even though it won't look perfect).
Fixes 9291.
I recall being confused as to the purpose of the `encrypted` property all the way back when working on PR 4750.
Looking at the history, this property was added in PR 1698 when password support was added to the API/viewer. However, its only purpose seem to have been to facilitate the addition of a `isEncrypted` function in the API. That function never, as far as I can tell, saw any use and was unceremoniously removed in PR 4144.
Since we want to avoid sending all non-essential data early during initial document loading (e.g. PR 4750), it seems correct to get rid of the `encrypted` property. Especially since it hasn't even been exposed in the API for over three years, with no complaints that I'm aware of.
Finally note that the `encrypt` property on the `XRef` instance isn't tied to the code that's being removed here. Given that we're calling `PDFDocument.parse` during `createDocumentHandler` in the worker which, via `PDFDocument.setup`, calls `XRef.parse` where the `Encrypt` data (if it exists) is always parsed.
This patch refactors the searching for 'endobj', to try and find the next occurance of "obj" and then check if it was in fact an 'endobj' and continue searching otherwise.
This approach is used to avoid having to first find 'endobj', and then re-check the entire contents of the object and having to run (potentially expensive) regular expressions on arbitrary long strings.
Fixes 9105.
Please note that while this could be considered a regression in user-facing behaviour, I'm not convinced that it's really a regression as such since prior to PR 8912 the Metadata would fail to parse (with an XML error) and thus be ignored when setting the viewer title.
With the refactored Metadata parsing we're now able to parse this, which uncovered issues with a subset of broken Ghostscript Metadata that uses HTML character names.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1424938
It is quite confusing that the custom function is called `log2` while it
actually returns the ceiling value and handles zero and negative values
differently than the native function.
To resolve this, we add a comment that explains these differences and
make the function use the native `Math` functions internally instead of
using our own custom logic. To verify that the function does what we
expect, we add unit tests.
All browsers except for IE support `Math.log2` for quite a long time
already (see https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Math/log2).
For IE, we use the core-js polyfill.
According to the microbenchmark at https://jsperf.com/log2-pdfjs/1,
using the native functions should also be faster, in my testing almost
six times as fast.
Note that no other image stream implements a special `getBytes` method, which makes `JpegStream` look somewhat odd.
I'm actually not sure what purpose this methods serves, since I successfully ran all tests locally with it commented out. Furhermore, I also ran tests with an added `if (length && length !== this.bufferLength) { throw new Error('length mismatch'); }` check, and didn't get a single test failure in that case either.
Looking at the history, it seems that this code originated back in PR 4528, but as far as I can tell there's no mention in either commit messages nor PR comments of why it was necessary to add a "special" `getBytes` function for the `JpegStream`.
My assumption is that there's a good reason why this method was added, e.g. to address a *specific* regression in one of the reference tests. However, I did check out commit https://github.com/mozilla/pdf.js/commit/58f697f977f4a60974d91b09b2da67133227837c locally and ran tests with this method commented out, and there didn't seem to be any image-related failures in that case either!?
Hence I'm suggesting that we attempt to simplify this code slightly be removing this special `getBytes` method. However, please note that there's perhaps a *small* risk of regressions in an edge-case where we currently have insufficient test-coverage.
Unless the debugging tools (i.e. `PDFBug`) are enabled, or the `browsertest` is running, the `PDFPageProxy.stats` aren't actually used for anything.
Rather than initializing unnecessary `StatTimer` instances, we can simply re-use *one* dummy class (with static methods) for every page. Note that by using a dummy `StatTimer` in this way, rather than letting `PDFPageProxy.stats` be undefined, we don't need to guard *every* single stats collection callsite.
Since it wouldn't make much sense to attempt to use `PDFPageProxy.stats` when stat collection is disabled, it was instead changed to a "private" property (i.e. `PDFPageProxy._stats`) and a getter was added for accessing `PDFPageProxy.stats`. This getter will now return `null` when stat collection is disabled, making that case easy to handle.
For benchmarking purposes, the test-suite used to re-create the `StatTimer` after loading/rendering each page. However, modifying properties on various API code from the outside in this way seems very error-prone, and is an anti-pattern that we really should avoid at all cost. Hence the `PDFPageProxy.cleanup` method was modified to accept an optional parameter, which will take care of resetting `this.stats` when necessary, and `test/driver.js` was updated accordingly.
Finally, a tiny bit more validation was added on the viewer side, to ensure that all the code we're attempting to access is defined when handling `PDFPageProxy` stats.
While this doesn't actually fix the underlying issue, it should prevent the ESLint errors and thus make future PDF.js updates easier.
Compared to updating (and testing) the preprocessor, this seems like a reasonable workaround given its simplicity.
There's a number of issues with the fonts in the referenced PDF file. First of all, they contain broken `ToUnicode` data (`NUL` bytes all over the place). However even if you skip those, the `ToUnicode` data appears to contain nothing but a `IdentityH` CMap which won't help provide a proper glyph mapping.
The real issue actually turns out to be that the PDF file uses the "Calibri" font[1], but doesn't include any font files. Since that one isn't a standard font, and uses a fairly different CID to GID map compared to the standard fonts, we're not able to render the file even remotely correct.
To work around this, I'm thus proposing that we include a (incomplete) glyph map for Calibri, and fallback to the standard Helvetica font. Obviously this isn't going to look perfect, but it's really the best that we can hope to achieve given that the PDF file is missing the necessary font data.
Finally, please note that none of the PDF readers I've tried (Adobe Reader, PDFium in Chrome) were able to extract the text (which isn't very surprising, given the broken `ToUnicode` data).
Fixes 9195.
---
[1] According to Wikipedia, see https://en.wikipedia.org/wiki/Calibri, Calibri is (primarily) a Windows font.
For sufficiently large page sizes, always limiting the minimum zoom level to 25% seem a bit too high. One example is pages, e.g. the first one, in:
Hence I think that it makes sense to lower `MIN_SCALE` slightly, since other PDF viewers (e.g. Adobe Reader) isn't limiting the minimum zoom level as aggressively. Obviously this will allow a greater number of pages to be visible at the same time in the viewer, but given that they will be small that shouldn't be an issue.
Note also that e.g. the `page-fit`/`page-width` zoom levels already allow `< MIN_SCALE` values, so I don't see why we shouldn't allow users the same functionality directly.
In some fonts, the included `ToUnicode` data is incomplete causing text-selection to not work properly. For simple fonts that contain encoding data, we can manually build a `ToUnicode` map to attempt to improve things.
Please note that since we're currently using the `ToUnicode` data during glyph mapping, in an attempt to avoid rendering regressions, I purposely didn't want to amend to original `ToUnicode` data for this text-selection edge-case.
Instead, I opted for the current solution, which will (hopefully) give slightly better text-extraction results in PDF file with incomplete `ToUnicode` data.
According to the PDF specification, see [section 9.10.2](http://www.adobe.com/content/dam/acom/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#G8.1873172):
> A conforming reader can use these methods, in the priority given, to map a character code to a Unicode value.
> ...
Reading that paragraph literally, it doesn't seem too unreasonable to use *different* methods for different charcodes.
Fixes 8229.
This was added, during the refactoring in PR 8556, to avoid outright breaking third-party users of the default viewer.
With PDF.js version `2.0`, where we're making API changes that aren't backwards compatible, we ought to be able to remove this piece of viewer code as well.
Split the existing `WebGLUtils` in two classes, a private `WebGLUtils` and a public `WebGLContext`, and utilize the latter in the API to allow various code to access the methods of `WebGLUtils`
We've never attempted to limit the errors displayed in the default viewer to the current PDF file, but that's not really been a problem before. However after PR 7926, it's now possible to get password related error messages for *previously* opened PDF files in the default viewer.
**STR:**
1. Open a password protected PDF file, e.g. `issue6010_1.pdf` from the test-suite.
2. Cancel the password prompt.
3. Open any new PDF file in the viewer.
**AR:**
The error UI is displayed, with a `No password given` message.
**ER:**
No error displayed, since it's only relevent for a now closed PDF file.
This is obviously a minor issue, caused by us now rejecting the still pending `pdfLoadingTask` during the `PDFViewerApplication.close` call, but I don't think that it (generally) makes sense to show errors if they're not relevant to the *currently* displayed PDF file.
adding commas each in two sentences in order to make readers fully understand when reading index.md and to organize the context properly based on grammatical rules.
In order to move viewer related options from the global `PDFJS` object and into the initialization of the relevant components, we'll need to parse the hash parameters *before* calling `PDFViewerApplication._initializeViewerComponents`.
We're currently disabling range requests and streaming for a number of configurations. A couple of those will no longer be supported (with PDF.js version 2.0), hence we ought to be able to clean up the compatibility code slightly.
This compatibility code is only relevant for browsers that will no longer be supported (with PDF.js version 2.0), hence we ought to be able to remove it.
The only reason for adding this parameter in the first place, all the way back in PR 4074, was that the "maintain document position on zooming" feature was landed and backed out a couple of times before it finally stuck.
Hence it seemed, at the time, like a good idea to have a simple way to disable that behaviour. However, that was almost four years ago, and it's just not likely that we'd want/need to ever disable it now.
Furthermore I really cannot imagine why anyone would actually *want* to reset the position whenever zooming occurs, since it results in a quite annoying UX.
*So, to summarize:* Based on the above, I think that we should try to remove this parameter now. On the off chance that anyone complains, re-adding it shouldn't be difficult.
If the sidebar is resized such that the thumbnails are displayed in multiple columns, then scrolling the currently active thumbnail into view doesn't work correctly in some cases.
The reason is that the code in `PDFThumbnailViewer.scrollThumbnailIntoView` implicitly assumes that the thumbnails will be present in just *one* column. Since that may no longer be the case, it's not sufficient to simply check if the thumbnail is visible. Instead we must explicitly check that *all*, i.e. 100 percent, of the thumbnail is already visible, and otherwise scroll it into view.
*This patch is the result of me starting to look into moving parameters from `PDFJS` into `getDocument` and other API methods.*
When familiarizing myself with the code, the signatures of the various network streams seemed to be unnecessarily cumbersome since `disableRange` is currently handled separately from other parameters.
I'm assuming that the explanation for this is probably "for historical reasons", as is often the case. Hence I'd like to clean this up *before* we start the larger, and more invasive, `PDFJS` parameter re-factoring.
The interface of all of the "image" streams look kind of weird, and I'm actually a bit surprised that there hasn't been any errors because of it.
For example: None of them actually implement `readBlock` methods, and it seems more luck that anything else that we're not calling `getBytes()` (without providing a length) for those streams, since that would trigger a code-path in `getBytes` that assumes `readBlock` to exist.
To address this long-standing issue, the `ensureBuffer` methods are thus renamed to `readBlock`. Furthermore, the new `ensureBuffer` methods are now no-ops.
Finally, this patch also replaces `var` with `let` in a number of places.
By making use of modern CSS features, in this case [CSS variables](https://developer.mozilla.org/en-US/docs/Web/CSS/Using_CSS_variables), implementing sidebar resizing is actually quite simple. Not only will the amount of added code be fairly small, but it should also be easy to maintain since there's no need for complicated JavaScript hacks in order to update the CSS. Another benefit is that the JavaScript code doesn't need to make detailed assumptions about the exact structure of the HTML/CSS code.
Obviously this will not work in older browsers, such as IE, that lack support for CSS variables. In those cases sidebar resizing is simply disabled (via feature detection), and the resizing DOM element hidden, and the behaviour is thus *identical* to the current (fixed-width) sidebar.
However, considering the simplicity of the implementation, I really don't see why limiting this feature to "modern" browsers is a problem.
Finally, note that a few edge-cases meant that the patch is a bit larger than what the basic functionality would dictate. Among those is first of all proper RTL support, and secondly (automatic) resizing of the sidebar when the width of the *entire* viewer changes. Another, pre-existing, issue fixed here is the incomplete interface of `NullL10n`.
*Please note:* This patch has been successfully tested in both LTR and RTL viewer locales, in recent versions of Firefox and Chrome.
Fixes 2072.
https://eslint.org/docs/rules/no-var
Please note that two files were excluded:
1. `web/debugger.js`, since there's code in other files that currently depend on the global availability of code in `web/debugger.js`. Furthermore, since that file isn't used in production, doing a ES6 conversion probably isn't a priority.
2. `web/grab_to_pan.js`, since that file could be considered to be "external" code. We have made smaller changes to that file over the years, however doing a full ES6 `class` conversion might be a step too far!?
In the PDF file, the `ToUnicode` data first maps the hyphen correctly, and then *overwrites* it to point to the softhyphen instead. That one cannot be rendered in browsers, and an empty space thus appear instead.
Fixes 9084.
Based on a number of opened issues, it seems that the "Is an extension" field might be causing some confusion as to its meaning. Without providing too much detail, I'm still thinking that we could attempt to clarify that it's referring to either of the *browser* extensions.
Nothing uses this option anymore, so setting it is a no-op now. We can
safely remove it.
Use `SKIP_BABEL` (instead of `PDFJS_NEXT`) now if you want to skip Babel
translation for a build.
Since we're already using core-js elsewhere in `compatibility.js`, we can reduce the amount of code we need to maintain ourselves.
https://github.com/zloirock/core-js#weakmap
If we want to (eventually) make it possible to resize the sidebar, then having its width indirectly affect the toolbar is going to wreck havoc on the media queries used to show/hide buttons in the main toolbar (since many of them depend on the toolbar state, and thus its width).
Updating all of the media queries dynamically with JavaScript seems like a non-starter, given that it'd cause *very* messy code. It thus seem to me that we'd need to fix the position of the sidebar, to have any hope of (in the short term) addressing issue 2072.
Hence, I'm suggesting that the we always layout the sidebar in a consistent vertical position, and only animate the `viewerContainer` rather than the entire `mainContainer`.
Fixes 4052.
Fixes bug 850591.
Some PDF files contain JavaScript actions that consist of nothing more that one, or possibly several, empty string(s). At least to me, printing a warning/showing the fallback seems completely unnecessary in that case.
Furthermore, this patch also makes use of an early `return`, so that we no longer will attempt to check for printing instructions when no JavaScript is present in the PDF file.
*Note:* It would perhaps make sense to change the API/core code, such that we ignore empty entries there instead. However, that would probably be considered a breaking changing with respect to backwards compatibility, hence this simple viewer only solution.
Fixes 5767.
I don't know if this is a regression, but I noticed earlier today that depending on the initial scale *and* sidebar state, the `annotationLayer` of the first rendered page may end up duplicated; please see screen-shot below.
[screen-shot]
I can reproduce this reliable with e.g. https://arxiv.org/pdf/1112.0542v1.pdf#zoom=page-width&pagemode=bookmarks.
When the document loads, rendering of the first page begins immediately. When the sidebar is then opened, that forces re-rendering which thus aborts rendering of the first page.
Note that calling `PDFPageView.draw()` will always, provided an `AnnotationLayerFactory` instance exists, call `AnnotationLayerBuilder.render()`. Hence the events described above will result in *two* such calls, where the actual annotation rendering/updating happens asynchronously.
For reasons that I don't (at all) understand, when multiple `pdfPage.getAnnotations()` promises are handled back-to-back (in `AnnotationLayerBuilder.render()`), the `this.div` property seems to not update in time for the subsequent calls.
This thus, at least in Firefox, result in double rendering of all annotations on the first page.
Obviously it'd be good to find out why it breaks, since it *really* shouldn't, but this patch at least provides a (hopefully) acceptable work-around by ignoring `getAnnotations()` calls for `AnnotationLayerBuilder` instances that we're destroying (in `PDFPageView.reset()`).
*It appears that this accidentally broken in PR 8775.*
Note that `PDFHistory.forward` is only used with certain named actions, and these aren't that commonly used, which ought to explain why this error managed to sneak in.
Steps to reproduce the issue (and verify the fix):
1. Navigate to e.g. http://mirrors.ctan.org/info/lshort/english/lshort.pdf
2. Click on a couple of links, or outline items, such that the history is populated with a few entries.
3. In the console, execute `PDFViewerApplication.pdfHistory.back()` one or more times, thus navigating back to a previous viewer position.
4. In the console, execute `PDFViewerApplication.pdfHistory.forward() one or more times.
At the last step above, no (forward) navigation happens with the current `master`; now compare with this patch.
Moreover, adjust one linked test case that did not conform to the
standard Web Archive URL format and adjust one linked test case because
the link was dead.
To support this, the following changes have to be made as well:
- Ignore `package-lock.json` since NPM creates it, but we should not
have it in the repository.
- Switch from `babel-preset-es2015` to `babel-preset-env` to resolve
the deprecation warning in the test logs. The latter is more recent
and flexible, but should be the same functionality-wise.
- `transform` now needs to have the `utf-8` encoding option provided.
If not given, it will call the callback with a `Buffer` object,
which results in an unhandled promise rejection since what is
returned from the callback is a string, not a `Buffer`.
This patch makes use of the existing `ignoreErrors` property in `src/core/evaluator.js`, see PRs 8240 and 8441, thus allowing us to attempt to recovery as much as possible of a page even when it contains broken XObjects.
Fixes 8702.
Fixes 8704.
Split the existing `PDFFunction` in two classes, a private `PDFFunction` and a public `PDFFunctionFactory``, and utilize the latter in `PDFDocument` to allow various code to access the methods of `PDFFunction`
*Follow-up to PR 8909.*
This requires us to pass around `pdfFunctionFactory` to quite a lot of existing code, however I don't see another way of handling this while still guaranteeing that we can access `PDFFunction` as freely as in the old code.
Please note that the patch passes all tests locally (unit, font, reference), and I *very* much hope that we have sufficient test-coverage for the code in question to catch any typos/mistakes in the re-factoring.
The `inline` parameter is passed to a number of methods/functions in `PDFImage`, despite not actually being used. Its value is never checked, nor is it ever assigned to the current `PDFImage` instance (i.e. no `this.inline = inline` exists).
Looking briefly at the history of this code, I was also unable to find a point in time where `inline` was being used.
As far as I'm concerned, `inline` does nothing more than add clutter to already very unwieldy method/function signatures, hence why I'm proposing that we just remove it.
To further simplify call-sites using `PDFImage`/`NativeImageDecoder`, a number of methods/functions are changed to take Objects rather than a bunch of (somewhat) randomly ordered parameters.
I don't have a good example at hand right know, but I recall seeing custom deployments of PDF.js that bundle a *specific* version of the `build/pdf.js` file and then set `PDFJS.workerSrc` to point to https://mozilla.github.io/pdf.js/build/pdf.worker.js.
That practice seems really bad since, besides (obviously) causing unnecessary server load, it will very quickly result in a version mismatch between the `pdf.js` and `pdf.worker.js` files in those PDF.js deployments.
Such a version mismatch could easily lead to either breaking errors, or even worse slightly inconsistent behaviour for an API call (if the API -> Worker interface changes, which does happen from time to time).
To avoid the problems described above, I'm thus proposing that we enforce that the versions of the `pdf.js` and `pdf.worker.js` files must always match.
The new `PDFSinglePageViewer` class extends the previously created abstract `BaseViewer` class.
There's *a lot* of existing functionality in `PDFViewer` that depends on all the pages being loaded and synchronously available, once the `setDocument` method has been called.
Given that initializing `PDFPageView` instances requires passing a DOM element to which the page is attached, the simplest solution I could come up with is to append all pages to a (hidden) document fragment and just swap them (one at a time) into the viewer when page switching occurs.
This patch introduces an abstract `BaseViewer` class, that the existing `PDFViewer` then extends. *Please note:* This lays the necessary foundation for the next patch.
Looking at `ColorSpace.parseToIR`, it will do one of the following things when called:
1. Return a String.
2. Return an Array.
3. Throw a `FormatError`.
4. In one case, return the result of *another* `ColorSpace.parseToIR` call.
However, under no circumstances will it ever return an `AlternateCS` instance.
Since it's often useful to understand why code, which has become unused, existed in the first place, let's grab a hard hat and a shovel and start digging through the history of this code :-)
The current condition was introduced in commit https://github.com/mozilla/pdf.js/commit/c198ec432335956ef564fae1a5e9e9c51bfbe53a, in PR 794, but it was actually already obsolete by that time.
The preceeding `instanceof SeparationCS` condition predates commit https://github.com/mozilla/pdf.js/commit/a7278b7fbc029f2fdf92a5bddeb5465cd0a2d124, in PR 700.
That condition was originally introduced all the way back in commit https://github.com/mozilla/pdf.js/commit/4e3f87b60c231061433f5cd07d38120380f97408, in PR 692. However, it was made obsolete by commit https://github.com/mozilla/pdf.js/commit/9dcefe1efcef436f22beef72ef30176e8da619bd, which is included in the very same PR!
Hence we're left with the conclusion that not only has this code be unused for *almost* six years, it was basically never used at all save for a few refactoring commits that're part of PR 692.
Bug 1392647 has a PDF where the default width of the font
is 0. It draws some charcodes that don't have glyphs, but
we were wrongly using the 1000 default width for these
charcodes causing some text to be overlapping.
The `DOMParser` is most likely overkill and may be less secure.
Moreover, it is not supported in Node.js environments.
This patch replaces the `DOMParser` with a simple XML parser. This
should be faster and gives us Node.js support for free. The simple XML
parser is a port of the one that existed in the examples folder with a
small regex fix to make the parsing work correctly.
The unit tests are extended for increased test coverage of the metadata
code. The new method `getAll` is provided so the example does not have
to access internal properties of the object anymore.
(for issue #6289)
This does the same for 16 bit as the existing 8 bit tiff predictor code, an addition of the last word to this word.
The last two "& 0xFF" may or may not be needed, I see this isn't done in the 8 bit code, but I'm not a JS developer.
Rather that registering a 'change' event listener on the `window`, which will thus (unnecessarily) fire in *a number* of other situations such as e.g. when the user changes the pageNumber or the current search term, we could/should just register it directly on the dynamically created `fileInput` DOM element instead.
I can see no really compelling reason why we actually need to listen for `file` changes on the `window` itself, and this way we're also able to keep the `fileInput` related code confined to one part of the code which should aid readability.
Furthermore, in custom deployments, there's less risk that we're going to interfere with "outside" code this way.
Finally, preprocessor guards were added to the `webViewerOpenFile` function, since that code doesn't make sense in e.g. the extension builds.
Currently `PDFFunction` is implemented (basically) like a class with only `static` methods. Since it's used directly in a number of different `src/core/` files, attempting to pass in `isEvalSupported` would result in code that's *very* messy, not to mention difficult to maintain (since *every* single `PDFFunction` method call would need to include a `isEvalSupported` argument).
Rather than having to wait for a possible re-factoring of `PDFFunction` that would avoid the above problems by design, it probably makes sense to at least set `isEvalSupported` globally for `PDFFunction`.
Please note that there's one caveat with this solution: If `PDFJS.getDocument` is used to open multiple files simultaneously, with *different* `PDFJS.isEvalSupported` values set before each call, then the last one will always win.
However, that seems like enough of an edge-case that we shouldn't have to worry about it. Besides, since we'll also test that `eval` is actually supported, it should be fine.
Fixes 5573.
Rather than (basically) duplicating the `SimpleLinkService` in `test/driver.js`, with potential test failuires if you forget to update the test mock, it seems much nicer to just re-use the viewer component.
Note that `SimpleLinkService` is already bundled into the `build/components/pdf_viewer.js` file. Hence we only need to expose it similar to the other viewer components in that file, and make sure that the `gulp components` command runs as part of the test-setup.
reference tests
This patch allows us to use the common styles as used by the viewer as a
baseline for the annotation layer reference tests. They are extended
with a small set of overrides to ensure that all elements are visible
during the test.
The overrides file now only contains the absolutely necessary rules to
make all elements visible and is therefore no longer an almost verbatim
copy of the common styles.
This patch provides a new unit tested factory for creating SVG
containers and elements. This code is duplicated twice in the
codebase, but with upcoming changes this would need to be duplicated
even more. Moreover, consolidating this code in one factory allows
us to replace it easily for e.g., supporting Node.js. Therefore, move
this to a central place and update/ES6-ify the related code.
Finally, we replace `setAttributeNS` with `setAttribute` because no
namespace is provided.
This changes both `PDFViewer` and `PDFThumbnailViewer` to return early in the `pagesRotation` setters if the rotation doesn't change.
It also fixes an existing issue, in `PDFViewer`, that would cause errors if the rotation changes *before* the scale has been set to a non-default value.
Finally, in preparation for subsequent patches, it also refactors the rotation code in `web/app.js` to update the thumbnails and trigger rendering with the new `rotationchanging` event.
When testing the new `PDFHistory` implementation in practice, I felt that the current value of `UPDATE_VIEWAREA_TIMEOUT` is too large to be truly useful.
The purpose of the timeout is to attempt to address (the PDF.js part of) https://bugzilla.mozilla.org/show_bug.cgi?id=1153393, and it's currently fairly easy for the user e.g. close the browser before the timeout had a change to finish.
Obviously, the timeout is a best-effort solution, but with the current value of `UPDATE_VIEWAREA_TIMEOUT` it's not as useful as one would want.
Please note that lowering it shouldn't be a problem, since it still prevents the browser history from updating at *every* 'updateviewarea' event or during (quick) scrolling, which is all that's really needed to not impact the UX negatively.
---
Furthermore, with this lower timeout, we can also simplify the part of the 'popstate' event handler that attempted to update the browser history with the current position before moving back. In most cases, the current position will now already exist in the history, and this *greatly* decreases the complexity of this code path.
The main impetus for this change though, is that I unfortunately found that given the asynchronous nature of updating the browser history, there is some *edge* cases where that code could cause history corruption.
In practice, the user could thus get "stuck" at a particular history entry and not be able to move back. I haven't got any reliable STR for this, since it's so difficult to trigger, but it involved navigating around in a document such that a number of destinations are added to the browser history and then changing the rotation before going back/forward in the history.
Rather that attempting to patch this code, and making it even more difficult to understand than it already is or adding more asynchronous behaviour, by far the easiest solution is to remove it and simply rely on the (lowered) `UPDATE_VIEWAREA_TIMEOUT` instead.
Since e.g. zooming can occur when navigating to a new destionation, ensure that a resulting 'updateviewarea' event doesn't trigger adding of a *temporary* position to the browser history at a bad time.
Rather than displaying links that does *nothing* when clicked, it probably makes more sense to simply not render them instead. Especially since it turns out that, at least at this point in time, this is *very* easy to both implement and test.
Fixes 3897.
It seems that the status check, for non-HTTP loads, causes the default viewer to *refuse* to open local PDF files.
***STR:***
1. Make sure that fetch support is enabled in the browser. In Firefox Nightly, set `dom.streams.enabled = true` and `javascript.options.streams = true` in `about:config`.
2. Open https://mozilla.github.io/pdf.js/web/viewer.html.
3. Click on the "Open file" button, and open a new PDF file.
***ER:***
A new PDF file should open in the viewer.
***AR:***
The PDF file fails to open, with an error message of the following format:
`Message: Unexpected server response (200) while retrieving PDF "blob:https://mozilla.github.io/a4fc455f-bc05-45b5-b6aa-2ecff3cb45ce".`
When looking briefly at using `Number.isInteger`/`Number.isNan` rather than `isInt`/`isNaN`, I noticed that there's a couple of not entirely straightforward cases to consider.
At first I really couldn't understand why `parseInt` is being used like it is in `XRef.fetchUncompressed`, since the `num` and `gen` properties of an object reference should *always* be integers.
However, doing a bit of code archaeology pointed to PR 4348, and it thus seem that this was a very deliberate change. Since I didn't want to inadvertently introduce any regressions, I've kept the `parseInt` calls intact but moved them to occur *only* when actually necessary.[1]
Secondly, I noticed that there's a redundant `isCmd` check for an edge-case of broken operators. Since we're throwing a `FormatError` if `obj3` isn't a command, we don't need to repeat that check.
In practice, this patch could perhaps be considered as a micro-optimization, but considering that `XRef.fetchUncompressed` can be called *many* thousand times when loading larger PDF documents these changes at least cannot hurt.
---
[1] I even ran all tests locally, with an added `assert(Number.isInteger(obj1) && Number.isInteger(obj2));` check, and everything passed with flying colours.
However, since it appears that this was in fact necessary at one point, one possible explanation is that the failing test-case(s) have now been replaced by reduced ones.
By using the (heuristic) `POSITION_UPDATED_THRESHOLD` constant, we can ensure that the current document position will be added to the browser history when a sufficiently "large" number of `updateviewarea` events have been dispatched.
This patch attempts to address an issue in the old `PDFHistory` implementation, where the current position wouldn't be correctly saved when the browser was closed.
In theory this *should* already be working, however as the discussion in https://bugzilla.mozilla.org/show_bug.cgi?id=1153393 showed, it seems that both `pagehide` and `beforeunload` arrive to late to successfully update the history during closing.
Hence a timeout is used to *temporarily* add the current position to the browser history when the viewer is idle.
Note that we need to take care not to update the browser history too often, since that would render the viewer more or unusable. Furthermore, if the timeout is *too* long it may end up effectively disable this whole functionality.
The `UPDATE_VIEWAREA_TIMEOUT` constant is thus a heuristic value, which we may need to tweak taking the above into account.
This patch completely re-implements `PDFHistory` to get rid of various bugs currently present, and to hopefully make maintenance slightly easier. Most of the interface is similar to the existing one, but it should be somewhat simplified.
The new implementation should be more robust against failure, compared to the old one. Previously, it was too easy to end up in a state which basically caused the browser history to lock-up, preventing the user from navigating back/forward. (In the new implementation, the browser history should not be updated rather than breaking if things go wrong.)
Given that the code has to deal with various edge-cases, it's still not as simple as I would have liked, but it should now be somewhat easier to deal with.
The main source of complication in the code is actually that we allow the user to change the hash of a already loaded document (we'll no longer try to navigate back-and-forth in this case, since the next commit contains a workaround).
In the new code, there's also *a lot* more comments (perhaps too many?) to attempt to explain the logic. This is something that the old implementation was serverly lacking, which is a one of the reasons why it was so difficult to maintain.
One particular thing to note is that the new code uses the `pagehide` event rather than `beforeunload`, since the latter seems to be a bad idea based on https://bugzilla.mozilla.org/show_bug.cgi?id=1336763.
The current implementation of `PDFHistory` contains a number of smaller bugs, which are *very* difficult to address without breaking other parts of its code.
Possibly the main issue with the current implementation, is that I wrote it quite some time ago, and at the time my understanding of the various edge-cases the code has to deal with was quite limited.
Currently `PDFHistory` may, despite most of those cases being fixed, in certain edge-cases lock-up the browser history, essentially preventing the user from navigating back/forward.
Hence rather than trying to iterate on `PDFHistory` to make it better, the only viable approach is unfortunately rip it out in its entirety and re-write it from scratch.
*It appears that this accidentally broke with PR 8394.*
Currently, the following will be printed in the console:
```
An error occurred while loading the PDF.
[object Promise],[object Promise]
```
With this patch we'll again get proper output, e.g. something with this format:
```
An error occurred while loading the PDF.
PDF.js v? (build: ?)
Message: unknown encryption method
```
By updating to uglify-es, rather than uglify-js, the minifier *itself* now supports ES6 code. This means that it's now possible to minify code built with `PDFJS_NEXT = true` set, i.e. with Babel transpilation disabled, which wasn't the case previously.
Note that uglify-es is based on the API of uglify-js v3, which differs from the one that we previously used.
Of particular importance is the fact that it's no longer possible to provide a path to a file for minification, but one must instead directly provide the source of the file.
For more information, please see https://github.com/mishoo/UglifyJS2/tree/harmony
Since this patch will now treat (some) `NUL` bytes as "ASCII", the number of `followingBytes` checked are thus increased to (hopefully) reduce the risk of introducing new false positives.
Fixes 8823.
According to the specification, see http://www.adobe.com/content/dam/Adobe/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#page=377, a `Dest` entry in an outline item should *not* contain a dictionary.
Unsurprisingly there's PDF generators that completely ignore this, treating is an `A` entry instead.
The patch also adds a little bit more validation code in `Catalog.parseDestDictionary`.
See https://eslint.org/docs/rules/for-direction; helps avoid typos that would cause infinite `for` loops.
Also, updates `eslint` and `eslint-plugin-mozilla` to the latest available versions.
Since the very early days of the viewer, it's been possible to pass in a `scale` when opening a PDF file. However, most of the time it was/is actually being ignored, which limits its usefulness considerably.
In older versions of the viewer, if a document hash was present (i.e. `PDFViewerApplication.initialBookmark` being set) or if the document existed in the `ViewHistory`, the `scale` passed to `PDFViewerApplication.open` would thus always be ignored.
In addition to the above, in the current viewer there's even more cases where the `scale` parameter will be ignored: if a (valid) browser history entry exists on document load, or if the `defaultZoomValue` preference is set to a non-default value.
Hence the result is that in most situation, a `scale` passed to `PDFViewerApplication.open` will be completely ignored.
A much better, not to mention supported, way of setting the initial scale is by using the `defaultZoomLevel` preference. In comparision, this also has the advantage of being used in situations where the `scale` would be ignored.
All in all this leads to the current situation where we have code which is essentially dead, since no part of the viewer (by default) relies on it.
To clean up this code, and to avoid having to pass (basically) unused parameters around, I'd thus like to remove the ability to pass a `scale` to `PDFViewerApplication.open`.
The property and the setter for text rise were already present, but they
were never used or called. This patch completes the implementation by
calling the setter when the operator is encountered and by using the
text rise value when rendering text.
See Bug 1389443 for more information. This allows us to register
PdfJs without waiting for file IO in nsHandlerService to finish.
Once that file IO is finished, we can set everything up properly
and double-check the registration.
The method signature was improved in PR 7440, which has now been present in a number of releases (starting with `v1.6.210`).
Hence we should be able to remove this now, and just print an error message if the old format is used.
This was added in PR 7793, which has now been present in a number of PDF.js releases (from version `v1.7.225`).
Hence we should be able to remove it now, considering that the migration code was only intended as a best effort solution to avoid wiping out all existing user data at once. Also, keep in mind that `ViewHistory` is already limited with regards to the number of documents it will simultaneous store data for.
I've gotten tired of having to manually re-install the `ttest` package on numerous occasions when benchmarking changes, so let's just list it in package.json to simplify things.
As discussed in PR 8673, we cannot solve the general issue (since that would require parsing every single page). However, we can mitigate the effect somewhat, by waiting for the FileAttachment annotations of the initially rendered page.
Rather than duplicating initialization code, we can just call `this.reset()` instead (which is also similar to other existing code, e.g. `PDFViewer`). This will also help ensure that the DOM is completely reset, before any outline items or attachments are displayed.
This bug is similar to the canvas bug of #6721.
I found this bug when I tried to run pdf2svg on a SVG file, and the generated
SVG could not be viewed in Chrome due to a SVG/XML parsing error:
"PCDATA invalid Char value 3"
Reduced test case:
- https://github.com/mozilla/pdf.js/files/1229507/pcdatainvalidchar.pdf
- expected: "hardware performance"
- Actual SVG source: "hardware\x03performance"
(where "\x03" is a non-printable character, and invalid XML).
In terms of rendering, this bug is similar to #6721, where an unexpected glyph
appeared in the canvas renderer. This was fixed by #7023, which skips over
missing glyphs. This commit follows a similar logic.
The test case from #6721 can be used here too:
- https://github.com/mozilla/pdf.js/files/52205/issue6721_reduced.pdf
expected: "Issue 6721"
actual (before this patch): "Issue ààà6721"
Implement a serialization "generator" for `DOMElement` in domutils.js
that yields the serialization of the SVG element. This method is used by
a newly added `ReadableSVGStream` class, which can be used like any
other readable stream in Node.js.
This reduces the memory requirements. Now, it is not needed to require
the serialization to fully fit in memory.
Note: The implementation of the serializer is a state machine in ES5
since the rest of the file is also in ES5. Its functionality is
equivalent to:
```
function* serializeSVGElement(elem) {
yield '<' + elem.nodeName;
if (elem.nodeName === 'svg:svg') {
yield ' xmlns:xlink="http://www.w3.org/1999/xlink"' +
' xmlns:svg="http://www.w3.org/2000/svg"';
}
for (let i in elem.attributes) {
yield ' ' + i + '="' + xmlEncode(elem.attributes[i]) + '"';
}
yield '>';
if (elem.nodeName === 'svg:tspan' || elem.nodeName === 'svg:style') {
yield xmlEncode(elem.textContent);
} else {
for (let childNode of elem.childNodes) {
yield* serializeSVGElement(childNode);
}
}
yield '</' + elem.nodeName + '>';
}
```
Since we're now using `Uint8ClampedArray`, rather than `Uint8Array`, doing manual clamping shouldn't be necessary given that that is now handled natively.
This shouldn't have any measurable performance impact, but just to sanity check that I've done some quick benchmarking with the following manifest file:
```json
[
{ "id": "S2-eq",
"file": "pdfs/S2.pdf",
"md5": "d0b6137846df6e0fe058f234a87fb588",
"rounds": 100,
"type": "eq"
}
]
```
which gave the following results against the current `master` (repeated benchmark runs didn't result in any meaningful differences):
```
-- Grouped By browser, stat --
browser | stat | Count | Baseline(ms) | Current(ms) | +/- | % | Result(P<.05)
------- | ------------ | ----- | ------------ | ----------- | --- | ----- | -------------
firefox | Overall | 100 | 592 | 592 | 1 | 0.12 |
firefox | Page Request | 100 | 3 | 3 | 0 | -9.88 |
firefox | Rendering | 100 | 588 | 589 | 1 | 0.18 |
```
When running browser tests, e.g. via `gulp unittest`, the test files are not
processed by babel, and neither by the "unittestcli" gulp target.
This commit copies the babelPluginReplaceNonWebPackRequire plugin from the
unittestcli target to the SystemJS config so that `__non_webpack_require__` is
replaced with `require` for all build targets, and adds a unit test to ensure
that this indeed works as expected.
This patch removes the `clamp0to255` helper function, as well as manual clamping code in `src/core/jpg.js`.
The adjusted constants in `_convertCmykToRgb` were taken from CMYK to RGB conversion code found in `src/core/colorspace.js`.
*Please note:* There will be some very slight movement in a number of existing test-cases, since `Uint8ClampedArray` appears to use `Math.round` (or equivalent) and the old code used (basically) `Math.floor`.
At this point, the default viewer is already not usable in older browsers in `gulp server` mode, since we only run the code through Babel as part of the build step.
Hence there shouldn't be much point in manually loading `compatibility.js` in `viewer.html` the way that we've been doing, especially considering that it's already being loaded by `src/shared/util.js`.
This appears to simply have been forgotten in the re-factoring in PR 4815, where the `coded` property was renamed to the *much* more descriptive `isType3Font` property.
From looking at blame, it seems that these checks became obsolete with PR 692 (which landed close to six years ago). Note how, after that PR, there's no longer anything being assigned to the `code` property of an Object.
In issue #8707, there's a char code mapped to a non-
existing glyph which shouldn't be drawn. However, we
saw it was missing and tried to then use the post table and
end up mapping it incorrectly.
This illuminated a problem with issue #5704 and bug
893730 where glyphs disappeared after above fix. This was
from the cmap returning the wrong glyph id. Which in turn was
caused because the font had multiple of the same type of cmap
table and we were choosing the last one. Now, we instead
default to the first one. I'm unsure if we should instead be
merging the multiple cmaps, but using only the first one works.
Added unit-tests for DeviceGray, DeviceRGB and DeviceCMYK
Added unit-tests for CalGray
Added unit-tests for CalRGB
Removed redundant code
Added unit-tests for LabCS
Added unit-tests for IndexedCS
Update comment
Change lookup to Uint8Array as mentioned in pdf specs(these tests will pass after PR #8666 is merged).
Added unit-tests for AlternateCS
Resolved code-style issues
Fixed code-style issues
Addressed issues pointed out in https://github.com/mozilla/pdf.js/pull/8611#pullrequestreview-52865469
The initial issue with #8255 was I added a missing glyphs
check to adjustMapping, but this caused us to skip re-mapping
a glyph if the fontCharCode was a missingGlyph which in turn
caused us to overwrite a valid glyph id with an invalid one. While
fixing this, I also added a warning if the private use area is full since
this also accidentally happened when I made a different mistake.
This brought to light a number of issues where we map
missing glyphs to notdef, but often the notdef is actually defined
and then ends up being drawn. Now the glyphs don't get
mapped in toFontChar and so they are not drawn by the canvas.
Fixing the above brought up another issue though in bug1050040.pdf.
In this PDF, the font fails to load by the browser and before we were still
drawing the glyphs because it looked like the font had them, but with the fixes
above the glyphs showed up as missing so we didn't attempt draw them. To
fix this, I now throw an error when the loca table is in really bad shape and
we fall back to trying to use a system font. We now also use this fall back if
there are any format errors during converting fonts.
The PDF file uses a non-embedded SegoeUISymbol font, which is *not* a standard font (and is mainly used by Microsoft, see https://en.wikipedia.org/wiki/Segoe).
Fixes 8697.
Issue 7696 was one of the issues fixed by PR 8580. The other ones were all cases of missing glyphs, however in this particular one glyphs did render but every single one was incorrect.
Hence it probably cannot hurt to have a small, reduced, reference test for that PDF file as well.
This replaces `assert` calls with `throw new FormatError()`/`throw new Error()`.
In a few places, throwing an `Error` (which is what `assert` meant) isn't correct since the enclosing function is supposed to return a `Promise`, hence some cases were changed to `Promise.reject(...)` and similarily for `createPromiseCapability` instances.
Fix TypeError that occurs in colorspace.js on accidentally passing an 'Array' instead of 'TypedArray'
Changed getRgbItem(...) to getRgbBuffer(...) since this.lookup has values in range[0, 255] whereas getRgbItem(...) expects those to be in range [0, 1]
Revert changes for IE9 compatibility
Looking at the blame, it seems that this typo was present even before PR 700 (almost six years ago).
The result of using `'num'`, rather than the *correct* `'numPages'` string, is that the `Catalog.numPages` getter isn't actually being shadowed.
Note that the PageMode, as specified in the API, will only be honoured when either: the user hasn't set the `sidebarViewOnLoad` preference to a non-default value, or a non-default `sidebarView` entry doesn't exist in the view history, or the "pagemode" hash parameter is included in the URL.
Since this is new functionality, the patch also includes a preference (`disablePageMode`), to make it easy to opt-out of this functionality if the user/implementor so wishes.
- Mark the test as async, and don't swallow exceptions.
- Fix the DOMElement polyfill to behave closer to the actual getAttributeNS
method, which excludes the namespace prefix.
Add UI for the cursorToolOnLoad pref in the UI of the Chrome extension.
Add logic to migrate the enableHandToolOnLoad pref to cursorToolOnLoad.
For past values in the mutable extension storage area:
1. If enableHandToolOnLoad=true, save cursorToolOnLoad=1.
2. Remove enableHandToolOnLoad.
For the managed extension storage, which is immutable since it is based
on administrative policies, use the following logic:
1. If enableHandToolOnLoad=true and cursorToolOnLoad=0 (default).
set cursorToolOnLoad=0 and assume enableHandToolOnLoad=false.
2. As usual, managed preferences can (and will) be overridden by the user.
The first migration logic is in extensions/chromium/options/migration.js
and can be removed after a few months / less than many years.
The second migration logic is in web/chromecom.js, and should be kept
around for a long while (many years).
The need for this migration logic arises from the change by:
https://github.com/mozilla/pdf.js/pull/7635
* Check for undefined
new URL(file, window.location.href) throws the following error in IE11 + iPad Safari:
Unable to get property 'replace' of undefined or null reference
* Adapting previous change to pdf.js code standards
Added curly braces
* Moved check for undefined above try/catch
According to the PDF specification, please see http://www.adobe.com/content/dam/Adobe/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#G6.2394361, if an Adobe JPEG marker is present it should always take precedence. This even seem to be consistent with the existing comment that is present in the code.
Hence it seems reasonable to interpret `transformCode === 0` as no color conversion being necessary.
Fixes the rendering of page 1 in `issue-4926` (from the test-suite), when the built-in `src/core/jpg.js` image decoder is used.
Use the environment's zlib implementation if available to get
reasonably-sized SVG files when an XObject image is converted to PNG.
The generated PNG is not optimal because we do not use a PNG predictor.
Futher, when our SVG backend is run in a browser, the generated PNG
images will still be unnecessarily large (though the use of blob:-URLs
when available should reduce the impact on memory usage). If we want to
optimize PNG images in browsers too, we can either try to use a DEFLATE
library such as pako, or re-use our XObject image painting logic in
src/display/canvas.js. This potential improvement is not implemented by
this commit
Tested with:
- Node.js 8.1.3 (uses zlib)
- Node.js 0.11.12 (uses zlib)
- Node.js 0.10.48 (falls back to inferior existing implementation).
- Chrome 59.0.3071.86
- Firefox 54.0
Tests:
Unit test on Node.js:
```
$ gulp lib
$ JASMINE_CONFIG_PATH=test/unit/clitests.json node ./node_modules/.bin/jasmine --filter=SVG
```
Unit test in browser: Run `gulp server` and open
http://localhost:8888/test/unit/unit_test.html?spec=SVGGraphics
To verify that the patch works as desired,
```
$ node examples/node/pdf2svg.js test/pdfs/xobject-image.pdf
$ du -b svgdump/xobject-image-1.svg
# ^ Calculates the file size. Confirm that the size is small
# (784 instead of 80664 bytes).
```
Do not directly export to global. Instead, export all stubs in domstubs.js and
add a method setStubs to assign all exported stubs to a namespace. Then replace
the import domstubs with an explicit call to this setStubs method. Also added
unsetStubs for undoing the changes. This is done to allow unit testing of the
SVG backend without namespace pollution.
Move the DEFLATE logic in convertImgDataToPng to a separate function.
A later commit will introduce a more efficient deflate algorithm,
and fall back to the existing, naive algorithm if needed.
`__pdfjsdev_webpack__` was used to skip evaluating part of an AST,
in order to not mangle some `require` symbols.
This commit removes `__pdfjsdev_webpack__`, and:
- Uses `__non_webpack_require__` when one wants the output to
contain `require` instead of `__webpack_require__`.
- Adds options to the webpack config to prevent "polyfills" for
some Node.js-specific APIs to be added.
- Use `// eslint-disable-next-line no-undef` instead of `/* globals ... */`
for variables that are not meant to be used globally.
Since we no longer, after PR 8555, allow changing the scale until the document is loaded, that hack is no longer necessary. Furthermore, no part of that event handling function needs to run unless a document is loaded.
The reason that this hack was initially added, is that previously the `ViewHistory` might be updated *before* `PDFViewerApplication.setInitialView` had run (in some cases leading to incorrect inital document scale). Since that is no longer possible, this is now dead code.
Don't allow setting various properties, such as `currentPageNumber`/`currentScale`/`currentScaleValue`/`pagesRotation`, before `{PDFViewer, PDFThumbnailViewer}.setDocument` has been called
This is a trivial follow-up to PR 5383, and it's a bit strange that this has been wrong since late 2014 without anyone noticing (maybe because inline images aren't too common).
So, apparently code works better if you actually spell correctly, who knew ;-)
Fixes 8613.
After PR 8394, where the l10n service was converted to be asynchronous, we're no longer calling `_adjustWidth` after updating the `findMsg` label. Hence it's currently possible that the width of the findbar won't be correct. The solution is simple though, just call `_adjustWidth` after the `findMsg` label has been (asynchronously) updated.
Another existing issue, which was an oversight in PR 8132, is that `PDFFindBar.updateResultsCount` may be called directly from `PDFFindController`. In that case, we're not calling `_adjustWidth` at all, which means that the findbar may also not have the correct width.
The simple solution here is to always call `_adjustWidth` at the end of `updateResultsCount` (which is why we no longer need the `_adjustWidth` call at the end of `updateUIState`).
Currently we're *only* hiding the label, but not actually resetting it until a new match is found.
Obviously it's being hidden, but it seems that it really ought to be completely reset as well (since e.g. `PDFFindBar.reset` won't technically reset *all* state otherwise).
Note that these files were among the first to be converted to ES6 classes, so it probably makes sense to do another pass to bring them inline with the most recent ES6 conversions.
All other code-paths already checks that the `MessageHandler` isn't terminated, but apparently `onFailure` was missing that check (compare e.g. with the `onSuccess` function).
From what I can tell, this is only an issue if workers are *disabled*, hence why I didn't bother adding a unit-test.
Fixes 8584.
These changes consists mainly of replacing `var` with `let`/`const`, adding a couple of default parameters to function signatures, and finally converting `EventBus`/`ProgressBar` to proper classes.
After PR 8510, we now always lookup the localized `page_scale_percent` string to prevent any possible ordering issues. Since the scaleSelect dropdown is updated asynchronous, there's really no point in having a helper function any more, hence this code can rather be placed inline in `Toolbar._updateUIState`.
In general, we may not know the stroke properties when path construction
happens. Since we must know the properties when we apply the stroke, we
should set the properties at that point. Note that we already do that
for the color and opacity, but not yet for the other properties.
The https://github.com/mozilla/pdf.js/wiki/Contributing article has been updated to explicitly mention that an ES6 browser is now a minimum requirement for development.
Since we recently have seen a couple of issues filed which seemed to indicate that people tried to use PDF.js in browsers without full ES6 support, it's probably a good idea to mention this more prominently in the README as well.
*This is an existing issue that I noticed while testing PR 8552.*
When zooming or rotation occurs, we'll try to use the current canvas as a (CSS transformed) preview until the page has been completely re-drawn.
If you manage to change the scale (or rotation) *very* quickly, it's possible that `PDFPageView.update` can be called *before* a previous `render` operation has progressed far enough to remove the `hidden` property from the canvas.
The result is thus that a page may be *entirely* black during zooming or rotation, which doesn't look very good. This effect can be a bit difficult to spot, but it does manifest even in the default viewer.
In the PDF from issue 8527, the clip operator (W) shows up before a path
is defined. The current SVG backend however expects a path to exist
before generating a `<svg:clipPath>` element.
In the example, the path was defined after the clip, followed by a
endPath operator (n).
So this commit fixes the bug by moving the path generation logic from
clip to endPath.
Our canvas backend appears to use similar logic:
`CanvasGraphics_endPath` calls `consumePath`, which in turn draws the
clip and resets the `pendingClip` state. The canvas backend calls
`consumePath` from multiple other places, so we probably need to check
whether doing so is also necessary for the SVG backend.
I scanned our corpus of PDF files in test/pdfs, and found that in every
instance (except for one), the "W" PDF operator (clip) is immediately
followed by "n" (endPath). The new test from this commit (clippath.pdf)
starts with "W", followed by a path definition and then "n".
# Commands used to find some of the clipping commands:
grep -ra '^W$' -C7 | less -S
grep -ra '^W ' -C7 | less -S
grep -ra ' W$' -C7 | less -S
test/pdfs/issue6413.pdf is the only file where "W" (a tline 55) is not
followed by "n". In fact, the "W" is the last operation of a series of
XObject painting operations, and removing it does not have any effect
on the rendered PDF (confirmed by looking at the output of PDF.js's
canvas backend, and ImageMagick's convert command).
Part of the rotation handling code, in what's now `web/app.js`, hasn't really changed since before the viewer was split into multiple files/components.
Similar to other properties, such as current page/scale, we should probably avoid tracking state in multiple places. Hence I'm suggesting that we don't store the rotation in `PDFViewerApplication`, and access the value in `PDFViewer` instead.
Since `PDFViewerApplication.pageRotation` has existed for a very long time, a getter was added to avoid outright breaking third-party code that may depend on it.
Currently a number of these properties do not work correctly if set *before* calling `setDocument`; please refer to the discussion starting in https://github.com/mozilla/pdf.js/pull/8539#issuecomment-309706629.
Rather than trying to have *some* of these methods working, but not others, it seems much more consistent to simply always require that `setDocument` has been called.
Wait for the completion of writing the generated SVG file before
processing the next page. This is to enable the garbage collector to
garbage-collect the (potentially large) SVG string before trying to
allocate memory again for the next page.
Note that since the PDF-to-SVG conversion is now sequential instead of
parallel, the time to generate all pages increases.
Test case:
node --max_old_space_size=200 examples/node/pdf2svg.js /tmp/FatalProcessOutOfMemory.pdf
Before this patch:
- Node.js crashes due to OOM after processing 20 pages.
After this patch:
- Node.js is able to convert all 203 PDFs to SVG without crashing.
Test case:
Using the PDF file from https://github.com/mozilla/pdf.js/issues/8534
node --max_old_space_size=200 examples/node/pdf2svg.js /tmp/FatalProcessOutOfMemory.pdf
Before this patch:
Node.js crashes due to OOM after processing 10 pages.
After this patch:
Node.js crashes due to OOM after processing 19 pages.
Since this call occurs *before* the `PDFViewer.setDocument` call, it won't actually cause any scale change.
Furthermore, moving it should not be necessary, since the `scale` is already used as the fallback case in `PDFViewerApplication.setInitialView` (provided it's non-zero, which isn't even the case in the default viewer).
Hence this patch should cause no functional changes at all, since it simply removes a piece of unnecessary code.
This patch adds Streams API support in getTextContent
so that we can stream data in chunks instead of fetching
whole data from worker thread to main thread. This patch
supports Streams API without changing the core functionality
of getTextContent.
Enqueue textContent directly at getTextContent in partialEvaluator.
Adds desiredSize and ready property in streamSink.
Currently, these properties are reset in what appears to be somewhat arbitrary locations (within the `load` and `open` methods respectively). The explanation is probably that both of these properties predates the existence of any centralized clean-up code in the viewer.
Hence I think that it makes sense to move the resetting of these properties to the `close` method, since that improves the overview of what's actually cleaned-up/reset when changing documents in the viewer.
The `ObjectLoader` currently takes an Object as input, despite actually working with `Dict`s internally. This means that at the (two) existing call-sites, we're passing in the "private" `Dict.map` property directly.
Doing this seems like an anti-pattern, and we could (and even should) simply provide the actual `Dict` when creating an `ObjectLoader` instance.
Accessing properties stored in the `Dict` is now done using the intended methods instead, in particular `getRaw` which (as the name suggests) doesn't do any de-referencing, thus maintaining the current functionality of the code.
The only functional change in this patch is that `ObjectLoader.load` will now ignore empty nodes, such that `ObjectLoader._walk` only needs to deal with nodes that are known to contain data. (This lets us skip, among other checks, meaningless `addChildren` function calls.)
This method is currently called from `PDFViewer._scrollUpdate` on *every* scroll event in the viewer.
However, I cannot see why this code is now necessary (assuming that it once was), since text-selection and searching still works *exactly* the same way with this patch as with the current `master`.
When `PDFPageView.updatePosition` is called, the page can be in either of these states:
1. The page hasn't been rendered, in which case the `textLayer` property doesn't exist yet.
2. The page is currently rendering, meaning that the `textLayer` property exists. Given that the `textContent` won't be fetched until the page has been successfully rendered, `TextLayerBuilder.render` will return immediately and effectively be a no-op (since there's nothing to render yet).
3. The has been been rendered, and the `textLayer` is currently rendering.
4. The page, and its `textLayer`, has been completely rendered. In this case, `TextLayerBuilder.render` will return immediately and effectively be a no-op.
Here, only the *third* case seem to require any further analysis:
When scrolling occurs while the `textLayer` is rendering, `TextLayerBuilder.render` will via a helper method call `TextLayerRenderTask.cancel` (in src/display/text_layer.js) to stop processing.
However, due to the run-to-completion nature of JavaScript, once `TextLayerRenderTask._render` has been invoked `appendText` will always run.[1]
So even though we cancel rendering of pending `textLayer`s during scrolling, via the repeated `TextLayerBuilder.render` calls from within the `PDFPageView.updatePosition` method, that does *not* prevent us from running the code inside of `TextLayerRenderTask._render` over and over for the *same* page; which all seems *very* inefficient to me.[2]
All this will thus have the effect of delaying the *actual* rendering of a `textLayer` ever so slightly while scrolling in the viewer. However, it does so at the expense of potentially hundreds of unnecessary `appendText` calls.[3]
Hence it seems to me that it's less resource intensive overall to simply let rendering of the `textLayer` complete, once it has started. Obviously, we still abort all rendering of a page, and its `textLayer`, when it's being destroyed (e.g. by being evicted from the page cache).
In case that there's any worry that the patch could affect e.g. highlighting of search results, please note that the existing code in `TextLayerBuilder.render` already calls `updateMatches` when the `TextLayerTask` resolves successfully.
*I'm sorry that this became quite long, but to try and summarize:*
`PDFPageView.updatePosition` doesn't actually do anything in *most* cases. In the one case where it matters, it seems that it's actually doing more harm than good; which is why I'm proposing that we just remove it.
---
[1] Although we may be able to skip the `render` call, provided that it happens *after* a `timeout` (as is the case in the default viewer).
[2] With current work being done to support streaming of `TextContent`, we'd also need to add just as many duplicate API calls to `PDFPageView.updatePosition`.
[3] The number of duplicate `appendText` calls is directly proportional not only to the scroll speed, but also to the number of pages in the document.
Since the localization service is now asynchronous, depending on the load the browser is under, there's a small risk that the lookup of the 'page_scale_percent' string could be delayed slightly.
If the scale would change a couple of times in *very* quick succession, there's perhaps a *theoretical* possibility that the Zoom dropdown would display an incorrect value.
Consider the following, somewhat contrived, theoretical example of two zoom commands being executed *right* after one another:
```javascript
PDFViewerApplication.pdfViewer.currentScale = 1.23;
PDFViewerApplication.pdfViewer.currentScaleValue = 'page-width';
```
Only the `currentScale` call will currently trigger a l10n lookup in `selectScaleOption`. However, as far as I understand, there's no *guarantee* that the l10n string is resolved *before* `selectScaleOption` is called again as a result of the `currentScaleValue` call.
This thus has the possibility of putting the Zoom dropdown into an inconsistent state, since it's currently updated synchronously for one code-path and asynchronously for another.
To avoid these issues, I'm proposing that we *always* update the Zoom dropdown asynchronously, such that we can guarantee that the ordering is correct.
After PR 8394, at least in Firefox, the Zoom dropdown now frequently displays an old custom scale instead of the correct one. To see this behaviour, the following STR works for me:
1. Open https://mozilla.github.io/pdf.js/web/viewer.html.
2. Zoom in, by clicking on the corresponding button in the toolbar.
3. Run `PDFViewerApplication.pdfViewer.currentScaleValue` in the console.
4. Compare what's displayed in the Zoom dropdown with what's printed in the console. (If no difference can be observed, try repeating steps 2 through 4 a couple of times.)
I really don't understand why this happens, but it seems that waiting until a custom scale has been set *before* selecting it fixes things in Firefox (and works fine in e.g. Chrome as well).
Note that this patch thus makes this particular piece of the code consistent with the state prior to PR 8394.
*As mentioned the last time that I touched this particular part of the font code, I'm sincerely hope that this doesn't cause any regressions!*
However, the patch passes all tests added in PRs 5770, 6270, and 7904 (and obviously all other tests as well). Furthermore, I've manually checked all the issues/bugs referenced in those PRs without finding any issues.
Fixes 8480.
Adds functionality to accept Queueing Strategy in
sendWithStream method. Using Queueing Strategy we
can control the data that is enqueued into the sink,
and hence regulated the flow of chunks from worker
to main thread.
Adds capability in pull and cancel methods.
Adds ready and desiredSize property in streamSink.
Adds unit test for ReadableStream and sendWithStream.
The click handler used in Presentation Mode didn't check if the first/last page was already reached, which after PR 7529 now causes an unnecessary console error.
Hence we should simply use the already existing `_goToPreviousPage`/`_goToNextPage` methods instead, since they do the necessary bounds checking.
Fixes 8498.
Moreover, rename `FindStates` to `FindState` since enumeration names are
usually not in plural, for readability and consistency with the ones in
`src/shared/util.js`.
http://eslint.org/docs/rules/comma-danglehttp://eslint.org/docs/rules/object-curly-spacing
*Please note:* This patch was created automatically, using the ESLint `--fix` command line option. In a couple of places this caused lines to become too long, and I've fixed those manually; please refer to the interdiff below for the only hand-edits in this patch.
```diff
diff --git a/gulpfile.js b/gulpfile.js
index d18b9c58..7d47fd8d 100644
--- a/gulpfile.js
+++ b/gulpfile.js
@@ -1247,7 +1247,8 @@ gulp.task('gh-pages-git', ['gh-pages-prepare', 'wintersmith'], function () {
var reason = process.env['PDFJS_UPDATE_REASON'];
safeSpawnSync('git', ['init'], { cwd: GH_PAGES_DIR, });
- safeSpawnSync('git', ['remote', 'add', 'origin', REPO], { cwd: GH_PAGES_DIR, });
+ safeSpawnSync('git', ['remote', 'add', 'origin', REPO],
+ { cwd: GH_PAGES_DIR, });
safeSpawnSync('git', ['add', '-A'], { cwd: GH_PAGES_DIR, });
safeSpawnSync('git', [
'commit', '-am', 'gh-pages site created via gulpfile.js script',
```
Fix inconsistent spacing and trailing commas in objects in `src/core/` files, so we can enable the `comma-dangle` and `object-curly-spacing` ESLint rules later on
Fix inconsistent spacing and trailing commas in objects in remaining `src/` files, so we can enable the `comma-dangle` and `object-curly-spacing` ESLint rules later on
Fix inconsistent spacing and trailing commas in objects in `test/` files, so we can enable the `comma-dangle` and `object-curly-spacing` ESLint rules later on
http://eslint.org/docs/rules/comma-danglehttp://eslint.org/docs/rules/object-curly-spacing
Given that we currently have quite inconsistent object formatting, fixing this in *one* big patch probably wouldn't be feasible (since I cannot imagine anyone wanting to review that); hence I've opted to try and do this piecewise instead.
Please note: This patch was created automatically, using the ESLint `--fix` command line option. In a couple of places this caused lines to become too long, and I've fixed those manually; please refer to the interdiff below for the only hand-edits in this patch.
```diff
diff --git a/src/display/canvas.js b/src/display/canvas.js
index 5739f6f2..4216b2d2 100644
--- a/src/display/canvas.js
+++ b/src/display/canvas.js
@@ -2071,7 +2071,7 @@ var CanvasGraphics = (function CanvasGraphicsClosure() {
var map = [];
for (var i = 0, ii = positions.length; i < ii; i += 2) {
map.push({ transform: [scaleX, 0, 0, scaleY, positions[i],
- positions[i + 1]], x: 0, y: 0, w: width, h: height, });
+ positions[i + 1]], x: 0, y: 0, w: width, h: height, });
}
this.paintInlineImageXObjectGroup(imgData, map);
},
diff --git a/src/display/svg.js b/src/display/svg.js
index 9eb05dfa..2aa21482 100644
--- a/src/display/svg.js
+++ b/src/display/svg.js
@@ -458,7 +458,11 @@ SVGGraphics = (function SVGGraphicsClosure() {
for (var x = 0; x < fnArrayLen; x++) {
var fnId = fnArray[x];
- opList.push({ 'fnId': fnId, 'fn': REVOPS[fnId], 'args': argsArray[x], });
+ opList.push({
+ 'fnId': fnId,
+ 'fn': REVOPS[fnId],
+ 'args': argsArray[x],
+ });
}
return opListToTree(opList);
},
```
Fix inconsistent spacing and trailing commas in objects in `web/` files, so we can enable the `comma-dangle` and `object-curly-spacing` ESLint rules later on
Fix inconsistent spacing and trailing commas in objects in `extensions/` files, so we can enable the `comma-dangle` and `object-curly-spacing` ESLint rules later on
http://eslint.org/docs/rules/comma-danglehttp://eslint.org/docs/rules/object-curly-spacing
Given that we currently have quite inconsistent object formatting, fixing this in in one big patch probably wouldn't be feasible (since I cannot imagine anyone wanting to review that); hence I've opted to try and do this piecewise instead.
Please note: This patch was created automatically, using the ESLint `--fix` command line option. In a couple of places this caused lines to become too long, and I've fixed those manually; please refer to the interdiff below for the only hand-edits in this patch.
```diff
diff --git a/extensions/firefox/content/PdfStreamConverter.jsm b/extensions/firefox/content/PdfStreamConverter.jsm
index ea91a71a..0d59dad1 100644
--- a/extensions/firefox/content/PdfStreamConverter.jsm
+++ b/extensions/firefox/content/PdfStreamConverter.jsm
@@ -773,7 +773,8 @@ class RequestListener {
response = function sendResponse(aResponse) {
try {
var listener = doc.createEvent("CustomEvent");
- let detail = Cu.cloneInto({ response: aResponse, }, doc.defaultView);
+ let detail = Cu.cloneInto({ response: aResponse, },
+ doc.defaultView);
listener.initCustomEvent("pdf.js.response", true, false, detail);
return message.dispatchEvent(listener);
} catch (e) {
```
After PR 8459, the run-time of the various `gulp` test commands has regressed quite badly on Windows. For me, `gulp test` now takes approximately *twice* as long when run locally on Windows.
The problem seems to be the Babel transpilation step, which takes well over five minutes to run.[1]
For someone like me, who runs tests a lot locally, this slowdown is really hurting the overall development experience.
To get around this I tested setting `PDFJS_NEXT = true` in `gulpfile.js`, since the transpilation step isn't necessary when testing in a modern browser.
However, having to edit `gulpfile.js` every time that I need to run tests isn't very practical. Hence this patch, which adds an environment variable that allows you to disable the transpilation simply by using e.g. `PDFJS_NEXT=true gulp test`.
I hope that this can be considered an acceptable solution, such that I don't need to maintain this patch locally (or worse, edit `gulpfile.js` locally before testing).
---
[1] This can also be observed on the Windows bot, but it seems fine on Linux.
This patch intends to simplify future ES6 refactoring of the thumbnail code, since the current code isn't going to work well together with proper `Class`es.
Please note that the `glyphlist.js` and `unicode.js` files are converted to CommonJS modules instead, since Babel cannot handle files that large and they are thus excluded from transpilation.
Add different code-paths to `{CMap, ToUnicodeMap}.charCodeOf` depending on length, since `Array.prototype.indexOf` can be extremely inefficient for very large arrays (issue 8372)
With the current way that the `HandTool` is implemented, if someone would try to also add a Zoom tool (as issue 1260 asks for) that probably wouldn't work very well given that you'd then have two cursor tools which may not play nice together.
Hence this patch, which attempts to refactor things so that it should be simpler to add e.g. a Zoom tool as well (given that that issue is marked as "good-beginner-bug", and I'm not sure if that really applies considering the current state of the code).
Note that I personally have no interest in implementing a Zoom tool (similar to Adobe Reader) since I wouldn't use it, but I figured that it can't hurt to make this code a bit more future proof.
PR 7341 added special handling for `nameddest`s that look like pageNumbers, to prevent issues since we previously *incorrectly* supported specifying a pageNumber directly in the hash; i.e. `#10` versus the correct `#page=10` format.
Since this behaviour wasn't correct, PR 7757 fixed and deprecated the old format, which means that we no longer need to maintain the `nameddest` hack in multiple files.
Also updates a `console.log` statement in the `gulp importl10n` command (since I forgot it in a previous patch, and it didn't seem necessary with a separate patch for it).
Added test for ReadableStream.
Adds ref-implementation license-header in streams-lib
and change gulp task to copy external/streams/ in build/
external/streams/ and build/dist/external/streams folder.
Adds README.md and LICENSE.md
This patch replaces a `var self = this;` statement with arrow functions, and `var` with `let` in `PDFLinkService.navigateTo`.
Furthermore, when I started looking at this method, it quickly became clear that its code is somewhat of a mess. Since I'm one of the persons that have touched this code over the years, I figured that it'd be a good idea to try and clean it up a bit.
Also replaces `var` with `let` in code that's touched in the patch. Please note that this should be completely safe, for two separate reasons, since trying to access let in a scope where it's not defined is first of all a runtime error and second of all an ESLint error (thanks to the `no-undef` rule).
Refactors `Driver._cleanup` to return a `Promise` which is resolved once all opened documents have been destroyed.
This is then used in `Driver._nextTask` to ensure that we wait for everything to be cleaned up, such that the tests run sequentially.
Currently this method first uses a loop to build a temporary array to hold Promises, which are then resolved from a recursive helper function once the textContent is fetched for each page.
To me, this is unncessarily complicated, since we can do everything within one loop by simply chaining the asynchronous calls to retrieve the textContent. (Note that this guarantees that the textContent of the pages is still fetched sequentially.)
Also replaces `var` with `let` in the functions/methods that are touched in the patch. Please note that this should be completely safe, for two separate reasons, since trying to access `let` in a scope where it's not defined is first of all a runtime error and second of all an ESLint error (thanks to the `no-undef` rule).
This is a downstream change introduced in [1]. That mozilla bug is adding a new property to channel's loadinfo object (nsILoadInfo) that protocol handlers has to set on channels when originalURI on the result channel is set to a different URI than the channel has been created for.
Existence of the new property on nsILoadInfo depends on landing [1].
[1] https://bugzilla.mozilla.org/show_bug.cgi?id=1319111
- the viewer was not loading in development mode on Safari, due to Safari
having crypto.webkitSubtle instead of crypto.subtle.
- the isCachingPossible check was amended to check for crypro.subtle
which is currently not in Safari but in Firefox and Chrome. This
essentially works around the issue by disabling caching for Safari
in development mode.
- maintainer sentiment: people who develop on Safari can get this speedup
once Safari drops prefix for SubtleCrypto.
- note: at time of writing Safari Version 10.1 (12603.1.30.0.34) has an
issue where caching can be enabled for PDF.js but must to be disabled for
worker, otherwise the two sides do not communicate.
- https://github.com/mozilla/pdf.js/pull/8387#issuecomment-299709961
- https://developer.mozilla.org/en-US/docs/Web/API/SubtleCrypto/digest
For some reason, we're putting all kind of images *except* JPEG into the `imageCache` in `evaluator.js`.[1]
This means that in the PDF file in issue 8380, we'll keep sending the *same* two small images[2] to the main-thread and decoding them over and over. This is obviously hugely inefficient!
As can be seen from the discussion in the issue, the performance becomes *extremely* bad if the user has the addon "Adblock Plus" installed. However, even in a clean Firefox profile, the performance isn't that great.
This patch not only addresses the performance implications of the "Adblock Plus" addon together with that particular PDF file, but it *also* improves the rendering times considerably for *all* users.
Locally, with a clean profile, the rendering times are reduced from `~2000 ms` to `~500 ms` for my setup!
Obviously, the general structure of the PDF file and its operator sequence is still hugely inefficient, however I'd say that the performance with this patch is good enough to consider the issue (as it stands) resolved.[3]
Fixes 8380.
---
[1] Not technically true, since inline images are cached from `parser.js`, but whatever :-)
[2] The two JPEG images have dimensions 1x2, respectively 4x2.
[3] To make this even more efficient, a new state would have to be added to the `QueueOptimizer`. Given that PDF files this stupid fortunately aren't too common, I'm not convinced that it's worth doing.
This patch contains the following improvements:
- Only fetch the various document properties *once* per PDF file opened, and cache the result (in a frozen object).
- Always update the *entire* dialog at once, to prevent inconsistent UI state (issue 8371).
- Ensure that the dialog, and all its internal properties, are reset when `PDFViewerApplication.close` is called.
- Inline, and re-factor, the `getProperties` method in `open`, since that's the only call-site.
- Always overwrite the fileSize with the value obtained from `pdfDocument.getDownloadInfo`, to ensure that it's correct.
- ES6-ify the code that's touched in this patch.
Fixes 8371.
Currently these methods accept a large number of parameters, which creates quite unwieldy call-sites. When invoking them, you have to remember not only what arguments to supply, but also the correct order, to avoid runtime errors.
Furthermore, since some of the parameters are optional, you also have to remember to pass e.g. `null` or `undefined` for those ones.
Also, adding new parameters to these methods (which happens occasionally), often becomes unnecessarily tedious (based on personal experience).
Please note that I do *not* think that we need/should convert *every* single method in `evaluator.js` (or elsewhere in `/core` files) to take parameter objects. However, in my opinion, once a method starts relying on approximately five parameter (or even more), passing them in individually becomes quite cumbersome.
With these changes, I obviously needed to update the `evaluator_spec.js` unit-tests. The main change there, except the new method signatures[1], is that it's now re-using *one* `PartialEvalutor` instance, since I couldn't see any compelling reason for creating a new one in every single test.
*Note:* If this patch is accepted, my intention is to (time permitting) see if it makes sense to convert additional methods in `evaluator.js` (and other `/core` files) in a similar fashion, but I figured that it'd be a good idea to limit the initial scope somewhat.
---
[1] A fun fact here, note how the `PartialEvaluator` signature used in `evaluator_spec.js` wasn't even correct in the current `master`.
Note that we have to use `fs.writeFileSync` since `.to()` is not
available anymore. Moreover, introduce `safeSpawnSync` to make sure that
we check the return codes of the spawned processes properly.
To run the regression tests, developers use `gulp browsertest` and the
bot uses `gulp bottest`. We're not passing the `noreftest` option
anywhere in the code (probably because the `bottest` command takes care
of this already), so we should remove this.
Note that by using `let` instead of `var` in `PartialEvaluator.setGState` and `TranslatedFont.loadType3Data`, we can get rid of further `bind` usages since `let` is block-scoped.
Also, the fact that `bind` wasn't used in the `Font` case inside of `setGState` is actually a bug which has been present ever since PR 5205, where a closure was replaced by a standard loop.[1]
---
[1] I'm not aware of any bugs caused by this, but that is probably more a happy accident than anything else, since e.g. just removing the `bind` from the `SMask` case without using block-scoped variables causes test failures.
Please note that I used the addon debugger to set a breakpoint in the `unload` function, in order to ensure that `this` still correctly refers to the `FindEventManager` scope.
Please see http://eslint.org/docs/rules/object-shorthand.
Unfortunately, based on commit https://github.com/mozilla/pdf.js/commit/9276d1dcd97b4b0dc45342608a7f1b1a7f1e9c5b, it seems that we still need to maintain compatibility with old Node.js versions, hence certain files/directories that are executed in Node.js are currently exempt from this rule.
Furthermore, since the files specific to the Chromium extension are not run through Babel, the `/extensions/chromium/` directory is also exempt from this rule.
Please see http://eslint.org/docs/rules/object-shorthand.
For the most part, these changes are of the search-and-replace kind, and the previously enabled `no-undef` rule should complement the tests in helping ensure that no stupid errors crept into to the patch.
This is a regression from commit https://github.com/mozilla/pdf.js/commit/3888a993b125e22fdfffd36bf603baf830e6fbe2.
It turns out the even though we have a `URL` polyfill, it's still dependent on the existence of native `URL.{createObjectURL, revokeObjectURL}` functions.
Since no such thing exists in Node.js, our `createObjectURL` utility function breaks there.
The test runner is automated, so if the default browser test is
performed, the browser hangs waiting for user input it never gets.
Disable the test to fix that.
Moreover, enable E10s now that it is mature. This may help with the
performance of the test runner as well.
Please see http://eslint.org/docs/rules/object-shorthand.
For the most part, these changes are of the search-and-replace kind, and the previously enabled `no-undef` rule should complement the tests in helping ensure that no stupid errors crept into to the patch.
Rather than having to manually use `initializedPromise`, which really ought to be a private property, to ensure that setting/getting values in the `ViewHistory` works as intended, this re-factoring simply changes all of its methods to be asynchronous.
Furthermore, a `getMultiple` method (mirroring the existing `setMultiple` one) is also added to `ViewHistory`.
Finally, this patch also addresses an existing issue, where certain preferences (e.g. the default zoom level) would be ignored when calling `setInitialView` if reading from the `ViewHistory` fails for some reason.
Please refer to the JBIG2 standard, see https://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-REC-T.88-200002-I!!PDF-E&type=items.
In particular, section "6.3.5.3 Fixed templates and adaptive templates" mentions that the offsets should be *subtracted*; where the offsets are defined according to "Table 6" under section "6.3.2 Input parameters".
Fixes 7145.
Fixes 7308.
Fixes 7401.
Fixes 7850.
Fixes 8270.
Please see http://eslint.org/docs/rules/object-shorthand.
For the most part, these changes are of the search-and-replace kind, and the previously enabled `no-undef` rule should complement the tests in helping ensure that no stupid errors crept into to the patch.
Please see http://eslint.org/docs/rules/object-shorthand.
For the most part, these changes are of the search-and-replace kind, and the previously enabled `no-undef` rule should complement the tests in helping ensure that no stupid errors crept into to the patch.
With the exception of just one test-case, all the current `ui_utils` unit-tests can run successfully on Node.js (since most of them doesn't rely on the DOM).
To get this working, I had to first of all add a new `LIB` build flag such that `gulp lib` produces a `web/pdfjs.js` file that is able to load `pdf.js` successfully.
Second of all, since neither `document` nor `navigator` is available in Node.js, `web/ui_utils.js` was adjusted slightly to avoid errors.
[Firefox addon] Remove the `window.FirefoxCom` hack from `web/viewer.js`, since it was made redunant by the `setExternalLocalizerServices` refactoring (PR 7202 follow-up)
This prevents issues with the filename detection being skipped, when trying to download the opened PDF attachment, since `getPDFFileNameFromURL` ignores `data:` URLs for performance reasons.
The patch also changes the `defaultFilename` to use the ES6 default parameter notation, and fixes the formatting of the JSDoc comment.
Finally, since `getPDFFileNameFromURL` currently has no unit-tests, a few basic ones are added to avoid regressions.
[Firefox addon] Upstream changes from: Bug 1346616 - Migrate callsites that are retrieving requested locale from pref, to use LocaleService::GetRequestedLocales
In https://bugzilla.mozilla.org/show_bug.cgi?id=1355216, the *third* parameter of `Services.obs.addObserver` was made optional.
However, omitting it in Firefox versions *without* that patch causes failures that completely prevents the addon from working (it won't even load).
As far as I can tell, there isn't *any* way to detect ahead of time if the third parameter can be safely omitted, hence we're forced to fallback to manually checking the version number :-(
*Note:* Since the `PdfJs.jsm` file is only used in the `MOZCENTRAL` build, we at least don't need to add any compatibility hacks there.
In the first commit in PR 8203, I changed how the `DownloadManager` was included/initialized in `GENERIC`/`CHROME` builds.
The change was prompted by the fact that you cannot have conditional `import`s with ES6 modules, and I wanted to avoid bundling the general `DownloadManager` into the various Firefox specific build targets.
What I completely missed though, is that the new code meant that `download_manager.js` will now be pulling in the *entire* viewer (through `app.js`).
This is a *really* stupid mistake on my part, since it causes the `dist/build/pdf_viewer.js` used with the viewer components to now include basically the entire default viewer.
The simplest solution that I could come up with, is to add a `genericcom.js` file (similar to the `firefoxcom.js`/`chromecom.js` files) which will be responsible for importing/initializing the `DownloadManager`.
The browsers have become smarter and made this hack no longer
functional. Since this is now enforced by practically all browsers,
there is nothing more we can do about this. It is up to the user to
serve the viewer over HTTPS or deal with the warning.
Note that this is in no way specific for PDF.js. Any site with password
inputs served over HTTP has this problem right now. This hack was
provided as a convenience for the users, but nothing more than that.
Since calling `PDFPageView.setPdfPage` will in turn call `PDFPageView.reset`, which cancels all rendering and completely resets the page, it's thus possible that we currently cause some unnecessary re-rendering during the initial loading phase of the viewer.
Depending on the order in which data arrives, it's possible (and in practice always seem to happen) that the `pdfPage` property of the *second* page has already been set during `PDFViewer.setDocument`, by the time that the request for the `pdfPage` is resolved in `PDFViewer._ensurePdfPageLoaded`.
Also, note how the `setPdfPage` call in `PDFViewer.setDocument` is already guarded by this kind of check.
In various viewer files, there's a number of cases where we basically duplicate the functionality of `createPromiseCapability` manually.
As far as I can tell, a couple of these cases have existed for a very long time, and notable even before the `createPromiseCapability` utility function existed.
Also, since we can write ES6 code now, the patch also replaces a couple of `bind` usages with arrow functions in code that's touched in the patch.
[api-minor] Always allow e.g. rendering to continue even if there are errors, and add a `stopAtErrors` parameter to `getDocument` to opt-out of this behaviour (issue 6342, issue 3795, bug 1130815)
This patch implements support for line annotations. Other viewers only
show the popup annotation when hovering over the line, which may have
any orientation. To make this possible, we render an invisible line (SVG
element) over the line on the canvas that acts as the trigger for the
popup annotation. This invisible line has the same starting coordinates,
ending coordinates and width of the line on the canvas.
Other PDF readers, e.g. Adobe Reader and PDFium (in Chrome), will attempt to render as much of a page as possible even if there are errors present.
Currently we just bail as soon the first error is hit, which means that we'll usually not render anything in these cases and just display a blank page instead.
NOTE: This patch changes the default behaviour of the PDF.js API to always attempt to recover as much data as possible, even when encountering errors during e.g. `getOperatorList`/`getTextContent`, which thus improve our handling of corrupt PDF files and allow the default viewer to handle errors slightly more gracefully.
In the event that an API consumer wishes to use the old behaviour, where we stop parsing as soon as an error is encountered, the `stopAtErrors` parameter can be set at `getDocument`.
Fixes, inasmuch it's possible since the PDF files are corrupt, e.g. issue 6342, issue 3795, and [bug 1130815](https://bugzilla.mozilla.org/show_bug.cgi?id=1130815) (and probably others too).
- When the minified version is used the resolver of the worker can not find it properly and throws 404 error.
- The problem was that:
- It was getting the current name of the file.
- It was replacing **.js** by **.worker.js**
- When it was loading the unminified version it was working fine because:
- *pdf.js - .js + .worker.js* = **pdf.worker.js**
- When it was loading the minified version it didtn't work because:
- *pdf.min.js - .js + .worker.js* = **pdf.min.worker.js**
- **pdf.min.worker.js** doesn't exist the real file name is **pdf.worker.min.js**
The display layer is responsible for creating the HTML elements for the
annotations from the core layer. If we need to ignore border styling for
the containers of certain elements, the display layer should do so and
not the core layer. I noticed this during the implementation of line
annotations, for which we actually need the original border width in the
display layer, even though we ignore it for the container. If we set the
border style to zero in the core layer, this becomes impossible.
To prevent this, this patch moves the container border removal code from
the core layer to the display layer. This makes the core layer output
the unchanged annotation data and lets the display layer remove any
border styling if necessary.
Note that as discussed on IRC, this makes the viewer slightly slower to load *only* in `gulp server` mode, however the difference seem slight enough that I think it will be fine.
Considering how extremely simple this patch turned out to be, I'm almost worried that I completely misunderstood why the current code looks like it does...
It doesn't really make sense to attempt to utilize the `NativeImageDecoder` in Node, since there's no native image support available, hence building on PR 8035 we can easily disable it in the example.
Fixes 7901.
The existing implementation of fakeRequestAnimationFrame
did not return a timer ID, so the frame could not be cancelled
if you wanted to cancel it. But if you do want to cancel it,
it needs to be cancelled through clearTimeout instead of
cancelAnimationFrame, because the timer IDs are different.
Signed-off-by: Jonathan Barnes <jbarnes@pivotal.io>
If we only invoke the bootstrap-enabled script when PdfJs.enabled is
true, then we don't need to check it again in the script.
This avoids a sync IPC call to the parent process.
It also keeps PdfJs.jsm from importing PdfjsContentUtils.jsm in the
child process until it is actually needed, which is one steps towards
not loading it until it is really needed.
pdfjschildbootstrap.js will always be run, but
pdfjschildbootstrap-enabled.js will only be run if PdfJs.enabled is
true. This will let us avoid some work in the child process in the
next patch.
This will need to be landed in the mozilla-central repository at the
same time as a change to nsBrowserGlue.js. See bug 1352218.
Refactor removing of the `zoomLayer` into a helper method, and use that in `PDFPageView.reset` to ensure that the entire `zoomLayer` is actually removed (issue 8209)
I happened to notice that the error handling wasn't that great, which I missed previously since there were no unit-tests for failure to load built-in CMap files.
Hence this patch, which improves the error handling *and* adds tests.
I found that PR 8105 unfortunately causes a *very serious* performance regression in long PDF documents where the `Pages` tree only has one level; my apologies for this!
Obviously we cannot revert that PR, since that would cause more issues than it solves. Hence it seems to me that the only viable solution here, is to add a simple `RefSetCache` to reduce the amount of redundant lookups.
Previously in PR 8105 caching was thought to be unnecessary, but as it turns out I don't think that we really have a choice in the matter any more.
See http://eslint.org/docs/rules/#ecmascript-6.
To try and enforce consistent rules and to help avoid some possible errors in ES6 code from the start, this patch adds a few basic ESLint rules.
Note that a two of the rules, `no-shadow` and `object-shorthand`, are currently disabled. While it'd certainly be nice to enable both of them, it's currently impossible since that would result in close to one thousand lint errors.
Remove unnecessary `xref` parameters from various method signatures in `PartialEvaluator`, since `this.xref` is already available in the relevant scope
For reasons I don't pretend to understand, we're passing around `xref` arguments to a bunch of methods despite `this.xref` being available in `PartialEvaluator`.
This patch is a small first small step towards cleaning up the, often unwieldy, signatures of methods in `PartialEvaluator`.
I really cannot understand why this change is necessary, since modern browsers such as Firefox and Chrome work just fine with the old code.
Hence this is patch is yet another "hack" that's needed just because IE apparently cannot just work like you'd expect.
For consistency, the Node factory used in the CMap unit-tests is changed as well.
Fixes 8193.
*My apologies for inadvertently breaking this in PR 8064; apparently we don't have any tests that cover this use-case :(*
Without this patch `getTextContent` will fail if called before `getOperatorList`, since loading of fonts during text-extraction may require fetching of built-in CMap files.
*Please note:* The `text` test added here, which uses an already existing PDF file, fails without this patch.
In core/document.js: `PDFDocument.prototype.parse` accesses a dictionary
property, which could throw if the underlying data is not yet available.
In core/obj.js: `get Catalog.prototype.metadata` calls
`stream.getBytes`, which can throw MissingDataException too when the
stream is a ChunkedStream.
Similar to other `try-catch` statements in `/core` code, we must re-throw `MissingDataException` to prevent issues with missing data during document loading.
Note that I'm not sure if/how we can test this, which is why the patch doesn't include any test(s).
Fixes 8180.
*After browsing through (a version of) the JPEG specification, see https://www.w3.org/Graphics/JPEG/itu-t81.pdf, I hope that this patch makes sense.*
Note that while issue 7828 became a problem after PR 7661, it isn't really a regression from than PR. The explanation is rather that we're now relying on `core/jpg.js` instead of the Native Image decoder in more situations than before, which thus exposed an *existing* issue in our JPEG decoder.
Another factor also seems to be that in many JPEG images, the DRI (Define Restart Interval) marker isn't present, in which case this bug won't manifest either.
According to https://www.w3.org/Graphics/JPEG/itu-t81.pdf#page=89 (at the bottom of the page):
"NOTE – The final restart interval may be smaller than the size specified by the DRI marker segment, as it includes only the number of MCUs remaining in the scan."
Furthermore, according to https://www.w3.org/Graphics/JPEG/itu-t81.pdf#page=39 (in the middle of the page):
"[...] If restart is enabled and the restart interval is defined to be Ri, each entropy-coded segment except the last one shall contain Ri MCUs. The last one shall contain whatever number of MCUs completes the scan."
Based on the above, it thus seem to me that we should simply ensure that we're not attempting to continue to parse Scan data once we've found all MCUs (Minimum Coded Unit) of the image.
Fixes 7828.
I happened to notice that some inequalities had the wrong order, and was surprised since I thought that the `yoda` rule should have caught that.
However, reading http://eslint.org/docs/rules/yoda#options a bit more closely than previously, it's quite obvious that the `onlyEquality` option does *exactly* what its name suggests. Hence I think that it makes sense to adjust the options such that only ranges are allowed instead.
There's still some work necessary if we want to be able to run (even a subset of) the API unit-tests on Travis.
However, this patch could be considered a small first step, since the relevant unit-tests will now rely on a `CanvasFactory` rather than using `document.createElement('canvas')` directly.
This patch gets rid of the only case in the code-base where we're throwing a plain `string`, rather than an `Error`, which besides better/more consistent error handling also allows us to enable the [`no-throw-literal`](http://eslint.org/docs/rules/no-throw-literal) ESLint rule.
The point of this patch is to fix a couple of inconsistencies in `viewer.html`, compared to the locale files, such that the viewer would work correctly even without the `l10n/` files present.
*Note:* Since this isn't changing any of the locale files, we should *not* need to update any of the string names.
Looking through the history of the findbar code, it seems that the `findPrevious`/`findNext` buttons have never had a `title` set (note PR 2168, which was the initial findbar implementation).
Furthermore, the `placeholder` of the `findInput` didn't agree 100% with locale file either, so this is also adjusted.
I noticed that we have a `.toolbarButton.group` CSS class, which is currently applied to some buttons in the viewer. Since it attempts to adjust the `margin-right` property, I was initially a bit puzzled as to why there wasn't different rules for LTR/RTL locales.
However, checking the viewer with the DevTools inspector, in both LTR and RTL locales, I quickly found that the rule in question is *always* being overridden by other CSS rules.
It thus seem to me that while this rule was probably useful at some point, it has been dead for years and could now be removed.
The find functionality is currently not available for small devices
because the find dialog is not responsive. This patch fixes that.
To achieve this goal, the HTML is changed to always show the find
button. To prevent issues because of the addition of an extra button for
small views, the previous/next page buttons are hidden if the view
becomes small. These buttons are not useful anyway because on small
devices navigation is usually done via scrolling. The find functionality
is much more useful to have in this case. Moreover, we wrap the existing
elements into separate `div`s so that the browser can position the
elements itself when the view becomes smaller and logically connected
elements stay together when this happens.
In the CSS, extra rules for the find bar have been added to ensure that
the dialog's doorhanger is always below the find button. All findbar
`div`s are forced to be 32 pixels high to prevent the find message text
being aligned under the checkboxes. Finally, the find message is only
visible when there is actually text to display. This prevents wrapping
issues because, by default, the label has padding and margin even if
there is no text.
Instead of just upstreaming the changes from [bug 1345253](https://bugzilla.mozilla.org/show_bug.cgi?id=1345253) as-is, it seemed better to simply get rid of the loops altogether and use the same approach as in `PDFViewer`/`PDFThumbnailViewer`.
This patch is another step towards enabling Babel. Since we're moving
towards ES6 modules, we will not be using UMD headers anymore, so we can
remove the validation.
*This is something that I noticed while working on PR 8126, which is (more) fallout from PR 6065.*
In general, it's actually *not* correct to return `Dict.empty` as the default value for non-existent properties. Please note that a prior PR, see https://github.com/mozilla/pdf.js/pull/5957#issuecomment-103112698, asked for that behaviour but I don't think that's right.
Obviously for properties that are (or should) be `Dict`s it makes sense, however certain properties can be e.g. Strings or Arrays instead. In the latter case, returning `Dict.empty` is just plain wrong, and it's quite fascinating that this hasn't caused any errors in practice. (The existing validation in the various getters has actually saved us here.)
Also, when looking at this code again, it seemed unnecessary to duplicate the `MAX_LOOP_COUNT` check since we could just return immediately instead.
When using content security headers to restrict connections to the same origin,
you may not make connections to `example.com`. This feature detection also
works with a request to the current location.
It appears that I accidentally broke this in PR 6065, sorry about that!
The issue in this particular PDF file is that there's `/Rotate` entries on different levels of the `/Pages` tree. We're supposed to use the `/Rotate` entry in the `/Page` dict (which is `0`), but because of an incorrect condition we instead ended up with the one from the `/Pages` dict (which is `180`).
Fixes 8125.
Given that the mailing list has now been closed, see [bug 1340296](https://bugzilla.mozilla.org/show_bug.cgi?id=1340296), and that there's no weekly meetings any more, we probably shouldn't mention either of those in CONTRIBUTING.md.
Instead, let's just suggest the IRC channel as a means of communication here.
Ideally, the `Annotation` class should not have anything to do with the
page's operator list. How annotations are added to the page's operator
list is logic that belongs in `src/core/document.js` instead where the
operator list is constructed.
Moreover, some comments have been added to clarify the intent of the
code.
Even though the PDF specification does not state that `Opt` fields are
inheritable, in practice there are PDF generators that let annotations
inherit the options from a parent.
This fixes something that I noticed while working with the code in `Catalog.getPageDict` when debugging issue 8088.
Note that while I don't have an example where this patch really matters, given that e.g. `PartialEvaluator.hasBlendModes` depends on the `objId` to avoid cyclic references this patch could potentially help for some PDF files.
As discussed on IRC, we need to check all nodes at the *bottom* of the tree to ensure that we find the correct `Page` dict.
Furthermore, this patch also gets rid of the caching present in a previous version, since it's not clear if that really helps.
Note that this patch purposely adds an `eq` test, using a reduced test-case, so that we can be sure that the algorithm actually finds the correct `Page` dict for each `pageIndex`.
Fixes 8088.
The download button in pdf.js doesn't work in iOS Chrome.
- It appears to be an issue with URLs from URL.createObjectURL.
The URL is correct, but iOS Chrome won't even load the URL
when `a.click()` is called in `download_manager.js`. Even
if you manually visit the URL, you get a blank page.
- Fix this by detecting iOS Chrome and disabling createObjectURL.
The `download_manager.js` `download` method wasn't checking
`PDFJS.disableCreateObjectURL`, so check it there, too.
- Move the navigator.msSaveBlob check earlier, so that
this doesn't change IE10 / IE11 behavior.
- Remove the !URL check since pdf.js has a URL polyfill
now.
[api-minor] Refactor fetching of built-in CMaps to utilize a factory on the `display` side instead, to allow users of the API to provide a custom CMap loading factory (e.g. for use with Node.js)
Currently the built-in CMap files are loaded in `src/core/cmap.js` using `XMLHttpRequest` directly. For some environments that might be a problem, hence this patch refactors that to instead use a factory to load built-in CMaps on the main thread and message the data to the worker thread.
This is inspired by other recent work, e.g. the addition of the `CanvasFactory`, and to a large extent on the IRC discussion starting at http://logs.glob.uno/?c=mozilla%23pdfjs&s=12+Oct+2016&e=12+Oct+2016#c53010.
Previously, we had a function called `getDefaultAppearance`. This name,
however, is misleading as the method gets the normal appearance (in the
`N` entry) and not the default appearance (in the `DA` entry). Moreover,
it was not entirely clear how it works just from reading the code. It
primarily lacks comments and explicit error case handling.
This patch improves the situation by fixing the issues mentioned above
and making this function a proper method of the `Annotation` class, just
like e.g., `setColor` and `setBorderStyle`.
This patch basically reverts one aspect of TrueType (3, 1) cmap parsing to the state prior to PR 4259. After that PR, a number of regressions occurred in this particular code-path, which necessitated a number of follow-ups such as PRs 5703, 5743, and 6425.
The empirical data suggests, at least to me, that we should always prefer a (3, 1) cmap for TrueType fonts when they have an encoding, regardless of the Symbolic font flag.
Obviously this patch passes all unit/font/reference tests locally, and I made sure that all the PRs mentioned above landed with test-cases included.
However, in my opinion, there's still a very real possibility that this patch could potentially cause new regressions.
Given that the PDF file in bug 1337429 has been broken for almost *three* years before anyone noticed, and considering that the code-path in question has been the source of numerous regressions, I do *not* intend to request uplift of this patch to previous Firefox versions (assuming that it's even accepted).
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1337429.
*Yet another thing that I unfortunately missed during review of PR 8023.*
Note that previously, in `make.js` this file was being preprocessed, however as far as I can tell that wasn't actually necessary. Hence this patch just copies the file to the proper output directory.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1338395#c8.
Consume the current character when encountering illegal characters in `Lexer.getObject`, in order to prevent infinite loops during reading of streams (issue 8061)
*Please note:* The rendering of the PDF file in issue 8061 first regressed in PR 7039, and then PR 7493 exacerbated the problem even further by causing an infinite loop.
In this particular case, when errors were encountered inside of the `Lexer.getObject` method *itself*, we didn't advance the stream position. This thus caused an inifinite loop in `parseCMap`, since the exact same character was then parsed over and over again.
Fixes 8061.
[Chromium addon] Prevent errors that break the addon, caused by the `DEFAULT_URL` constant being replaced by a `defaultUrl` viewer configuration parameter (PR 8046 follow-up)
If users want to download, they can quickly click on the Download button
in the newly opened viewer.
The blobUrl logic for Firefox relies on `disableCreateObjectURL` is
never false in Firefox. If the assumption is invalid, then PDF
attachments at the attachment view will not correctly be displayed,
because a data-URL will be generated and `?<filename>` is treated as
part of the data:-URL.
Determine the page rotation at the same place as where the page size is
determined. This allows us to implement custom print page rotation logic
in one place, in the future.
Note that I initially tried to add this as a parameter to the `PDFPageProxy.render` method, such that it could be passed to `PartialEvaluator.getOperatorList`.
However, given all the different code-paths that call `getOperatorList` (there's a bunch only in `annotation.js`), this seemed to very quickly become unwieldy and thus difficult to maintain compared to simply using the existing `evaluatorOptions`.
The download method (and the PDF document properties) detect the
file name using `getPDFFileNameFromURL`. The title ought to also
display the PDF filename if available.
This patch lets the AcroForms example make use of the built-in interactive
forms functionality in PDF.js. This makes the example:
- much easier to understand;
- more feature-complete;
- in sync with the core when new functionality is added;
- similar to the other examples in terms of structure.
Fixes extra canvas create calls.
Fixes unnecessary call of `new DOMCanvasFactory`.
Fixes undefined error of DOMCanvasFactory.
Fixes failures in some of the tests.
Fixes expected behaviour.
Remove unused vars.
The `Driver._cleanup` method is removing all stylesheets between test runs, which causes "TypeError: styleElement.parentNode is null" console errors in `FontLoader.clear`.
As can also be seen during various tests, some of the changes I made in PR 7972 unfortunately causes console errors.
It seems that I didn't test this properly, since it *should* have been obvious to me that while tests are triggered using Node.js, the files in question are run within the *browser*.
My apologies for not testing this thoroughly, and for causing unnecessary churn in the code!
The last (and only) usage of `MOZ_CENTRAL` was removed in PR 3036, so it's been unused for almost four years now.
If we need to have different code-paths for `FIREFOX`/`MOZCENTRAL` builds, the preprocessor should (and has) been used instead.
See http://eslint.org/docs/rules/brace-style.
Having the opening/closing braces on the same line can often make the code slightly more difficult to read, in particular for `if`/`else if` statements, compared to using new lines.
This patch also, for consistency with `mozilla-central`, enables the [`no-iterator`](http://eslint.org/docs/rules/no-iterator) rule. Note that this rule didn't require a single code change.
When a blob or data-URL is opened with the extension, viewer.html
rewrites the URL. But when the viewer is refreshed (e.g. F5), Chrome
would fail to display the viewer because the extension router was not
set up to recognize such URLs.
Now it is.
The regular expression incorrectly marked a group as capturing.
For `http://example.com/#file.pdf`, the expected result is "file.pdf",
but instead "document.pdf" was returned.
This property was added all the way back in PR 542, but hasn't actually been relied upon ever since PR 692.
Note that there's a `isStream()` utility function which replaced the property years ago, hence the `isStream` property is now dead code.
Other PDF viewers, e.g. Adobe Reader, seem to append `FileAttachment`s to their attachments views.
One obvious difference in PDF.js is that we cannot append all the annotations on document load, since that would require parsing *every* page. Despite that, it still seems like a good idea to add `FileAttachment`s, since it's thus possible to access all the various types of attachments from a single place.
*Note:* With the previous patch we display a notification when a `FileAttachment` is added to the sidebar, which thus makes appending the contents of these annotations to the sidebar slightly more visible/useful.
A longstanding issue with the viewer is that you cannot tell if a PDF document includes an outline and/or attachments without actually opening the sidebar.
This patch contains a suggested solution for that, by displaying an hide-on-interaction notification on the `sidebarToggle` button (and the relevant sidebar view buttons). Note that this was inspired by e.g. the update notification that is displayed on the menu button in Firefox.
For an initial implementation, I've tried to do this in such a way that the notification isn't too distracting. Without being an UX expert, I don't think that we'd want something too in-your-face, in order to keep the viewer toolbars reasonable clean. (We probably do *not* want e.g. an entire notification bar in these situations, since that would take up unnecessary screen space and require actions from the user to close.)
However it's certainly possible that the current notification might simply be *too* inconspicuous to be truly helpful to users, but we could probably iterate on that if the feature itself is deemed useful.
Please see http://eslint.org/docs/rules/no-unused-vars; note that this patch purposely uses the same rule options as in `mozilla-central`, such that it fixes part of issue 7957.
It wasn't, in my opinion, entirely straightforward to enable this rule compared to the already existing rules. In many cases a `var descriptiveName = ...` format was used (more or less) to document the code, and I choose to place the old variable name in a trailing comment to not lose that information.
I welcome feedback on these changes, since it wasn't always entirely easy to know what changes made the most sense in every situation.
Given the nature of `EOF` and `isEOF`, it seems to me that they really ought to be placed in `core/primitives.js` instead.
In general, it doesn't seem great to have to depend on the entire `core/parser.js` file for such simple primitives/helper functions.
In particular, while `core/ps_parser.js` is completely separate from `core/parser.js` with regards to its function, it still depends on the latter for just *one* primitive.
Note that compared to e.g. PR 7389, this will not reduce the number of dependencies for `core/ps_parser`, however the new dependency IMHO makes more sense.
Given that this patch causes a lot of churn in the addon code, I wouldn't really mind if we ultimately decide against doing this and just add a rule exception in mozilla-central instead.[1]
---
[1] Note that I used the ESLint `--fix` option, hence writing this commit message actually took longer time than the creation of the patch :-)
PR 7322 added the `PdfJsNetwork.jsm` file, instead of the general `src/core/network.js` file for the Firefox addon. However, `make.js` wasn't updated to actually stop including the now obsolete network file.
*This fixes a regression from commit https://github.com/mozilla/pdf.js/pull/7738/commits/c9a0955c9c7ed857c98696732c6edeb6f4901f43, i.e. PR 7738.*
Currently if you quickly rotate a document at least *twice*,[1] such that rendering of a page hasn't finished for the first rotation before the last rotation is triggered, the `cssTransform` method can fail to update the page correctly leading to it looking temporarily distorted.
The reason why things break is that previously we stored the `viewport` on the canvas DOM element, meaning that when it was accessed in `cssTransform` is was guaranteed to point to the `viewport` of the `zoomLayer` canvas.
Generally you want to avoid storing data on DOM elements this way, and during the `PDFPageView` refactoring needed to support SVG rendering, the previous `viewport` was instead stored directly on `PDFPageView`.
However, the problem is first of all that the `paintedViewport` only stores the *last* `viewport` computed, and second of all that there're no guarantees that it actually applies to the current `zoomLayer` canvas.
If a document is rotated slowly enough that rendering finishes *before* the next rotation then this problem doesn't exist, but for sufficiently quick rotations rendering will be cancelled at least once and the `paintedViewport` could thus be bogus.
The solution for the above problems is to ensure that we track the correct `viewport` for each DOM element (canvas or svg),[2] which seemed easist to do with a `WeakMap`.[3]
---
[1] I'm able to reproduce this using the `tracemonkey` file, but please note that for pages with few operations, i.e. that render very quickly, the effect may be hard to spot.
[2] One other possible solution that I briefly considered, was to wait until rendering finished before storing the current `viewport`. However, that would have caused issues with rotating a page before the *first* rendering operation had finished.
[3] This regression took me way longer to both figure out, and fix, than I'd like to admit :-)
*Please note: ignoring whitespace changes is most likely necessary for the diff to be readable.*
This patch addresses all the current, in `mozilla-central`, linting failures in the addon. It should thus be possible to change the `.eslintignore` entry for PDF.js in `mozilla-central` from `browser/extensions/pdfjs/**` to `browser/extensions/pdfjs/build/**` and `browser/extensions/pdfjs/web/**` instead.
Note that we cannot, for backwards compatibility reason of the general PDF.js library, at this time make similar changes for files residing in the `build` and `web` directories in `mozilla-central`.
The main changes in this patch are that we now use [classes](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Classes) instead of our previous "class-like" functions, and also use the more compact [object shorthand notation](https://developer.mozilla.org/en/docs/Web/JavaScript/Reference/Operators/Object_initializer#New_notations_in_ECMAScript_2015).
A couple of functions were also converted to [arrow functions](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Functions/Arrow_functions), to reduced usages of `bind(this)` and `var self = this`.
One caveat with ES6 classes is that it's not (yet) possible to define private constants/helper functions within them, which is why the `NetworkManagerClosure` was kept to not change the visibility of those constant/functions.
Besides testing in Firefox Nightly 53, this patch has also been tested in Firefox ESR 45 and SeaMonkey 2.46.
However, I'd gladly welcome help with testing the patch more, to ensure that nothing has gone wrong during the refactoring.
Fixes the first bullet point of issue 7957.
http://eslint.org/docs/rules/no-lone-blocks
Note that we currently have no code that violates this rule in the source files, but it seems that the built files are possibly affected (see issue 7957).
Please see http://eslint.org/docs/rules/spaced-comment.
Note that the exceptions added for `line` comments are intended to still allow use of the old preprocessor without linting errors.
Also, I took the opportunity to improve the grammar slightly (w.r.t. capitalization and punctuation) for comments touched in the patch.
Further adjust the heuristics used to detect OpenType font files with CFF data, to ensure that all Type0 fonts are handled the same way regardless of font Subtype (issue 7901)
Rename `annotation_layer_spec.js` to `annotation_spec.js` to better describe what is actually tested, and simplify the `FileAttachmentAnnotation` unit-test to avoid having to use the entire API in the test
Every other unit-test in `annotation_spec.js` is already only testing the annotation code. Hence it seems unnecessarily convoluted to make use of the API here, when we can (fairly) simply provide the necessary data explicitly as in all the other annotation unit-test.
Currently you have to open the files to be certain which tests each one will run, which we can avoid by appending the directory names to the file names of the tests themselves.
We're currently making use of `uniquePrefix`/`idCounters` in multiple files, to create unique object id's, and adding a new occurrence of them requires some care to ensure that an object id isn't accidentally reused.
Furthermore, having to pass around multiple parameters as we currently do seem like something you want to avoid.
Instead, this patch adds a factory which means that there's only *one* thing that needs to be passed around. And since it's now only necessary to call a method in order to obtain a unique object id, the details are thus abstracted away at the call-sites which avoids accidental reuse of object id's.
To test that this works as expected a very simple `Page` unit-test is added, and the existing `Annotation layer` tests are also adjusted slightly.
Using `else` after `return` is not necessary, and can often lead to unnecessarily cluttered code. By using the `no-else-return` rule in ESLint we can avoid this pattern, see http://eslint.org/docs/rules/no-else-return.
According to https://wiki.mozilla.org/RapidRelease/Calendar#Past_branch_dates: The *last* ESR version of Firefox 38 was released in April this year, and since June the only available ESR version has been based on Firefox 45.
Now that Seamonkey has *finally* released a new version, i.e. 2.46 which should correspond to Firefox 49, there doesn't seem to be any reason to keep the fallback code around in the addon anymore.
[api-minor] Ensure that the `getDocument` Promise is rejected if the `loadingTask` is destroyed, or an `Error` is thrown, inside of the `onPassword` callback (issue 7806)
We're already passing in a, currently unused, `PdfManager` instance when initializing the `XRef`. To avoid having to pass a single `password` parameter around, we could thus simply get the `password` through the `PdfManager` instance instead.
This patch also removes the `UpdatePassword` message, in favour of using the `sendWithPromise` method of `MessageHandler`.
Furthermore, the patch also refactors the `BasePdfManager_updatePassword`/`BasePdfManager_passwordChanged` methods (in pdf_manager.js), and the `pdfManagerReady` function (in worker.js).
I recently happened to look at the code I wrote for PR 5964, which fixed [bug 1157493](https://bugzilla.mozilla.org/show_bug.cgi?id=1157493), and I quickly realized that the solution is way too simplistic.
The fact that only using the `length` of a `Differences` array worked seems more like a happy accident for a particular set of font data, but could just as easily be incorrect for other PDF files.
Note that in practice, the case where the `Encoding` entry is a regular `Dict` (and not a `Ref` or `Name`) is very rare, hence I don't think that we really need to worry about having to reparse this data.
Also, the performance of this code-block is quite a bit better by updating the `hash` with the data from the *entire* `Differences` array, instead of at every loop iteration.
Let `finishPaintTask` in pdf_page_view.js return a promise instead, to avoid having to throw in the `paintTask.promise` rejection handler, and don't reject the `PDFPageView_draw` promise when rendering is `cancelled`
Changing this particular code makes me somewhat nervous about regressions, since PR 5770 necessitated the follow-up PR 6270.
However, the patch passes all tests added in those PRs (and obviously all other tests). Furthermore, I've manually checked all the issues/bugs referenced in PRs 5770 and 6270 without finding any issues.
**Please note:** This patch fixes *only* the font bug, not the SVG conversion, present on pages two and three of the PDF file in issue 7901.
As mentioned on IRC yesterday, we currently throw even when rendering is `cancelled`, which is annoying when the devtools are active. Furthermore, since `cancelled` isn't really an error, rejecting the `PDFPageView_draw` promise seems somewhat strange in that case.
Modern browsers support styling radio buttons and checkboxes with CSS.
This makes the implementation much easier, and the fallback for older
browsers is still decent.
I haven't got an example where the current code breaks, but given all the previous cases we've seen where PDF generators use indirect objects in Arrays it makes sense to fix this pro-actively.
I've modified the relevant unit-tests slightly, and they would *not* pass without the code changes in this patch.
*Note:* `Dict_getArray` only dereferences Array elements on the "top-level", to avoid recursion issues. Furthermore if you have to loop through the Array at the call-site anyway, then using `Dict_get` in combination with `XRef_fetchIfRef` is a tiny bit more efficient.
*Please note that most of the necessary code adjustments were made in PR 7890.*
ESLint has a number of advantageous properties, compared to JSHint. Among those are:
- The ability to find subtle bugs, thanks to more rules (e.g. PR 7881).
- Much more customizable in general, and many rules allow fine-tuned behaviour rather than the just the on/off rules in JSHint.
- Many more rules that can help developers avoid bugs, and a lot of rules that can be used to enforce a consistent coding style. The latter should be particularily useful for new contributors (and reduce the amount of stylistic review comments necessary).
- The ability to easily specify exactly what rules to use/not to use, as opposed to JSHint which has a default set. *Note:* in future JSHint version some of the rules we depend on will be removed, according to warnings in http://jshint.com/docs/options/, so we wouldn't be able to update without losing lint coverage.
- More easily disable one, or more, rules temporarily. In JSHint this requires using a numeric code, which isn't very user friendly, whereas in ESLint the rule name is simply used instead.
By default there's no rules enabled in ESLint, but there are some default rule sets available. However, to prevent linting failures if we update ESLint in the future, it seemed easier to just explicitly specify what rules we want.
Obviously this makes the ESLint config file somewhat bigger than the old JSHint config file, but given how rarely that one has been updated over the years I don't think that matters too much.
I've tried, to the best of my ability, to ensure that we enable the same rules for ESLint that we had for JSHint. Furthermore, I've also enabled a number of rules that seemed to make sense, both to catch possible errors *and* various style guide violations.
Despite the ESLint README claiming that it's slower that JSHint, https://github.com/eslint/eslint#how-does-eslint-performance-compare-to-jshint, locally this patch actually reduces the runtime for `gulp` lint (by approximately 20-25%).
A couple of stylistic rules that would have been nice to enable, but where our code currently differs to much to make it feasible:
- `comma-dangle`, controls trailing commas in Objects and Arrays (among others).
- `object-curly-spacing`, controls spacing inside of Objects.
- `spaced-comment`, used to enforce spaces after `//` and `/*. (This is made difficult by the fact that there's still some usage of the old preprocessor left.)
Rules that I indend to look into possibly enabling in follow-ups, if it seems to make sense: `no-else-return`, `no-lonely-if`, `brace-style` with the `allowSingleLine` parameter removed.
Useful links:
- http://eslint.org/docs/user-guide/configuring
- http://eslint.org/docs/rules/
Move the `Preferences` initialization/fetching code to the top of `PDFViewerApplication.initialize`, to enable using them when initializing e.g. the viewer components
Ideally we'd remove the 'localized' event from the `eventBus`, but for backwards compatibility we keep it in `GENERIC` builds.
Note that while we want to ensure that the direction attribute of the HTML is updated as soon as the `localized` Promise is resolved, we purposely wait until the viewer has been initialized to ensure that the 'localized' event will always be dispatched.
With `bindEvents()` now being called after the viewer has been initialized, we no longer need to have `PDFViewerApplication.initialized` checks in the event handler functions.
Furthermore by moving the `window.addEventListener`s to a helper method, `PDFViewerApplication.initialized` checks are no longer necessary in the event handlers, hence we thus address part of issue 7797 here as well.
Note that in quick testing using `console.time/timeEnd`, both locally and with the Firefox addon, the total run time of the *entire* `PDFViewerApplication.initialize` function does not seem to change with this patch.
This patch refactors the `CropBox` code to combine fetching and
validation code in a getter, like we already did for the `MediaBox`
property. Combined with variable name changes, this improves readability
of the code and makes the `view` getter simpler as well.
I've not actually, thus far, come across a PDF file that this patch fixes. However, given the string of recent patches that has fixed issues with indirect objects in arrays, I think that it makes sense to proactively avoid any issues in this code.
I just realized that none of our current unit-tests cover this particular part of the Page Label parsing code, hence this patch adjusts an existing test PDF to include a "St" entry in the Page Label dictionary.
This patch avoids the creation of extra arrays when initializing an
array with default (zero) values. Doing this additionally makes the code
more readable by allocating enough space for the number of color
components.
It seems that for normal web pages, at least in Firefox, the keyboard shortcuts <kbd>Ctrl</kbd> + <kbd>Up</kbd>/<kbd>Down</kbd> are functionally equivalent to <kbd>Home</kbd>/<kbd>End</kbd>. This is obviously an edge-case, but can be easily implemented by using the same logic as we do for <kbd>Home</kbd>/<kbd>End</kbd>.
Fixes 7852.
*Please note:* I'm finding it slightly difficult to interpret issue 7852, and bug 1285719, since among other things: the title includes the word "reverse" with no other mention of it, and the STR makes reference to print preview which doesn't seem applicable to the PDF viewer.
However, compared to regular web pages in Firefox, I think the behavior of this patch makes sense here.
For Arabic characters, the Unicode character codes are mapped to Unicode
character types using the character codes for indexing. However, the
character code 0x061D is undefined (and therefore invalid) in the
Unicode standard. The imported list does not contain this entry, but not
having it in the list breaks indexing for items after it. Therefore, put
an empty string on its position to make indexing work properly and issue
a warning in the unlikely event that we encounter this character.
This patch moves the user agent checks to the top of the file to reduce
duplication and to provide a clear overview of which user agent we are
detecting.
Moreover, we extract inline user agent checks as well and use existing
checks in more places. Finally, we fix the indenting in one place for
consistency.
*This patch fixes something that I noticed while debugging https://bugzilla.mozilla.org/show_bug.cgi?id=1308536.*
The PDF file contains a font called "NuptialScript", which unfortunately is not embedded. Since that is a non-standard font we will not be able to render it entirely correct. However, by adding "NuptialScript" to the `getNonStdFontMap`, we can at least improve the rendering slightly by using an italic (serif) fallback font.
This patch adds support for non-embedded Arial Black fonts, that use a `Arial-Black...` format for the font names.
Also, this patch changes `canvas.js` such that we always render Arial Black fonts with the maximum weight, which actually improves a number of existing test-cases. This should thus explain the test "failures", which are clear improvements compared with e.g. Adobe Reader.
Fixes 7835.
For consistency, I also renamed the `FIREFOX/MOZCENTRAL` sessionStorage key, but given that sessionStorage is a lot less permanent than localStorage it didn't seem necessary to migrate any existing values.
Fixes 7760.
Currently if you try to enable SVG rendering in the addons, a `TypeError` is thrown by the browser since we have code that depends on what `paintOnCanvas`/`paintOnSvg` (should) return.
For commands with with too few arguments, clear out `args` if it's an Array instead of replacing it with `null` in `EvaluatorPreprocessor_read` (issue 7804)
For `PartialEvaluator_getTextContent`, the same `args` Array should be re-used for every `EvaluatorPreprocessor_read` call. Hence we want to ensure that it's not accidentally replaced with `null` in `EvaluatorPreprocessor_read`, since otherwise corrupt PDF files (with too few arguments for certain commands) will cause errors in `PartialEvaluator_getTextContent`.
Perhaps a micro-optimization, but this patch also changes two `!args` comparisons to `args === null`, since that should be a tiny bit more efficient.
According to e.g. issue 7754, it appears that the current `isSafari` check is failing in newer version of the browser. Despite the fact that checking the userAgent is an anti-pattern, which should be avoided, it's currently the simplest solution.
By only allowing very specific type of `JavaScript` actions, and also utilizing the existing `URL` validation, this patch shouldn't pose too much risk.
Fixes one of the points in issue 3897 (with the PDF file taken from issue 3438).
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=843699 (probably, since that bug doesn't contain a test-case).
Given that the `customScaleOption` should already be hidden, provided that the browser supports that, this patch also prevents it from being accessible via the keyboard.
As far as my testing goes in various browsers, this doesn't seem to have any ill effects, and note that we're already explicitly ignoring the `custom` value in the `select` event handler.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1315608.
This patch resolves the responsiveness issues for the toolbar in the
viewer. Depending on the language (for example the Dutch language),
elements could overlap when the viewport size is reduced.
The main issue here is that the CSS rules are unnecessarily complex and
handle lots of different cases (LTR/RTL, displacements for specific
viewport widths, et cetera). By removing this complexity and letting the
browser handle the responsiveness, we not only get simpler CSS rules and
HTML mark-up, but the responsiveness issues are mostly fixed at the same
time. We no longer have to position the elements manually (by setting
their `left` attribute value) anymore.
It seems that certain bad PDF generators can create badly encoded "Prefix" entries for Page Labels, one example being http://ukjewishfilm.org/wp-content/uploads/2015/09/Jewish-Film-Festival-Programme-ONLINE.pdf.
Unfortunately I didn't come across such a PDF file while adding the API support for Page Labels, but with them now being used in the viewer I just found this issue. With this patch, we now display the Page Labels in the same way as Adobe Reader.
The original code is difficult to read and, more importantly, performs
actions that are not described in the specification. It replaces empty
names with a backtick and an index, but this behavior is not described
in the specification. While the specification is not entirely clear
about what should happen in this case, it does specify that the `T`
field is optional and that multiple field dictionaries may have the same
fully qualified name, so to achieve this it makes the most sense to
ignore missing `T` fields during construction of the field name. This is
the most specification-compliant solution and, judging by opened issue #6623, also the required and expected behavior.
I got tired of staring at a bunch of localization warnings every time that I open the console, hence this patch adds the missing translations to the Swedish locale.
- Renamed startPrint to performPrint to emphasize that the method
does not start the print process (preparing pages for the printer),
but that it does the actual printing (sending pages off to the
printer).
- Put performPrint in the PDFPrintService, so that it can be
overridden if needed.
- Move the global scratchCanvas to PDFPrintService. This is mainly to
make it easier to reason about the state of scratchCanvas. In practice
there is no difference because only one PDFPrintService instance can
be instantiated at any given time.
- Move all logic of using the rendered page to one location.
This makes it easier to replace the printing logic later, when I add
special handling to out-of-process frames in the Chrome extension.
Make sure that the print service is stopped as soon as possible when
aborted, and that it is not possible for a (slow) promise to
accidentally wipe the state of a print job that was started later.
Remove/deprecate specifying a pageNumber directly after the hash symbol (#), to improve compatibility since other PDF viewers don't support this form (issue 7746)
There's no mention of our `#{pagenum}` form in http://www.adobe.com/content/dam/Adobe/en/devnet/acrobat/pdfs/pdf_open_parameters.pdf, and Adobe Reader doesn't seem to support it either.
Hence this patch removes support for it in the extensions, but keeps it in the `GENERIC` build with a deprecation warning and a fallback to handle it as a destination.
Fixes 7746.
This patch implements the page label functionality in a similar way as Adobe Reader.
For documents with page labels, if a non-existent page label is entered we'll try to fallback to the page number instead.
The patch also includes a preference (`disablePageLabels`), to make it easy to opt-out of using page labels if the user/implementor so wishes.
The way that `get/set currentPageLabel` is implemented in `PDFViewer`, is as wrappers for the corresponding `get/set currentPageNumber` functions, since that seemed like the cleanest solution.
The page labels are purposely *only* added to the page controls in the viewer UI, and not stored in e.g. the `ViewHistory`. Since doing so would mean adding unnecessary code complexity, without any real added value, and would also mean delaying the inital loading of PDF documents.
Note that this patch will ignore page labels if they are identical to standard page numbering, since in this case displaying the page labels adds no value (but only UI noise). The reason for handling this case specially, is that in practice a surprising number of PDF files include "pointless" page labels.
The following reasoning was used for deciding to remove the "Page: " label, and replace it with a tooltip, from the main toolbar:
- We have no other visible labels in the *main* toolbar (e.g. the Zoom dropdown doesn't have a label, but only a tooltip).
- We already hide the "Page: " label when the viewer is narrow.
- The varying width of the "Page: " label in different locales is already causing issues for many languages, with overlap in the main toolbar as a result.
Trying to create responsive CSS styles that works well in all locales is already difficult, and if we add support for page labels that will only further compound the issues.
- Some PDF viewers (e.g. Adobe Reader, pdfium in Chrome) doesn't show labels in the UI by default.
In some PDF files, the first element (i.e. the one containing either a `Ref` or a `Number` pointing to a page) of the explicit destination Array may be bogus.
One such example is actually the file `pdf.pdf` in the test-suite, where some destinations are incompletely specified. One such example being the `G1.998360` destination whose explicit destination Array contains `[null, /XYZ, 54, 488, null]`, i.e. the destination page is `null`.
Hence this patch adds a bit more validation for that case. It also adds an additional check to ensure that the resulting `pageNumber` is non-negative, and finally a couple more error messages for existing cases of malformed data.
In general we neither want, nor can, support arbitrary `Launch` actions. But in practice, all the cases we've seen so far just contains relative URLs to other PDF files. Building on PR 7689, we can thus at least support basic `Launch` actions.
Note that in `FIREFOX/MOZCENTRAL/CHROME` builds of the standard viewer the `docBaseUrl` parameter will be set by default, since in that case it makes sense to use the current URL as a base.
For the `GENERIC` viewer, or the API itself, it doesn't make sense to try and set the `docBaseUrl` by default. However, custom deployments/implementations may still find the parameter useful.
Note that this will automatically reject any relative URL.
To make the API more useful to consumers, URLs that are rejected will be available via the `unsafeUrl` property in the data object returned by `PDFPageProxy_getAnnotations`.
The patch also adds a bit more validation of the data for `Named` actions.
This not only reduces code duplication, but it also allow us to easily support the same kind of URLs we currently do for Link annotations in the Outline as well.
This requires the `run-sequence` dependency because Gulp does not have a
way to run sequences of tasks inside a Gulp task. Gulp 4.0 will have
support for this, but until that is released this is the recommended way
to implement it.
The outline toggle button has a feature where it can be double-clicked
to expand/collapse all items shown therein. Although this is described
in the FAQ, can go potentially unnoticed. This, however, being a useful
feature, advertise on the tool tip itself.
l10n translation for en-US and IDs updated.
Signed-off-by: Jeenu Viswambharan <jeenuv@gmail.com>
Each well-formed SVG image has the following structure:
SVG element
- Definitions element
- Root group
- Other group 1
- ...
- Other group n
This patch factors out initialization code into a private method in such
as way that the creation of this structure is clear from the code. The
root group is the replacement for the parent group from before. We need
this group as we cannot apply the viewport transform on the SVG element
itself (this caused issues in Chrome). If other code appends groups to
the SVG image, in reality it is appending those groups to the root
group, but this detail is abstracted away by this patch.
This patch ensures that we only create transformation groups when it is
actually required and that we re-use transform groups as much as possible.
It reduces the number of transform groups for the Tracemonkey paper from
2790 to 1271, thereby making the DOM much lighter and rendering/scrolling
smoother. Moreover, it simplifies the code and prevents duplication.
Finally, we issue a warning when an unimplemented graphic state is
encountered. Before, this was ignored silently, making debugging harder.
This patch:
- resolves a warning in the console about missing character encoding;
- makes the viewer use the same background color and PDF file as the
regular viewer;
- simplifies the example to bring it more in line with the other
examples.
Let `Parser_makeFilter` pass in the `DecodeParms` data to various image `Stream`s, instead of re-fetching it in various `[...]Stream.prototype.ensureBuffer` methods
In `Parser_filter` the `DecodeParms` data is fetched and passed to `Parser_makeFilter`, where we also make sure that a `Ref` is resolved to a direct object.
We can thus pass this along to the various image `Stream` constructors, to avoid the current situation where we lookup/resolve data that is already available.
Note also that we currently do *not* handle the case where `DecodeParms` is an Array entirely correct in the various image `Stream`s, and this patch fixes that for free.
While the array argument to TJ should only contain strings and numbers, other
unfortunate items are found in PDFs in the wild, e.g.:
[(Grandes) 0.0 Tc
-250.0 (Client\350les,) 0.0 Tc
-250.0 (Financements) 0.0 Tc
-250.0 (et) 0.0 Tc
-250.0 (March\351s) ] TJ
getOperatorList already properly ignores any non-string, non-numeric values in
TJ arrays; without this patch to getTextContent, returned text items can have
NaN widths due to calculations being applied to those non-numeric values.
For PDF files with multiple `/Filter`s, where the `/Length` entry is zero, we fail to render the file correctly. The reason is that `maybeLength` is `null` for the every filter except the first, and `!maybeLength` is thus truthy.
Hence it seems that we should completely ignore the `/Length` entry and also explicitly check `maybeLength === 0`.
Note that I've not (yet) come across a PDF file with this issue in the wild, but given all the stupid things PDF generators do I wouldn't be surprised if such a file actually exists. In order to prevent a possible future bug, I'm submitting this patch which includes a hand-edited PDF file that we currently cannot render correctly (but e.g. Adobe Reader can).
Directly use the hexadecimal representation, just like the
`AnnotationFlags`, to avoid calculations and to improve readability.
This allows us to simplify the unit tests for text widget annotations as
well.
When opening a PDF file that triggers the browser fallback bar in the Firefox addon/built-in version, e.g. http://web.archive.org/web/20110918100215/http://www.irs.gov/pub/irs-pdf/f1040.pdf (with forms *disabled*), and then reloading the document an error can be thrown.
The reason is that displaying the fallback bar triggers 'resize' events, and they can arrive *before* the viewer has had a chance to run all the necessary initialization code.
Unfortunately PR 7633 missed, and I didn't catch it during review, to update `test/driver.js` such that the `forms` tests takes the correct code-path.
This resultet in the `forms` reference test images looking better than they should, and more problematicly differing from the rendering in the viewer.
With this patch, the tests now correctly skip over any `Appearance` streams.
The `forms` tests now highlights quite clearly (e.g. look at `annotation-tx2.pdf`/`annotation-tx3.pdf`) that we cannot just skip the `Appearance` streams when rendering forms. Hence we're going to have to find a way to fix that *before* enabling forms by default, since both display *and* print would look completely wrong otherwise.
Finally, this patch also uncovers one more existing bug that still needs to be fixed, since the current `rasterizeAnnotationLayer` in `test/driver.js` isn't able to handle the contents of e.g. `<input>` and `<textarea>`.
When debugging issue 7643, I noticed that the `forms` tests currently doesn't look like the rendering in the viewer (with `renderInteractiveForms = true` set).
After scratching my head for a little while, I realized that PR 7633 make the implicit assumption that `Page_getOperatorList` (in `core/document.js`) is called *before* fetching the annotation with `PDFPageProxy_getAnnotations` (in `display/api.js`).
Hence this patch, that changes it so that we instead pass in the `renderInteractiveForms` parameter to `Annotation_appendToOperatorList` to ensure that it's always correctly set.
A user encountered a response that looks like:
URL: some gibberish
Headers:
Content-Type: application/octet-stream
Content-Disposition: attachment; filename="something.pdf"
In the Chrome extension, the "attachment" content disposition is almost
always ignored (i.e. the PDF Viewer will try to view it anyway). So we
need to fall back to the Content-Disposition header if the URL check is
inconclusive.
If interactive forms are enabled, then the display layer takes care of
rendering the form elements. There is no need to draw them on the canvas
as well. This also leads to issues when values are prefilled, because
the text fields are transparent, so the contents that have been rendered
onto the canvas will be visible too.
We address this issue by passing the `renderInteractiveForms` parameter
to the render task and handling it when the page is rendered (i.e., when
the canvas is rendered).
This patch improves the unit tests by testing the support for read-only
and multiline fields. Moreover, we add a reference test to ensure that
the text widgets are not only rendered, but also that their contents are
styled properly.
Finally, we perform minor improvements in `src/core/annotation.js`, for
example adding missing comments.
For embedded Type1 fonts without included `ToUnicode`/`Encoding` data, attempt to improve text selection by using the `builtInEncoding` to amend the `toUnicode` map (issue 6901, issue 7182, issue 7217, bug 917796, bug 1242142)
I intended to provide proper benchmarking results here, as outlined in https://github.com/mozilla/pdf.js/wiki/Benchmarking-your-changes, but after wasting a couple of hours over the weekend getting weird results I gave up.
It appears that there's a lot of, i.e. way too much, variance between subsequent runs of `text` tests for the results to be meaningful.
(Previously I've only benchmarked `eq` tests, so I don't know if the `text` tests has never worked well or if it's a newer problem. For reference, please see the results of back-to-back benchmark runs on the current `master` with a *very* simple manifest file: [link here].)
Instead I used `console.time/timeEnd` in `appendText` and `expandTextDivs` to be able to compare the performance with/without the patch. The entire viewer was (skip-cache) reloaded between measurements, and the result are available here: [link here].
Given the troubles I've had with benchmarking, I've not yet computed any statistics on the results (e.g. mean, variance, confidence intervals, and so on).
However, just by looking at the data I think it's safe to say that this patch first of all doesn't seem to regress the current performance. Secondly it certainly looks *very* likely that this patch actually improves the performance, especially for the one-glyph-per-text-div case (cf. issue 7224).
Re: issue 7584.
Even though this patch passes all tests (unit/font/reference) locally, including the new ones that I added in PR 7621, I'm still a bit nervous about modifying the code that choose the fallback encoding for fonts without an `/Encoding` entry.
Note that over the years this code has been changed on a number of occasions, see a possibly incomplete [list here], to deal with various cases of incorrect font data.
According to the PDF specification, see http://www.adobe.com/content/dam/Adobe/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#G8.1904184, it seems that we should fallback to the `StandardEncoding` for Nonsymbolic fonts.
There's obviously a risk that fixing this particular issue *could* break other PDF files for which we don't have tests. However I've tried to change the logic as little as possible in this patch, to hopefully reduce possible breakage.
Based on debugging numerous font issue, it seems that a lot of fonts actually set the Symbolic flag, even when they are in fact *not* Symbolic. Fonts actually marked as Nonsymbolic seem to be somewhat less common, which I hope should reduce the risk of the patch somewhat.
Fixes 7580.
Note that in `parseCodestream` I purposly left the `throw new Error` instances inside of the `try` block, since we don't want to throw any `Errors` while in recovery mode.
Finally somewhat unrelated to the rest of the patch, but I moved the `doNotRecover` variable declaration outside of the `try` block to avoid variable hoisting given that it's accessed inside the `catch` block.
Note that in order to prevent any possible issues, this patch does *not* try to amend the `toUnicode` data for Type1 fonts that contain either `ToUnicode` or `Encoding` entries in the font dictionary.
Fixes, or at least improves, issues/bugs such as e.g. 6658, 6901, 7182, 7217, bug 917796, bug 1242142.
Moreover, we refactor the code a bit to extract code that is shared
between the two branches and we only apply text alignment (and create
the array) when it is actually defined, since it's optional and left is
already the default.
I've started to look into how we can fix issue 7580, but quickly became worried that fixing it could easily mean that we'd trade one fixed PDF file for a multitude of broken ones.
Hence I started going through the history of the code that choose the fallback encoding, and noticed that it has been changed a number of times over the years to deal with various cases of weirdness/errors in non-embedded fonts.
To my relief it turned out that almost all the PRs, please see a possibly incomplete [list here], that changed this code actually included `eq` test-cases.
However, in one case it appears that a PR missed to add a test-case. Furthermore since the fallback encoding may also be the only source for creating a `toUnicode` map, changing the encoding could possibly regress only the text-selection despite a PDF file still rendering correctly.
Therefore, this PR adds one new `eq` test, and also a number of additional `text` tests for PDF files already present in the test-suite.
Note that it's obviously possible that there's a certain overlap between the added tests, but I'd be *a whole lot* more concerned with causing regressions.
Don't duplicate the first entry in the `charCodeToGlyphId` map for CIDFontType2 fonts with a `CIDToGIDMap` that already mapped the first entry to a non-zero `glyphId` (issue 7544)
There's really no good reason to attempt to adjust the `max-height` of the `SecondaryToolbar` when it's not visible, so let's not do that anymore.
Also, we can listen for the `resize` event in `SecondaryToolbar`, to avoid having to manually call the `max-height` adjust function from various event handlers in `app.js`. Please note that by always adjusting the `max-height` when the toolbar is opened we no longer need the `localized` event, since it was mainly used to set a correct inital `max-height` value.
Please note that this is a hack, but I think that it should be OK for now to atleast get the preference landed. Refer to the code comment for further information.
Re: issue 7584 and PR 7586.
This patch is the first step towards implementing support for
interactive forms (AcroForms). It makes it possible to render text
widget annotations exactly like Adobe Reader/Acrobat.
Everything we implement for AcroForms is disabled by default using a
preference, mainly because it is not ready to use yet, but has to
implemented in many steps to avoid complexity. The preference allows us
to work with the code while not exposing the behavior by default. Mainly
storing entered values and printing them is still absent, which would be
minimal requirements for enabling this by default.
This patch is yet another instalment in the (never ending) series of patches for PDF files that specify completely incorrect Type/Subtype for its fonts. In this case Type1/Type1C, when in fact OpenType would have been correct.
Fixes 7598.
Currently, we only support text widget annotations (field type 'Tx')
partially. However, the current code does not make this entirely clear
and does not provide a warning when an unsupported field type is
encountered, making it harder to determine why rendering fails.
Moreover, in the display layer we make no distinction between the
various types of widget annotations, causing the code for text widget
annotations to also be executed for other types of widget annotations in
a fallback situation.
This patch improves the structure of the widget annotation code. In the
core layer, we use the same structure we use for non-widget annotations
in the factory and provide a clear warning when an unsupported type is
encountered. In the display layer, we do the same and split the
`WidgetAnnotationElement` class into two classes, namely
`TextWidgetAnnotationElement` for text widget annotations and
`WidgetAnnotationElement` for other unsupported annotations as a
fallback. From this it clear that we only support text widget
annotations and nothing else.
In the case where the document was destroyed, we were rejecting the `Promise` in `JpegDecode` with a string instead of an `Error`. The patch also brings the wording more inline with other such rejections.
Use the `isInt` utility function when validating the `pageNumber` parameter in `WorkerTransport_getPage`, to make it more obvious what's actually happening. There's also a couple more unit-tests added, to ensure that we always fail in the expected way.
Finally, we can simplify the rejection handling in `WorkerTransport_getPageIndexByRef` somewhat. (Note that the only reason for using `catch` here is that since the promise is rejected on the worker side, the `reason` becomes a string instead of an `Error` which is why we "re-reject" on the display side.)
This patch avoids having to calculate the scale twice by saving it in
the properties object.
Moreover, we remove a temporary variable and place parentheses around a
calculation inside a string concatenation.
This allows us to remove the `try/catch` statements used in `src/core/stream.js` when parsing JPEG images.
As far as I can tell, the only reason for the current usage of plain `throw` is that `jpg.js` originally was external code. Given that this code now lives in our repo, this patch brings the JPEG code more in line with e.g. `src/core/jpx.js` and `src/core/jbig2.js`.
Assign the `quantizationTables` after parsing the entire JPEG image, to prevent issues when the DQT (Define Quantization Tables) marker is encountered after SOF{n} (Start of Frame) markers (issue 7406)
This patch improves performance by avoiding unnecessary type
conversions, which also help the JIT for optimizations.
Moreover, this patch fixes issues with the div expansion code where
`textScale` would be undefined in a division. Because of the `dataset`
usage, other comparisons evaluated to `true` while `false` would have
been correct. This makes the expansion mode now work correctly for cases
with, for example, each glyph in one div.
The polyfill for `WeakMap` has been provided by @yurydelendik.
Move the `Preferences` initialization/fetching code to the top of `PDFViewerApplication.initialize`, and add a `enhanceTextSelection` preference to the viewer
We pass many parameters to `appendText` while we might as well pass the
`task` object that contains them. This saves a few lines of code and
makes the signature of `appendText` more clear. We do the same for
`expand`, which is useful for the next commit in which we replace
`div.dataset` with a `WeakMap`.
Furthermore, this patch adds a missing parameter to a comment block to
make it clear which parameters remain.
When clicking on the `pageNumber` input, or when using the keyboard shortcut <kbd>Ctrl</kbd>+<kbd>Alt</kbd>+<kbd>G</kbd>, the element isn't just focused but its contents is actually *selected* so that the user doesn't need to clear the input before entering a new `pageNumber`.
However, the `GoToPage` named action is only using `focus`, so let's change this to be consistent.
The following document can be used for testing: http://www2.informatik.uni-freiburg.de/~frank/ENG/beamer/example/Beamer-class-example1.pdf#zoom=page-fit
[screeenshot]
1. Expanding divs to improve text selection. (Yury)
2. Adding enhanceTextSelection as an option.
3. Moving feature functionality from text_layer_builder.js to text_layer.js.
4. Added expandTextDivs method to only load expanded divs on first click, and only show on subsequent clicks
The PDF file contains an image that we're allowed to use, since it's just the PDF.js logo.
The logo image was simply inverted (so that it requires a /Decode entry in the image dictionary that triggers the use of `jpg.js` instead of the browser), converted to JPEG, and finally edited by hand to change the order of the DQT/SOF{n} markers.
Ensure that the zoom buttons are disabled correctly if the `scale` is smaller/larger than `MIN_SCALE/MAX_SCALE` in `PDFViewerApplication._updateUIToolbar`
In the `zoom{In, Out}` functions in `PDFViewerApplication`, we prevent the zoom value from becoming smaller/larger than `MIN_SCALE/MAX_SCALE`.
However, if the user sets the zoom level through the hash parameter the zoom buttons might not be correctly disabled; try http://mozilla.github.io/pdf.js/web/viewer.html#zoom=10.
Note that this issue has been present since "forever", but given that the solution is so simple I think that we should just fix this. (I'm also fixing a stupid typo I previously made in the JSDoc comment.)
Since the `mobile-viewer` example is based on the old FirefoxOS/B2G PDF viewer, it didn't need to have the same kind of `open/close` methods as the default viewer.
However, now that it has been re-purposed as a simple `mobile-viewer` example, it seems like a good idea to ensure that it has proper asynchronous `open/close` methods.
Fixes 7571.
When adding new entries to `ProblematicCharRanges`, you have to be careful to not make any mistakes since that could cause glyph mapping issues.
Currently the existing reference tests should probably help catch any errors, but based on experience I think that having a unit-test which specifically checks `ProblematicCharRanges` would be both helpful and timesaving when modifying/reviewing changes to this code.
Hence this patch which adds a function (and unit-test) that is used to validate the entries in `ProblematicCharRanges`, and also checks that we don't accidentally add more character ranges than the Private Use Area can actually contain.
The way that the validation code, and thus the unit-test, is implemented also means that we have an easy way to tell how much of the Private Use Area is potentially utilized by re-mapped characters.
Instead of having `Parser_getObj` fail unconditionally for the referenced PDF file, this patch attempts to let searching for the main trailer continue even if there are errors.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1250079.
This is similar to the existing `isCmd` and `isDict` functions, which already support similar kind of checks.
With the updated `isName` function, we'll be able to simplify many callsites from: `isName(someVariable) && someVariable.name === 'someName'` to: `isName(someVariable, 'someName')`.
With the viewer code now being split into various components/files, having an obsolete comment in `PDFViewer` that references thumbnails despite there being no other mentions of them in the entire file seems strange.
*Note:* This comment is simply a left-over from older versions of PDF.js, where the *entire* default viewer code was placed in just one file (and where we unconditionally created thumbnails, regardless whether they were visible or not).
There are PDF generators which create destinations with e.g. too large top values, which cause the wrong page to be scrolled into view because the offset becomes negative.
By ignoring negative offsets, we can prevent this issue, and get a similar behaviour as in Adobe Reader.
However, since we're also using `PDFViewer_scrollPageIntoView` in more cases than just when links (in the document/outline) are clicked, the patch adds a way to allow the caller to opt-out of this behaviour.
In e.g. the following situations, I think that we still want to be able to allow negative offsets: when restoring a position from the `ViewHistory`, when the `viewBookmark` button is used to obtain a link to the current position, or when maintaining the current position on zooming.
Rather than adding another parameter to `PDFViewer_scrollPageIntoView`, I've changed the signature to take an parameter object instead. To maintain backwards compatibility, I've added fallback code enclosed in a `GENERIC` preprocessor tag.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=874482.
With PR 7502 we no longer dispatch an event when the `val` is out of bounds, so to better communicate why nothing happens this patch logs an error in that case (similar to the logging of errors when trying to set an invalid scale).
The way that the default viewer is currently implemented, means that e.g. keyboard short-cuts could trigger the new error. Hence this patch also adds the necessary validation code, both to `app.js` and `pdf_link_service.js` to prevent unnecessary errors.
The AMO extension is no longer supported as PDF.js is integrated in
Firefox. It was already removed from the bot's preview messages around a
year ago and has not been used or updated since.
This patch improves the performance of issue 5808, but I'm not sure if it's enough to call it fixed. On average, this patch reduces the number of textLayer div's by a factor of 3, and it also reduces the time spend in `getTextContent` by a factor of ~2.
The PDF file is generated by `Scribus PDF`, which for reasons I cannot understand is placing redundant `Tf` commands before *every* showText command.
Note how the PDF file also contains lots of (basically) identical fonts, but with slightly different names, which causes unnecessary font-switching. This causes some unnecessary breaking of textLayer div's, but this issue cannot be easily worked around.
[api-minor] Add a parameter to `PDFPageProxy_getTextContent` that controls whether `PartialEvaluator_getTextContent` will attempt to combine same line text items
Note that I used a separate warning message for this case, instead of utilizing the same one as in the unsupported subtype case, to more clearly indicate that the PDF file itself is to blame rather than PDF.js.
Fixes 7446.
This patch attempts to cleanup a couple of things:
- Remove the `previousPageNumber` paramater. Prior to PR 7289, when the events were dispatched even when the active page didn't change, it made sense to be able to detect that in an event listener. However, now that's no longer the case, and furthermore other similar events (e.g. `scalechanging`/`scalechange`) don't include information about the previous state.
- Don't dispatch the events when the value passed to `set currentPageNumber` is out of bounds. Given that the active page doesn't change in this case, again similar to PR 7289, I don't think that the events should actually be dispatched in this case.
- Ensure that the value passed to `set currentPageNumber` is actually an integer, to avoid any issues (note how e.g. `set currentScale` has similar validation code).
Given that these changes could possibly affect the PDF.js `mochitest` integration tests in mozilla-central, in particular https://dxr.mozilla.org/mozilla-central/source/browser/extensions/pdfjs/test/browser_pdfjs_navigation.js, I ran the tests locally with this patch applied to ensure that they still pass.
Fonts that are not referenced by `Ref`s are very uncommon in practice, but it can unfortunately happen. In this case, we're currently not caching them in the usual way, i.e. by `Ref`, which leads to failures when a page is rendered after `cleanup` has run.
The simplest solution would have been to remove the `font.translated` workaround, but since this would have meant loading these kind of fonts over and over, the patch attempts to be a bit clever about this situation.
Note that if we instead loaded fonts per *page*, instead of per document, this issue wouldn't have existed.
Originally, I was just going to change this code to use `Ref_toString` in a couple more places. When I started reading the code, I figured that it wouldn't hurt to clean up a couple of comments. While doing this, I noticed that the logic for the (rare) `isDict(fontRef)` case could do with a few improvements.
There should be no functional changes with this patch, but given the added reference checks, we will now avoid bogus `Ref`s when resolving font aliases. In practice, as issue 7403 shows, the current code can break certain PDF files even if it's very rare.
Note that the only thing that this patch will change, is the `font.loadedName` in the case where a `fontRef` is a reference *and* the font doesn't have a descriptor. Previously for `fontRef = Ref(4, 0)` we'd get `font.loadedName = 'g_d0_f4_0'`, and with this patch `font.loadedName = g_d0_f4R`, which is actually one character shorted in most cases. (Given that `Ref_toString` contains an optimization for the `gen === 0` case, which is by far the most common `gen` value.)
In the already existing fallback case, where the `fontName` is used to when creating the `font.loadedName`, we allow any alphanumeric character. Hence I don't see how (as mentioned above) e.g. `font.loadedName = g_d0_f4R` would be an issue here.
From the discussion in issue 7445, it seems that there may be cases where an API consumer would want to get the text content as is, without combined text items.
After PR 7039, the PDF file in issue 7492 no longer renders at all, but note that text selection wasn't working correctly previously.
The problem with the PDF file in issue 7492 is that the `cMap`, in the `toUnicode` entry in the font, contains an invalid name:
```
/CMapName /-usr-share-fonts-truetype-Panton-Panton Family-Fontfabric - Panton.otf,000-UTF16 def
```
When we parse that line, things obviously break because there are spaces present in the wrong places.
To avoid that issue, the patch simply lets `parseCMap` continue when errors are encountered, to try and recover usable data. Note that by not aborting immediatly when an error is encountered, we are also able to fix the text selection.
Obviously, it could be argued that we should just immediatly reject a corrupt `cMap`. But given that they usually are correct, it seems that trying to recover as much data as possible from corrupt one can only be a good thing for both glyph mapping and text selection.
Fixes 7492.
In the PDF file in the issue, some of the glyphs end up being mapped to the Lepcha Unicode block; see https://en.wikipedia.org/wiki/Lepcha_(Unicode_block).
This didn't use to matter, but after HarfBuzz updates that improved support for Lepcha fonts, in particular https://bugzilla.mozilla.org/show_bug.cgi?id=1249861, some glyphs are now moved horizontally.
To avoid that, this patch adds the Lepcha block to the list of Unicode ranges that we skip when building the glyph mapping.
Fixes 7426.
After PR 7289, we'll now reset the current page view in cases where I don't think we should. To avoid this, this patch ensures that we'll not modify the position when the page number is out-of-bounds.
**STR:**
1. Open http://mozilla.github.io/pdf.js/web/viewer.html#page=1&zoom=auto,-98,696
2. Enter an invalid number, e.g. `1000`, in the `pageNumber` input.
**ER:**
The current position in the document shouldn't change, since the page number wasn't valid.
**AR:**
The document resets to the top of the page `1`.
After PR 7441, where `recoverGlyphName` is used a lot more than before, many PDF files will generate a lot of warnings the console. For normal usage, compared to debugging/development, this is probably more annoying than helpful.
Fallback to attempt to recover standard glyph names when amending the `charCodeToGlyphId` with entries from the `differences` array in `type1FontGlyphMapping` (issue 7439)
The method signature was improved in PR 6546, which was included in the `1.2.109` release from last November.
Hence I think that we should now be able to remove the fallback code for the old method signature. Note that this patch now throws an `Error` in `GENERIC` builds, to clearly indicate that the `open` callsite must be modified.
In PR 7273, we simplified the `MOZCENTRAL` specific check for fullscreen support. Unfortunately I've just, by coincidence, found out about https://bugzilla.mozilla.org/show_bug.cgi?id=1268749, which disabled the unprefixed fullscreen support in release versions of Firefox. If we do nothing, this will lead to Presentation Mode being unavailable in future Firefox releases.
This patch should fix things for now, and I'm afraid that we'll need to uplift it to Firefox 49 as well.
With the changes in PR 7289, we no longer dispatch a 'pagechanging' event on load. Since most PDF documents open on the first page, this means that the `previous` and `firstPage` buttons are no longer correctly disabled.
To avoid this, this patch moves the code that updates various UI toolbar state into one method, which is then called on document initialization and from the various existing event handling functions.
To reduced the amount of compatibility code that we need to maintain in the addon, I think that we should be more explicit about which Firefox versions we intend to support. In the future, this should allow us to more quickly remove old compatibility code.
Furthermore, this patch also tries to be more explicit about the addon support for Seamonkey, by defining it in terms of the underlying Firefox version.
Move the Chromium-specific URL rewriting code to the top of viewer.js
(before the PDF.js library) to make sure that the URL is as expected.
This ensures that variables that synchronously depends on the URL
(e.g. PDFViewerApplication.initialBookmark) are properly set.
Currently `importl10n` fails, since mxr.mozilla.org has been down for a while. Based on discussions in #developers, it isn't clear to me if/when it'll be available again.
Furthermore, I'm not sure why we're even getting the l10n files through a code search tool, rather than just using the official Mozilla repository directly.
From the discussion in issue 7386, it wasn't really clear if we can restrict addon support to Firefox `45` (i.e. the version that corresponds to the *current* ESR version).
However, we have a bunch of code for *very* old Firefox versions. Hence this patch changes the minimum supported version to Firefox `38` (which was released on `2015-05-12`, and correspond to the *previous* ESR version), and removes code that only applies to old Firefox versions.
Regardless what we end up deciding regarding addon support for previous Firefox versions, given the amount of code that even the Firefox `>= 38` condition lets us remove, I certainly think that there is value in doing this.
Test:
1. Build the Chrome extension and load it.
2. Visit https://robwu.nl/pdfjs/object-embed.html
3. Verify that all displayed blocks have the same width and
height as the reference ("Expected dimension").
In fonts with only upper-case glyphs, that are also missing a space glyph, `get spaceWidth` won't be able to return anything useful.
By adding upper-case `I` as a fallback, we can thus improve text-selection in some PDF files.
Note that locally, the patch causes slight movement in a few existing `text` tests, but in my opinion this actually looks like slight improvements.
Fixes 7180.
Currently the `isSpace` utility function is a member of `Lexer`, which seems suboptimal, given that it's placed in `core/parser.js`. In practice, this means that in a number of `core/*.js` files we thus have an *otherwise* completely unnecessary dependency on `core/parser.js` for a one-line function.
Instead, this patch moves `isSpace` into `shared/util.js` which seems more appropriate for this kind of utility function. Not to mention that since all the affected `core/*.js` files already depends on `shared/util.js`, this doesn't incur any more file dependencies.
Privacy policy: https://github.com/Rob--W/pdfjs-telemetry#privacy-policy
Unit tests (offline):
```
node test/chromium/test-telemetry.js
```
Server tests (requires that Nginx is installed):
```
git clone https://github.com/Rob--W/pdfjs-telemetry.git
cd pdfjs-telemetry/
python testserver.py TestHttp TestHttps
```
Integration test (extension + server):
- Build the extension
- Edit build/chromium/telemetry.js and remove the check for
chrome.runtime.id.
- Start Chrome (preferably a new profile):
chromium --user-data-dir=/tmp/pdftest --no-first-run
- Open chrome://net-internals#events
- Visit chrome://extensions and enable Developer mode.
- Load unpacked extension, select build/chromium.
- Go to the chrome://net-internals tab and filter on pdfjs.robwu.nl.
- Click on URL_REQUEST and verify that the server replied with 204.
- Reload the extension.
- Verify that chrome://net-internals did not contain a new log request.
Currently `setGState` is completely broken, and looking through the history of that code, it seems to me that this may never have worked correctly.
This patch fixes the text-selection in `extgstate.pdf` in the test-suite, which is also added as a `text` test.
Fixes http://www.pdf-archive.com/2013/09/30/file2/file2.pdf.
Note how it's not possible to show the various Popup Annotations in the above document.
To fix that, this patch lets the Popup inherit the flags of the parent, in the special case where the parent is `viewable` *and* the Popup is not.
In general, I don't think that a Popup must have the same flags set as the parent. However, it seems very strange to have a `viewable` parent annotation, and then not being able to view the Popup.
Annoyingly the PDF specification doesn't, as far as I can find, mention anything about how this case should be handled, but this patch seem consistent with the actual behaviour in Adobe Reader.
Use chrome.storage.sync to store preferences instead of
chrome.storage.local, to allow settings to be synchronized if the user
chooses to sign in in Chrome and enables synchronization of extension
preferences.
Commit https://github.com/mozilla/pdf.js/commit/df10513e103cc165352fb1ee2e2cf5122a73eee1 unfortunately broke the options dialog of the Chromium extension because the logic required to work with the preference was not added. This patch adds the required logic to show the preference in the options dialog and to persist it to the preferences storage.
Verified using Chromium 50 on Arch Linux.
Currently for explicit destinations, compared to named destinations, we manually try to build a hash that often times is a quite poor representation of the *actual* destination. (Currently this only, kind of, works for `\XYZ` destinations.)
For PDF files using explicit destinations, this can make it difficult/impossible to obtain a link to a specific section of the document through the URL.
Note that in practice most PDF files, especially newer ones, use named destinations and these are thus unnaffected by this patch.
This patch also fixes an existing issue in `PDFLinkService_getDestinationHash`, where a named destination consisting of only a number would not be handled correctly.
With the added, and already existing, type checks in place for destinations, I really don't think that this patch exposes any "sensitive" internal destination code not already accessible through normal hash parameters.
*Please note:* Just trying to improve the algorithm that generates the hash is unfortunately not possible in general, since there are a number of cases where it will simply never work well.
- First of all, note that `getDestinationHash` currently relies on the `_pagesRefCache`, hence it's possible that the hash returned is empty during e.g. ranged/streamed loading of a PDF file.
- Second of all, the currently computed hash is actually dependent on the document rotation. With named destinations, the fetched internal destination array is rotational invariant (as it should be), but this will not hold in general for the hash. We can easily avoid this issue by using a stringified destination array.
- Third of all, note that according to the PDF specification[1], `GoToR` destinations may actually contain explicit destination arrays. Since we cannot really construct a hash in `annotation.js`, we currently have no good way to support those. Even though this case seems *very* rare in practice (I've not actually seen such a PDF file), it's in the specification, and this patch allows us to support that for "free".
---
[1] http://www.adobe.com/content/dam/Adobe/en/devnet/acrobat/pdfs/PDF32000_2008.pdf#G11.1951685
Currently the `getPageIndex` method will happily return `0`, even if the `Ref` parameter doesn't actually point to a proper /Page dictionary.
Having the API trust that the consumer is doing the right thing seems error-prone, hence this patch which adds a check for this case.
Given that the `Catalog_getPageIndex` method isn't used in any hot part of the codebase, this extra check shouldn't be a problem.
(Note: in the standard viewer, it is only ever used from `PDFLinkService_navigateTo` if a destination needs to be resolved during document loading, which isn't common enough to be an issue IMHO.)
These have been found using `gulp lint` in combination with the `unused:
true` parameter for JSHint. Unfortunately there are too many false
positives to enable this feature, but now that most globals have been
removed because of the conversion to UMD the results are much more
useful than before.
This was only ever useful for the Opera extension because the API
requires a whitelisted extension ID. Opera ditched PDF.js from their
extension gallery, so we don't need to keep this in the tree.
I've seen the above error occasionally when the scale is updated many times in quick succession, but I've not been able to pinpoint exactly why it happens.
Since the error isn't caught, this means that the `pageViewDrawCallback` function doesn't run to completion.
Unfortunately, given the very intermittent nature of the issue, I haven't got any good STR for reliably reproducing this issue. However, I hope that this patch can be accepted anyway, since it's simple and should help prevent unnecessary errors.
In the font in question, there are a couple of `topDict` entries that have invalid values (`0xF 0xF`, i.e. just eof markers without any actual numbers).
This causes the `parseFloatOperand` function, inside `CFFParser_parseDict`, to return `NaN`. Currently we pass this broken font onto the browser, which OTS unsurprisingly rejects.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1068432.
As evident from e.g. PRs 6485 and 7118, some bad PDF generators unfortunately create Arrays where *some* elements are indirect objects (i.e. `Ref`s). This seems to mostly affect Arrays that contain numbers, such as e.g. `Matrix/FontMatrix/BBox/FontBBox/Rect/Color/...`, and has manifested itself in PDF files that fail to render correctly (some elements are missing).
The problem in both the cases above, besides broken rendering, was that there were *no* errors/warnings that indicated what the problem was, making it difficult to pinpoint the issue.
Hence this patch, where I've audited all usages of `Dict_get` in `src/core/` files, and replaced it with `Dict_getArray` where appropriate to try and prevent unnecessary future bugs.
Re: issue 7261.
Given the we have `gulp fonttest`, which tests the `fonts.js` functionality at a higher level, and that we have *a lot* of font specific reference tests, I'm not convinced that we *also* need unit-tests for it.
We're already, since quite some time, using the standard Fullscreen API provided that it's available in the browser. The warning is only caused by the code that checks if the Fullscreen API is supported.
This patch uses a simple preprocessor tag to avoid the warning, since I'm assuming that in general, we want to try and remain backwards compatible with the prefixed versions of the Fullscreen API.
Fixes 7270.
Multiple shadow roots are not supported any more in Chrome 51+
(https://crbug.com/603448#c6), so this patch changes the way that PDFs
are rendered in `<embed>` / `<object>` tags.
I used shadow roots because their content is not visible from the web
page, so the odds of conflicts were minimal. Now I have to render the
PDF frame directly in the page, which can be observed from the page
(unfortunately).
Now the following happens when an embedded PDF tag is detected:
- `<embed>` tags: The type and src attributes are updated.
- `<object>` tags: The type attribute is changed and the fallback
content is set and displayed.
When Firefox is run in e10s mode, which will soon be the default, the PDF.js zoom dropdown menu doesn't look right. This is apparently because the `<select>` DOM element is rendered in the parent, and that all the necessary style information isn't sent up from the child. See the discussion in https://bugzilla.mozilla.org/show_bug.cgi?id=910022.
Besides this causing the PDF.js UI to *look* worse in e10s, notably it also means that the `customScaleOption` isn't hidden like it ought to be.
To work-around that, this patch utilizes the `hidden` attribute, since https://bugzilla.mozilla.org/show_bug.cgi?id=1242450 at least made that work in e10s.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1194700.
This naming issue has been present since PR 3529, but at least I cannot find any issues/bugs that seem to have been caused by it, which is good.
The patch also removes an unnecessary `else` branch, since an already existing `break` means that it's redundant.
Fixes 7232.
Furthermore we introduce two new methods named `setCallback` and
`setReason` so external code does not change the properties of the class
directly. Finally we update various names of properties and methods to
be more self-explanatory.
With the recent PR 7172, which made the viewer modular, there's now a couple of modules that are no longer easily accessible (e.g. through the console).
This can make testing/debugging more difficult, and means that e.g. https://github.com/mozilla/pdf.js/wiki/Debugging-PDF.js#enabling no longer works in the generic viewer.
For now, as a simple solution, this patch just exposes those non-classes on `PDFViewerApplication` to ensure that they are available, and to avoid polluting the `window` scope.
Persist the state of content sidebar while browsing away from viewer and
initializing the same on returning back to the viewer. The state is saved
in persistent store preferences and used upon viewer initialization.
Fixes#6935
We cannot piggy-back on the `updateviewarea` event in order to update the stored sidebar state, since there're a number of cases where opening/switching the sidebar view won't fire a `updateviewarea` event.
Note that `updateviewarea` only fires when the position changes in the *viewer*, which means that it won't fire if e.g. the viewer is narrow, such that the sidebar overlays the document transparently; or when switching views, without the document position also changing.
This patch also moves the handling of `forceOpen` parameter in `PDFSidebar_switchView`, to prevent triggering back-to-back rendering and dispatching of events.
We currently don't have *any* unit-tests for `LinkAnnotation`s, so it seemed a good idea to add a few. These tests are taken from various actual PDF files.
- Add support for the 'NewWindow' property.
- Ensure that destinations are applied to the *remote* document, instead of the current one.
- Handle the `F` entry being a standard string, instead of a dictionary.
This patch also adds/improves utility functions for checking if the passwords are correct/incorrect, and replaces `string2binary` with `stringToBytes`. Finally the patch does away with the `DictMock`, in favour of using actual `Dict`s.
Re: issue 6905.
Using `new {Name,Cmd}` should be avoided, since it creates a new object on *every* call, whereas `{Name,Cmd}.get` uses caches to only create *one* object regardless of how many times they are called.
Most of these are found in the unit-tests, where increased memory usage probably doesn't matter very much. But it still seems good to get rid of those cases, since no part of the codebase ought to advertise that usage.
Given the small size of the patch, I'm also tweaking a few comments and class names.
This is a regression from PR 7097.
(Also, out of scope for this PR, but I think that a `setTimeout` value of `1000 ms` is too large. Switching from scrolling to zooming can fell sluggish, and give the impression that nothing happens.)
Previously the commands were not properly parsed as such by GitHub
because they need to be indented with four spaces.
Furthermore, address some minor textual nits.
In the API unit-tests, we're currently loading the `basicapi.pdf` before every sub-test in `PDFDocument` and `Page`, which slows down the unit-tests quite a bit.
Locally this patch reduces the run time for `gulp unittest` by at least 40% for me.
Some browsers render certain special characters with width 0, others with strictly positive width. (For example, the Greek Delta, Δ, has width 0 in Google Chrome, and a positive width in Firefox.) The `if` clause in operation so far results in different text layer DOM trees for different browsers.
This commit fixes that by adding the elements independently of their width.
Note that in the PDF files provided by the reporter, this issue was limited to `Rect` arrays in AcroForm entries (which we currently don't support).
However, since a bad PDF generator could create this problem in *any* kind of annotation, the reduced test-case included here uses a simple LinkAnnotation instead.
Fixes 7115.
According to "The table directory" under https://developer.apple.com/fonts/TrueType-Reference-Manual/RM06/Chap6.html#Directory, TrueType font tables should have `uint32` checksums.
This is something that I noticed, and was initially confused about, while debugging a TrueType issue.
As far as I can tell, the current (`int32`) checksums we use doesn't cause any issues in practice. However, I do think that this should be addressed to agree with the specification, and to reduce possible confusion when reading the font code.
This is regression from PR 7063, causing `gulp importl10n` to fail:
```
$ gulp importl10n
[19:45:02] Using gulpfile c:\Users\Jonas\Git\pdfjs\gulpfile.js
[19:45:02] Starting 'importl10n'...
Downloading ach...
[19:45:02] 'importl10n' errored after 4.42 ms
[19:45:02] Error: EEXIST, file already exists 'c:\Users\Jonas\Git\pdfjs\l10n\ach
'
at Error (native)
at Object.fs.mkdirSync (fs.js:747:18)
at downloadLanguageFiles (c:\Users\Jonas\Git\pdfjs\external\importL10n\local
es.js:59:8)
at next (c:\Users\Jonas\Git\pdfjs\external\importL10n\locales.js:90:5)
at Object.downloadL10n (c:\Users\Jonas\Git\pdfjs\external\importL10n\locales
.js:91:5)
at Gulp.<anonymous> (c:\Users\Jonas\Git\pdfjs\gulpfile.js:92:11)
at module.exports (c:\Users\Jonas\Git\pdfjs\node_modules\gulp\node_modules\o
rchestrator\lib\runTask.js:34:7)
at Gulp.Orchestrator._runTask (c:\Users\Jonas\Git\pdfjs\node_modules\gulp\no
de_modules\orchestrator\index.js:273:3)
at Gulp.Orchestrator._runStep (c:\Users\Jonas\Git\pdfjs\node_modules\gulp\no
de_modules\orchestrator\index.js:214:10)
at Gulp.Orchestrator.start (c:\Users\Jonas\Git\pdfjs\node_modules\gulp\node_
modules\orchestrator\index.js:134:8)
```
Pass the `PDFJS.postMessageTransfer` parameter to the worker, so that the `MessageHandler` can be setup correctly in `createDocumentHandler` (issue 6957)
Currently there's a lot of duplicate code for non-embedded `toFontChar`, which this patch simplifies by extracting the code into a helper function instead.
This patch adds a `getUnicodeForGlyph` helper function, which is used to recover Unicode values for non-standard glyph names.
Some PDF generators, e.g. Scribus PDF, use improper `uniXXXX` glyph names which breaks the glyph mapping. We can avoid this by converting them to "standard" glyph names instead.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1132849.
Fixes 6893.
Fixes 6894.
[PDFThumbnailView] Re-factor the `canvas` to `image` conversion such that we always render to a `canvas`, and then replace it with an `image` once rendering is done
In the PDF file in question, some of the 'name' table entries have `record.length === 0`. This becomes problematic in the non-unicode case, since `font.getBytes(0)` will fetch the *entire* stream.
Given that OTS rejects 'name' entries larger than `2^16`, this thus explain the sanitizer errors.
Fixes 7020.
Functionality wise, `ensureThumbnailVisible` is essentially just a shorthand for `scrollThumbnailIntoView`. (And note that `PDFViewer` doesn't implement a `ensurePageVisible` method.)
The only remaining usage of `PDFThumbnailViewer_ensureThumbnailVisible` is inside `PDFPresentationMode`, which introduces an otherwise unnecessary `PDFThumbnailViewer` dependency there.
We're already relying on the `presentationmodechanged` event, in various files, to track the state of Presentation Mode. Thus we can simply listen for that event in `PDFSidebar` too, and update the thumbnails if necessary.
*This is a follow-up to PRs 6299 and 6441.*
The patch also adds an option to `PDFThumbnailView`, that disables the canvas-to-image conversion entirely, which might be useful in the context of issue 7026.
The sidebar code has, except for minor fixes/additions (such as attachments), been largely untouch for years.
To avoid having a bunch of sidebar code sprinkled throughout viewer.js, this patch moves the sidebar code into a separate file (pdf_sidebar.js), similar to how most other functionality has been moved in the last few years.
Besides simply moving code around, this patch also has the added benefit that we now keep track of the sidebar state (not just opened/closed).
This now makes it possible to handle both `Preferences` *and* `ViewHistory` settings for the sidebar state in a cleaner way, preventing strange and confusing interactions between the two.
Changes `PDFOutlineView`/`PDFAttachmentView` to be initialized once, since we're always creating them, and refactor their `render` methods to instead pass in the `outline`/`attachments`.
For consistency with other "classes", the `PDFOutlineView`/`PDFAttachmentView` are renamed to `PDFOutlineViewer`/`PDFAttachmentViewer`.
Also, make sure that the outline/attachments are reset when the document is closed. Currently we keep the old ones around until the `getOutline`/`getAttachments` API calls are resolved for a new document.
Re: issue 7017.
This should prevent the error, but does not fix the broken rendering.
The rendering issues are caused by `svg.js` not supporting the various text rendering modes, in this case specifically "invisible" (as indicated in the console `Warning: Unimplemented method setTextRenderingMode`).
Compared to the canvas case, where we just ignore invisible text, a smarter solution is probably required for the SVG case. Since just ignoring invisible text in `svg.js` would mean that text-selection isn't possible.
To reduce code duplication, the initialization code now uses the `reset` method.
Also, this patch moves `charactersToNormalize` out of `PDFFindController`, since it seemed better suited to be a "constant".
Open http://www.puolustusvoimat.fi/wcm/61ba4180411e702ea19ee9e364705c96/luonnonmuonaohjelmalumo1985.pdf?MOD=AJPERES#pagemode=bookmarks.
Note how the outline looks entirely empty, but hovering over it you'll see that there are entries present. There's two separate issues here, first of all the fact that you cannot visually make out the outline items, and secondly that the lack of text means that the clickable area becomes too small.
In Adobe Reader this issue is somewhat mitigated, since they use an icon for every item. For PDF.js, the easist way to address this seems to be making use of a default title.
This issue should be *very* rare in practice, but given the simplicity of the solution I think that we should fix this.
The `hasImage` property is a left-over from older thumbnail code, and has been made obsolete by `renderingState`.
Having two different properties tracking (basically) the same state is asking for trouble, since it's very easy to forget to update one of them, with annoying bugs as the result.
This is required to be able to use it in the annotation display code,
where we now apply it to sanitize the filename of the FileAttachment
annotation. The PDF file from https://bugzilla.mozilla.org/show_bug.cgi?id=1230933 has shown that some PDF generators include the path of the file rather than the filename, which causes filenames with weird initial characters. PDF viewers handle this differently (for example Foxit Reader just replaces forward slashes with spaces), but we think it's better to only show the filename as intended.
Additionally we add unit tests for the `getFilenameFromUrl` helper
function.
*A more robust solution for issue 6066.*
As a temporary work-around for (the upstream) [bug 1164199](https://bugzilla.mozilla.org/show_bug.cgi?id=1164199), we parsed *all* images in the Firefox addon during a short time.
Doing so uncovered an issue with our image handling (see 6066), for JPEG images with a `DeviceGray` ColorSpace *and* `bpc !== 1` (bits per component).
As long as we let the browser handle image decoding in this case, this isn't going to be an issue, but I do think that we should proactively fix this to avoid future issues if we change where the images are decoded (in `jpg.js` vs in browser).
Also, we currently don't seem to have a test-case for that kind of image data.
Currently the `C` entry in an outline item is returned as is, which is neither particularly useful nor what the API documentation claims.
This patch also adds unit-tests for both the color handling, and the `F` entry (bold/italic flags).
`Dict_getAll` is problematic for a number of reasons. First of all, as issue 6961 shows, it can be really bad for performance, since it dereferences all indirect objects.
Second of all, all the derefencing can lead to data being unncessarily requested when ranged/chunked loading is used, thus unnecessarily delaying rendering.
Note: For cases where `Dict_getAll` was previously used, `Dict_getKeys` in combination with `Dict_get` can be used instead. This has the advantage that data isn't requested until it's actually needed.
For the operators that we currently support, the arguments are not `Dict`s, which means that it's not really necessary to use `Dict_getAll` in `EvaluatorPreprocessor_read`.
Also, I do think that if/when we support operators that use `Dict`s as arguments, that should be dealt with in the corresponding `case` in `PartialEvaluator_getOperatorList` which handles the operator.
The only reason that I can find for using `Dict_getAll` like that, is that prior to PR 6550 we would just append certain (currently unsupported) operators without doing any further processing/checking. But as issue 6549 showed, that can lead to issues in practice, which is why it was changed.
In an effort to prevent possible issue with unsupported operators, this patch simply ignores operators with `Dict` arguments in `PartialEvaluator_getOperatorList`.
For the `CalGray`/`CalRGB`/`Lab` colour spaces, we're currently using `getAll` to retrieve the parameters. However that's not really necessary, since we may just as well explicitly `get` the needed parameters instead.
Some bad PDF generators, in particular "Scribus PDF", duplicates resources *a lot* at various levels of the PDF files. This can lead to `PartialEvaluator_hasBlendModes` taking an unreasonable amount of time to complete.
The reason is that the current code is using `Dict_getAll`, which recursively dereferences *all* indirect objects, which can be really slow. This patch instead uses `Dict_getKeys`, and then manually looks up only the necessary indirect objects.
I've added the PDF file as a `load` test. The most important thing here is probably to ensure that the file remains available in the repo, and the comment should help reduced the chance of regressions. (Note that locally, the `load` test times out without this patch, but we cannot really assume that that always happens.)
Fixes 6961.
The Chrome extension enforces that local files cannot be embedded in
non-local web pages. The previous check was too strict (because the
origin of a file:-URL is "null"), and prevented local PDF from being
viewed in local files).
This patch fixes that problem, by querying the actual tab URL via the
background page.
Steps to verify:
1. Create a HTML file: `<iframe src=test.pdf width=100% height=100%>`
2. Build and load the extension.
3. Allow file access to the extension at `chrome://extensions`
4. Open the HTML file from a file:// URL.
5. VERIFY: The extension should attempt to load the PDF file.
6. Now open the following (replace ID with the extension ID, which you
can find at `chrome://extensions`):
`data:text/html,<iframe src="chrome-extension://ID/file:///test.pdf">`
7. VERIFY: The next error should be displayed:
"Refused to load a local file in a non-local page for security reasons."
Reverts "Hack to avoid intermidiate Chrome failures during tests."
(2b2c521213).
require.js uses importScript asynchronously, which activates the worker
GC bug in WebKit. This patch works around a bug in a way that is similar
in the upcoming (but not yet released) require.js 2.1.23
The advantage of the new work-around is that it allows the runtime to
garbage-collect idle Workers.
References:
- https://crbug.com/572225
- https://webkit.org/b/153317
*This patch is based on something I noticed while debugging some of the PDF files in issue 6931.*
In a number of the cases in `setGState`, we're implicitly assuming that we're not dealing with indirect objects (i.e. `Ref`s). See e.g. the 'Font' case, or the various cases where we simply do `gStateObj.push([key, value]);` (since the code in `canvas.js` won't be able to deal with a `Ref` for those cases).
The reason that I didn't use `Dict_forEach` instead, is that it would re-introduce the unncessary closures that PR 5205 removed.
The intention of PR 5192 was to avoid adding empty `setGState` ops to the operatorList. But the patch accidentally used `>=`, which means that it's not actually working as intended, since empty arrays always have `length === 0`.
Even though the currently known test-cases render correctly without this patch, that seems more like a lucky coincidence, given that there's no guarantee that `transferMap[255] === 0` for every possible transfer function.
This patch fixes an issue that I inadvertently introduced in PR 5815, where we accidentally modify the `Differences` array in the encoding dictionary for indirect objects.
Instead of this change, we could also have used the now existing `Dict_getArray`. However in this case I don't think that would have been a good idea, since it would mean iterating through the array *twice*.
Prevent failures in the "scanning for endstream" code, in `Parser_makeStream`, by handling the case where 'endstream' is split between contiguous chunks (issue 1536)
Currently the `deprecated` message is using `warn`, meaning that it's possible to disable warnings about deprecated API usage through the `PDFJS.verbosity` setting.
I don't think that it should be possible to opt out of deprecation messages,[1] since it might mean that in a custom deployment of PDF.js these messages could be overlooked, leading to PDF.js being broken (seemingly without any warning) when updating to a future version.
Obviously this could be considered the responsibility of the people doing custom PDF.js implementations, but in order to reduce the support burden later on, it seems better to "annoy" people upfront.
Compared to various `info`/`warn`/`error` messages, `deprecated` messages should be very simple to get rid of -- just update the API usage and the message goes away!
---
[1] In e.g. Firefox it doesn't seem possible to prevent deprecation warnings from being displayed (in the Browser Console).
Fixes 6898.
Note that this doesn't prevent the warning for PDF files that *do* ask for a password, e.g. http://async5.org/moz/passwordOU.pdf, but it's not clear to me if/how we could avoid that.
Re: issue 5089.
(Note that since there are other outline features that we currently don't support, e.g. bold/italic text and custom colours, I thus think we can keep the referenced issue open.)
As part of the link cleanup in issue 6854, obtaining this file through the Internet Archive didn't work.
However, given that the file was added in order to test an issue with `CropBox/MediaBox`, a reduced test-case should do just fine instead.
Please refer to issue 1155, and PR 1212.
It seems to be fairly common for OCR software to include incomplete TrueType fonts, notable missing the "glyf" table, in PDF files. Since we currently reject such fonts, the result is that text-selection/copying is broken.
This patch contains a suggested approach to try and use these kind of broken fonts, by using existing code in `sanitizeGlyphLocations` to replace a missing "glyf" table with dummy data.
Fixes 4684.
Fixes 6007.
Fixes 6829.
The Firefox addon currently fails with:
```
SyntaxError: missing ; before statement pdf.js:1692:12
TypeError: PDFJS.shadow is not a function viewer.js:6228:12
```
The test case was changed in https://github.com/mozilla/pdf.js/commit/1faca190216deed1ef6084697d04a1d90621e210 because the original file was not available anymore. However, its hash was also changed, meaning that we do not test the intended version anymore.
This patch makes sure that we test the intented version by reverting to
the original hash and using a link, also pointing to the Internet
Archive, with the original file.
This test was disabled in PR 4732, because the file was no longer available. The motivation being that there were two other files which should be good replacements. However, one of those has since been replaced with a reduced test-case (which doesn't exercise the same code-path), and in the other one the error does not appear to be entirely identical.
Hence it seems reasonable to re-add the 'aboutstacks.pdf' test, since it was possible to find it on the Internet Archive (by searching using a different URL, compared to the current one).
Issue 6854 - Update the link locations to point to the Internet Archive for tests: 20130226130259, JST2007-5, P020121130574743273239, SFAA_Japanese, TaroUTR50SortedList112
Note that despite the new file having a different hash than the the current one, it does render *identically* and most importantly it uses *the same* JBIG2 functionality.
For reference, please see issue 3666 and PR 3738.
*This patch follows a similar idea as PR 5756.*
The patch is based on the nice debugging done by Brendan in the referenced issue 6782.
A better way to handle this, and similar issues, would probably be to completely ignore what the PDF file claims about font type/subtype, and just check the actual data. But until that kind of rewrite happens, this patch should help.
Fixes 6782.
When generating new references locally on Windows, after PR 6724, I get the following output:
```
WARNING: Unable to open file for reading "Error: ENOENT, open 'c:\Users\Jonas\Git\pdfjs\test\pdfs\issue_3694_reduced.pdf'".
Unable to verify the checksum for the files that are used for testing.
Please re-download the files, or adjust the MD5 checksum in the manifest for the files listed above.
```
Compared to the name of the file (`issue3694_reduced.pdf`), you see that the manifest entry has a superfluous underscore in the "file" entry.
Apparently some PDF files can have annotations with `URI` entries ending with `null` characters, thus breaking the links.
To handle this edge-case of bad PDFs, this patch moves the already existing utility function from `ui_utils.js` into `util.js`, in order to fix those URLs.
Fixes 6832.
Currently we're not applying Patterns for text, but only for graphics.
This patch is unfortunately not a complete solution, but rather a step on the way, since there are still some PDF files where the Patterns look more like a solid colour, rather than the intended gradient.
I've been unable to fix these issues completely, and I've not managed to determine if the remaining issues are caused either by the pattern code, the canvas code, or perhaps both.
However, given that even this simple patch improves the current situation quite a bit, I figured that it couldn't hurt to submit it as-is.
- Fixes 5804.
- Fixes 6130.
- Improves 3988 a lot, since the text is now visible. However, it looks like the text is *one* solid colour, instead of the correct gradient.
- Improves 5432, since the text is no longer gray. (This file also suffers from the same problem as the previous one.)
Most code for Popup annotations is already present for Text annotations.
This patch extracts the popup creation logic from the Text annotation
code so it can be reused for Popup annotations.
Not only does this add support for Popup annotations, the Text
annotation code is also considerably easier. If a `Popup` entry is
available for a Text annotation, it will not be more than an image. The
popup will be handled by the Popup annotation. However, it is also
possible for Text annotations to not have a separate Popup annotation,
in which case the Text annotation handles the popup creation itself.
This PDF file (see issue 4914) originally regressed in PR 4318, and was subsequently fixed in PR 4915.
I added the PDF file as a (linked) test-case in PR 6481, in an effort to prevent regressions. Since we at that time didn't have the necessary framework in place, in order to correctly test annotations, this almost regressed *again* in PR https://github.com/mozilla/pdf.js/pull/6672#issuecomment-158689392.
In that PDF file, some of the annotations are both printable and hidden, and should definitely *not* be visible on normal display. Hence this patch, which adds the `annotations` flag to the manifest in order to ensure that those annotations won't be rendered when `intent === 'display'`.
This patch makes the naming consistent with the `TextLayerBuilder`, and also the new `AnnotationLayer`, and should thus help reduce possible confusion when working with the code.
Please note that the files were renamed using `git mv`, in order to preserve blame.
Now we have a full list of all possible annotation types and the
numbering corresponds to the order in the specification. Not only is
this more consistent and complete, it also prevents having to add these
constants when a new annotation type is implemented.
Additionally fix an issue where a regular Widget annotation would not
have `data.annotationType` set. It was only set for a
TextWidgetAnnotation, but instead move it to the base Widget annotation
class to add it for all Widget annotations (since TextWidgetAnnotation
inherits from WidgetAnnotation it will have it too).
[Firefox] Prevent internal links from displaying "resource://pdf.js/web/" on hover, by tweaking the fallback case in `PDFLinkService_getDestinationHash`
Additionally simplify the div creation logic (it needs to happen only
once, so it should not be in the for-loop) and remove/rename variables
for shorter code.
In `Font_checkAndRepair` we can decide that a font isn't TrueType, and instead parse it as CFF. In that case it's quite possible that the `fontMatrix` will be changed, and without calling `adjustWidths` we're failing to update the glyph widths correctly.
Fixes 5027.
Fixes 5084.
Fixes 6556.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1204903.
After PR 6590, `font.spaceWidth` is now called in more cases than before (in `PartialEvaluator_getTextContent`), which exposed an underlying issue with `IdentityToUnicodeMap_charCodeOf` throwing an error.
This breaks text-selection in some PDF files found in the wild, hence this patch replaces the `error` with an actual function instead (modelled after `IdentityCMap_charCodeOf`).
This patch improves the code structure of the annotation code.
- Create the annotation border style object in the `setBorderStyle` method instead of in the constructor. The behavior is the same as the `setBorderStyle` method is always called, thus a border style object is still always available.
- Put all data object manipulation lines in one block in the constructor. This improves readability and maintainability as it is more visible which properties are exposed.
- Simplify `appendToOperatorList` by removing the promise capability and removing an unused parameter.
- Remove some unnecessary newlines/spaces.
*This is a regression from PR 3424.*
The PDF file in the referenced issue is using `Type3` fonts. In one of those, the `/CharProcs` dictionary contains an entry with the name `/#`. Before the changes to `Lexer_getName` in PR 3424, we were allowing certain invalid `Name` patterns containing the NUMBER SIGN (#).
It's unfortunate that this has been broken for close to two and a half years before the bug surfaced, but it should at least indicate that this is not a widespread issue.
Fixes 6692.
[api-minor] Add a parameter to `PDFPageProxy_getTextContent` that enables replacing of all whitespace with standard spaces in the textLayer (issue 6612)
This patch goes a bit further than issue 6612 requires, and replaces all kinds of whitespace with standard spaces.
When testing this locally, it actually seemed to slightly improve two existing test-cases (`tracemonkey-text` and `taro-text`).
Fixes 6612.
This patch:
- Updates the JSDoc comment in `api.js`, to more clearly point out that `PDFJS.openExternalLinksInNewWindow` is deprecated, and explains what to use instead.
- Changes the `warn`, in `isExternalLinkTargetSet()`, to use the new `deprecated` function instead. Also updates the message with more detailed information about what to use instead.
- Changes the pre-processor tag to ensure the deprecation warning is seen in all build types where it could possibly matter (in case people are using `PDFJS.openExternalLinksInNewWindow` in e.g. custom-built extensions).
These changes were prompted by seeing http://stackoverflow.com/questions/33813373/pdf-js-how-to-open-hyperlinks-in-a-new-tab-window, since it seems to me that the current comments/warnings might not be worded well enough.
Currently `getAnnotations` will *only* fetch annotations that are either `viewable` or `printable`. This is "hidden" inside the `core.js` file, meaning that API consumers might be confused as to why they are not recieving *all* the annotations present for a page.
I thus think that the API should, by default, return *all* available annotations unless specifically told otherwise. In e.g. the default viewer, we obviously only want to display annotations that are `viewable`, hence this patch adds an `intent` parameter to `getAnnotations` that makes it possible to decide if only `viewable` or `printable` annotations should be fetched.
While browsing through the latest PDF.js update on mozilla-central, see https://hg.mozilla.org/integration/fx-team/rev/aef06cd725fc, I noticed that the `'use strict';` directives were missing at the top of a number of files.
This is fallout from the changes made in `make.js` in PR 6627, since `stripCommentHeaders` previously relied on the existence of the mode-lines.
I'm assuming that we do want *all* of the code (e.g. the viewer too) to execute in strict mode, hence this patch tweaks `stripCommentHeaders` to make it less greedy.
This patch makes it possible to set and get all possible flags that the PDF specification defines. Even though we do not support all possible annotation types and not all possible annotation flags yet, this general framework makes it easy to access all flags for each annotation such that annotation type implementations can use this information.
We add constants for all possible annotation flags such that we do not need to hardcode the flags in the code anymore. The `isViewable()` and `isPrintable()` methods are now easier to read. Additionally, unit tests have been added to ensure correct behavior.
This is another part of #5218.
In my opinion, this event shouldn't be dispatched in `PDFPageView.cssTransform`, since that would cause *two* events to be fired when "normal" zooming is used (once in `PDFPageView.cssTransform`, and once in `PDFPageView.draw`).
Fixes 6463.
- replace // with /
(otherwise http://localhost:8888// links to e.g. http://src/ instead
of http://localhost:8888/src).
- Solve XSS issue (file names should be sanitized, not output as-is).
- Prevent server from crashing if there is a stat error (e.g. permission
error or file not found (e.g. broken symlink)).
For documents with different page sizes, ensure that the correct page becomes visible on load when e.g. the 'page' hash parameter was provided (bug 1191279, issue 6393)
I received multiple reports about the following cryptic error in the
Chrome extension when the user tried to open a local file:
> PDF.js v1.1.527 (build: 2096a2a)
> Message: Cannot read property 'Symbol(Symbol.iterator)' of null
This error most likely originated from core/stream.js:
function Stream(arrayBuffer, start, length, dict) {
this.bytes = (arrayBuffer instanceof Uint8Array ?
arrayBuffer : new Uint8Array(arrayBuffer));
^^^^^^^^^^^
`arrayBuffer` is `null`, and that in turn is caused by the fact that
for non-existing files, there is no data. I've applied two fixes:
1. Never call onDone with a void buffer, but call the error handler
instead.
2. Show a sensible error message for local files with status = 0.
This is a quick fix to get `scrollIntoView` working again for search results, but we really ought to re-write that function since it seems way too brittle.
Fixes 6596.
This patch contains an alternative solution (see the referenced PRs in 4718), in order for the indeterminate loading bar to have a consistent speed regardless of width.
Fixes 4718.
Currently this constant is present in `viewer.js`, but it is not used there at all. Instead, it is used in `view_history.js` where we have a global for it. We might as well move the constant to `view_history.js` as that is the only place where it is used, thereby removing a global and an unused constant from `viewer.js`.
If your Internet Explorer 11 default compatibility settings are set to "I.E 7 Compatibility", the PDF plugin will not load. This fix is the same one used by AngularJS to force the browser to use HTML5 mode.
When I submitted PR 3576, I included a linked test-case. The reason was that I didn't know enough about the PDF format, in order to successfully create a reduced test-case.
Considering that the link points to a Dropbox, there's no guarantee that the PDF file will remain available, hence it seems worthwhile to replace the test-case.
*Note:* Since this is a `load` test, `makeref` won't be necessary.
In the `RenderPageRequest` handler in `worker.js`, we attempt to print an `info` message containing the rendering time and the length of the operator list. The latter is currently broken (and has been for quite some time), since the `length` of an `OperatorList` is reset when flushing occurs.
This patch attempts to rectify this, by adding a getter which keeps track of the total length.
*Follow-up to PR 6546.*
If rendering has already started when the document is destroyed, then `this.pageCache[data.pageIndex]` may already have been cleared when the `StartRenderingPage`/`RenderPageChunk` messages are recieved in `api.js`, which results in `TypeError`s being thrown.
We should never actually try to execute `paintXObject` in canvas.js, but in some cases where we fail to parse the PDF file correctly it can happen. Currently this will potentially cause an entire page to fail to render, which seems suboptimal.
With this patch, we will instead continue rendering with a warning that things might not work correctly.
`operatorList.addOp` adds the arguments to the list which is then
passed as-is by postMessage to the main thread. But since we don't
parse these operations, they are raw PDF objects and may therefore
cause a serialization error.
This is a conservative patch, and only affects operators which are
known to be unsupported. We should ignore all unknown operators,
but I haven't really looked into the consequences of doing that.
Fixes#6549
The standard viewer implicitly depends on the `onPassword` and `onProgress` callbacks, in order to open password protected PDF files, respectively to report file loading progress. We currently have no unit-tests for this functionality, which seems unfortunate; hence this patch.
*Please note:* Rather than adding more unit-tests to `api_spec.js`, I slightly extended/reworked two existing tests. Specifically for the password test, this *does not* really change what we actually test, just how the test is done.
In PR 6485 I somehow missed to account for the case where `xref` is undefined. Since a dictonary can be initialized without providing a reference to an `xref` instance, `Dict_getArray` can thus fail without this added check.
Replaces `PDFJS.openExternalLinksInNewWindow` with a more generic configuration option.
*Note:* `PDFJS.openExternalLinksInNewWindow = true;` is equal to `PDFJS.externalLinkTarget = PDFJS.LinkTarget.BLANK;`.
*This regressed in PR 5356.*
Rather than just backing out the offending code, this patch restores scrolling in PresentationMode by making the `overflow: hidden;` check optional and letting the callers that need it (e.g. `PDFFindController`) opt-in to use it.
One of the patches in https://bugzilla.mozilla.org/show_bug.cgi?id=1202902, specifically [`Mass replace toplevel 'let' with 'var' in preparation for global lexical scope. (rs=jorendorff)`](https://hg.mozilla.org/mozilla-central/rev/380817d573cd), touches PDF.js code. Unfortunately it was landed upstream without, as far as I can tell, notifying us about it.
This patch uplifts the relevant changes to avoid future merge conflicts, and for consistency also tweaks `PdfJs-stub.jsm`.
When the user edits the URL and changes the reference fragment (hash),
PDF.js intercepts this action, and saves the then-current history state
in the previous history entry. This is implemented by navigating back,
editing the history and navigating forward again.
The current logic has a flaw: It assumes that calling history.back() and
history.forward() immediately updates the history state. This is however
not guaranteed by the web standards, which states that calling e.g.
history.back "must traverse the history by a delta -1", which means that
the browser must QUEUE a task to traverse the session history, per spec:
http://w3.org/TR/2011/WD-html5-20110113/history.html#dom-history-backhttps://html.spec.whatwg.org/multipage/browsers.html#dom-history-back
Firefox and Internet Explorer deviate from the standards by immediately
changing the history state instead of queuing the navigation.
WebKit derived browsers (Chrome, Opera, Safari) and Opera presto do not.
The user-visible consequence of strictly adhering to the standards in
PDF.js can be shown as follows:
1. Edit the URL.
2. Append #page=2 for example.
3. Press Enter.
-> Presto and WebKit: PDF.js reverts to the previous URL.
-> Gecko and Trident: PDF.js keeps the new URL, as expected.
To fix the issue, modification of the previous history item happens in
a few asynchronous steps, guided by the popstate event to detect when
the history navigation request has been committed.
--
Some more implementation notes:
I have removed the preventDefault and stopPropagation calls, because
popstate is not cancelable, and window is already the last target of the
event propagation.
The previous allowHashChange logic was hard to follow, because it did
not explain that hashchange will be called twice; once during the
popstate handler for history.back() (which will reset allowHashChange),
and again for history.forward() (where allowHashChange will be false).
The purpose of allowHashChange is now more explicit, by incorporating
the logic in the replacePreviousHistoryState helper function.
*This regressed in PR 4920.*
The main motivation for PR 4920 was to quickly get rid of old canvases when pages are evicted from the `PDFPageViewBuffer` cache. However it inadvertently broke the use-case where the `canvas` is used as a preview, on scale or rotation changes, until the re-rendering is finished.
Fixes 6467.
In PR 5552, specifically commit https://github.com/mozilla/pdf.js/commit/9f384bbb41f780b258d832d1c7170b570aedac95, the meaning of `this.annotationLayer` changed in `PDFPageView`. Previously it referred directly to a DOM element, but now it's instead an instance of `AnnotationsLayerBuilder`.
This patch tweaks things so that we won't try to hide a non-existent `annotationLayer` div in `PDFPageView_reset`, and also so that we don't attempt to insert empty (`null`) DOM elements in `PDFPageView_draw`.
The file (`lshort.pdf`) has changed a couple of times since the test was added, hence there's no guarantee that the current version accurately reflects the issues the test was added to check.
In this patch, I'm updating the link location to point to the *intended* file version (hosted on the "Internet Archive").
According to the PDF spec 5.3.2, a positive value means in horizontal,
that the next glyph is further to the left (so narrower), and in
vertical that it is further down (so wider).
This change fixes the way PDF.js has interpreted the value.
*Follow-up to PR 6299.*
This patch reduces unnecessary code duplication for the `canvas` to `image` conversion. It also does a bit of re-ordering (and adds new lines) in `_getPageDrawContext`, since that function currently is a bit hard to read.
Since some browsers render `null` characters, and others don't, this patch adds a way to remove them to prevent display issues in the viewer UI.
Given that documents may contain very long outlines, I've added a utility function to avoid creating a lot of unnecessary `RegExp` objects.
To avoid any future issues, this utility function is used for both the outline and the attachments.
Fixes 6416.
This patch adjusts `get fingerprint` to also check that the `/ID` entry contains (non-empty) strings, to prevent more possible failures when loading corrupt PDF files (follow-up to PR 5602).
Note that I've not actually encountered such a PDF file in the wild. However given that `stringToBytes` will assert that the input is a string, and that we'll thus fail to load a document unless `get fingerprint` succeeds, making this more robust seems like a good idea to me.
For (1, 0) cmaps, we have two different codepaths depending on whether the font has/hasn't got an encoding. But with (3, 1) cmaps we don't have a good fallback when the encoding is missing, hence this patch changes `readCmapTable` to only choose a (3, 1) cmap table if the font is non-symbolic *and* an encoding exists. Without this, we'll not be able to successfully create a working glyph map for some TrueType fonts with (3, 1) cmap tables.
Fixes 6410.
Prior to PR 6242, the width of all outline items was set such that their right (or left, in RTL locales) edges lined up vertically. In my opinion that looked more consistent, therefore this patch adjusts the CSS to make sure that this will be the case again.
The patch also makes the `border-radius` values of outline items a bit more consistent.
This issue is actually, in a sense, "caused" by the fact that the API/viewer supports partial loading/rendering. Previously when the *entire* document was always fetched before rendering begun, we knew all page sizes in advance and this issue didn't exist.
Now we use the size of *one* page in order to set the initial size of every page, until we've fetched the pages and thus know their correct sizes.
This means that during loading the size of the pages can change, which may cause the initial position to become scrolled out of view.
The most naive solution to this problem would perhaps be to delay setting the initial position on load for all documents, until all pages are fetched. However I think that would be a *really* bad idea, since doing so would make the initial rendering slower and make it feel sluggish for most documents.
Since there is generally no way of knowing if a document has different sized pages prior to loading it, we can only check once the pages are available.
Hence this patch, which treats documents with different sized pages as a special case, by re-applying the initial position when all pages have become available.
This code was added in PR 1214, but was made obsolete by PRs 1488/1493. Prior to the latter ones, `Dict_get` retured the raw objects. However, afterwards (and currently) `Dict_get` now resolves indirect objects, which makes `Parser_fetchIfRef` redundant.
*Potential risks with this patch:*
This patch passes all tests locally, but there's a *small* possibility that it could break some weird PDF files.
In the current code, wrapping `Dict_get` inside `Parser_fetchIfRef` will potentially mean two back-to-back call of `XRef_fetch`, if a reference points directly to another reference. I'm not sure if this can actually happen in practice, and I'd think that if that were the case we'd already have run into it elsewhere in the code-base, given that `Parser` is the only place where we try to "double" resolve references.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1200096.
The problematic font has a `format 2` cmap, which we've never supported properly. Prior to PR 2606, we were able to fallback to a working state, despite not having proper support for that cmap format.
Obviously the best/correct solution would be to implement actual support for more cmap formats[1]. However, I'm hoping that a simple patch will be OK for now, given that:
- `format 2` cmaps seem to be quite rare in practice, since this has been broken for 2.5 years before anyone noticed.
- Having a simple patch will make potential uplifts a lot easier.
[1] See the specification at https://developer.apple.com/fonts/TrueType-Reference-Manual/RM06/Chap6cmap.html
Re: PR 4731.
Since the URL points to the Internet Archive, I think that adding a linked test-case should be OK. (Also, it's difficult to create reduced, or even `unit`, tests that accurately captures the brokenness of real-world PDF files.)
*Please note:* Since this is a `load` test, `makeref`ing won't be needed.
Our current test coverage for vertical text is somewhat lacking, as evident from e.g. issue 6387. That regression could easily have been avoided if the `taro` test-case would have been an `eq` test, as well as an `text` test.
Currently we only attempt to test the *first* page, since `lastPage == 1`, but given that it's subsequently skipped we end up not testing anything.
*Note:* I've verified that the *second* page actually contains the kind of Colour space that the test was intended to check for.
It appears to me that the `expanded/collapsed` part of the names got switched around, since I'd expect the following convention:
- `v` == expanded
- `>` == collapsed
Having a warning here would have meant that issue 6360 could have been solved in approximately five minutes, instead of an hour. To avoid that happening again, this patch adds a warning whenever we treat a stream as empty.
- This commit adds a '>' before every outline item that has subitems.
- Click on the '>' to collapse all subitems under that item (which turns
the '>' in a 'v').
- Shift + click expands/collapses all descendant items under that tree.
- Double-clicking on the "Show Document Outline" button in the toolbar
expands/collapses all outline items.
This patch improves the detection of `xref` in files where it is followed by an arbitrary whitespace character (not just a line-breaking char).
It also adds a check for missing whitespace, e.g. `1 0 obj<<`, to speed up `readToken` for the PDF file in the referenced issue.
Finally, the patch also replaces a bunch of magic numbers with suitably named constants.
Fixes 5752.
Also improves 6243, but there are still issues.
The problem with the PDF files in the issue, besides the obviously broken XRef tables which we're able to recover from, is that many/most of the streams have Dictionaries where the `Length` entry is set to `0`. This causes us to return `NullStream`, instead of the appropriate one in `Parser_makeFilter`.
Fixes 6360.
The JSON file is generated as follows.
1. Go to the src/chrome/app/resources directory of Chromium's source.
2. Find the translation ID of the "Allow access to file URLs" string:
grep 'Allow access to file URLs' generated_resources_en-GB.xtb
3. With the ID that you've found, locate the other translations.
grep 3341703758641437857 generated_resources_*.xtb
4. If the result looks OK, serialize the result as JSON and save it.
> path/to/pdf.js/web/chrome-i18n-allow-access-to-file-urls.json \
python -c "import json;print(json.dumps({ \
$(grep 3341703758641437857 generated_resources_*.xtb | \
sed "s@generated_resources_\([^.]\+\)\.xtb:<translation[^>]\+>\(.\+\)</translation>@'\1':'''\2''',@" \
)}, sort_keys=True, indent=2))"
(Strings are taken from Chromium 45.0.2448.0 (ccrev.com/337313).
In some cases, such as in use with a CSP header, constructing a function with a
string of javascript is not allowed. However, compiling the various commands
that need to be done on the canvas element is faster than interpreting them.
This patch changes the font renderer to instead emit commands that are compiled
by the font loader. If, during compilation, we receive an EvalError, we instead
interpret them.
Serialize errors to make sure that the callback is still invoked when
an error is thrown.
Firefox:
"DataCloneError: The object could not be cloned."
Chrome:
"DataCloneError: Failed to execute 'postMessage' on 'WorkerGlobalScope': An object could not be cloned."
CMaps may be sparse. Array.prototype.forEach is terribly slow in Chrome
(and also in Firefox) when the sparse array contains a key with a high
value. E.g.
console.time('forEach sparse')
var a = [];
a[0xFFFFFF] = 1;
a.forEach(function(){});
console.timeEnd('forEach sparse');
// Chrome: 2890ms
// Firefox: 1345ms
Switching to CMap.prototype.forEach, which is optimized for such
scenarios fixes the problem.
Currently in `switchSidebarView` there is code that attempts to handle the `outline` or `attachments` being disabled. However, given the placement of it, that code does not actually accomplish anything. Even more strange is the fact that the way that the buttons are disabled, that code won't ever be hit.
(Looking back with `git blame`, it seems that it has never worked as you'd expect.)
Hence this patch, which correctly disables switching to the `outline` or `attachments` views when they are disabled.
Once PR 6314 restores the 'pagemode' hash parameter, this patch thus makes it impossible to switch to a disabled sidebar view by modifying the 'pagemode' hash of an already loaded document.
This regressed in https://github.com/mozilla/pdf.js/commit/0ef6212b646427604b6968b8bc3cf2d6c048f012.
Since the 'pagemode' hash parameter requires certain viewer functionality (e.g. thumbnails and an outline) in order to work, it seemed reasonable to move the functionality from `pdf_link_service.js` into `viewer.js`.
Similar to `namedaction`, this patch makes use of an event to forward the 'pagemode' parameter.
pi is an index in the stream and is explained on page 201 of the 32000-spec (however 1-based there), and ps is an index to something in PDF.js. I used the code from flag 0 (which works) to understand which is which. It is also important to understand that for flags 1,2 and 3, the stream is always assigned to the same coordinates and colors. What changes is which "old" coordinates and colors are assigned to what is "missing" in the stream. This is why for these flags, the code is identical except for the assignments in the first "row". (Same principle as in #6304). Note that this change will not improve the lamp_cairo.pdf file, only the two files mentioned in #6305.
Short story: somebody got lost in two different indices. pi is an index in the stream and is explained on page 198 of the 32000-spec (however 1-based there), and ps is an index to something in PDF.js. I used the code from flag 0 (which works) to understand which is which. It is also important to understand that for flags 1,2 and 3, the stream is always assigned to the same coordinates and colors. What changes is which "old" coordinates and colors are assigned to what is "missing" in the stream. This is why for these flags, the code is identical except for the assignments in the first "row".
This is intended as a temporary solution, in order to get the "simpleviewer" example to work, until we've re-factored `scrollIntoView` to work in both the standard and components-based viewers.
Currently, `src/core/core.js` uses the `fromRef` method on an `Annotation` object to obtain the right annotation type object (such as `LinkAnnotation` or `TextAnnotation`). That method in turn uses a method `getConstructor` to find out which annotation type object must be returned.
Aside from the fact that there is currently a lot of code to achieve this, these methods should not be part of the base `Annotation` class at all. Creation of annotation object should be done by a factory (as also recommended by @yurydelendik at https://github.com/mozilla/pdf.js/pull/5218#issuecomment-52779659) that handles finding out the correct annotation type object and returning it. This patch implements this separation of concerns.
Doing this allows us to also simplify the code quite a bit and to make it more readable. Additionally, we are now able to get rid of the hardcoded array of supported annotation types. The factory takes care of checking the annotation types and falls back to returning the base annotation type (and issuing a warning, which the current code also does not do well) when an annotation type is unsupported.
I have manually tested this commit with 20 test PDFs with different annotation types, such as /Link, /Text, /Widget, /FileAttachment and /FreeText. All render identically before and after the patch, and unsupported annotation types are now properly indicated with a warning in the console.
Built-in DOM properties are slower than plain JS properties.
A few lines before, textContent is assigned as follows:
textDiv.textContent = geom.str;
So replacing textDiv.textContent.length with geom.str.length slightly
improves performance without side effects.
Since the zoom value should be in percent, using `PDFViewer.currentScale` will be wrong by a factor of 100, potentially causing the zoom level of the document to become wrong on load.
*This fixes a regression from PR 6192.*
Under some circumstances, the `resize` event handler in `viewer.js` is fired before the scale has been set. This can lead to PDF documents being rendered at the wrong zoom level when they are opened.
It seems that a way to reliably trigger this is to, using the Firefox addon, open a PDF file that triggers the `fallback` bar, in a new background tab (i.e. middle clicking on a link, or use the context menu).
Prior to PR 6192, we checked the selected option in the `scaleSelect` dropdown instead. Since `pageAutoOption` is selected by default in `viewer.html`, this should explain why the issue wasn't visible previously.
This patch refactors the code responsible for setting the annotation's rectangle. Its goal is to:
- Actually check that the input array is actually an array, and if so, that it contains exactly four elements.
- Only call `normalizeRect` if the input array is valid, i.e., we do not call it for the default rectangle anymore.
Unit tests are provided just like with the other patches in this series.
Fixes#6106
To avoid future regressions, two new unit tests were added:
1. A new PDF based on the report from #6106, which contains an
OpenAction of type JavaScript and a string "this.print({...}".
2. An existing PDF from https://bugzil.la/1001080 (from #4698).
Although it does not matter, since we don't execute the JavaScript code,
I have also changed "print(true)" to "print({})" since the print method
takes an object (not a boolean). See "Printing PDF documents", page 62:
http://adobe.com/content/dam/Adobe/en/devnet/acrobat/pdfs/js_developer_guide.pdf
The special handling of the 'custom' scale value is only relevant for the `scaleSelect` dropdown in the standard viewer, hence I think that it should be placed in `viewer.js` instead.
Features / bug fixes in the preprocessor:
- Add word boundary after regex for preprocessor token matching.
Previously, when you mistakenly used "#ifdef" instead of "#if", the
line would be parsed as a preprocessor directive (because "#ifdef"
starts with "#if"), but without condition (because "def" does not
start with a space). Consequently, the condition would always be false
and anything between "#ifdef" and "#endif" would not be included.
- Add validation and error reporting everywhere, to aid debugging.
- Support nested comments (by accounting for the whole stack of
conditions, instead of only the current one).
- Add #elif preprocessor command. Could be used as follows:
//#if !FEATURE_ENABLED
//#error FEATURE_ENABLED must be set
//#endif
- Add #error preprocessor command.
- Add end-of-line word boundary after "-->" in the comment trimmer.
Otherwise the pattern would also match "-->" in the middle of a line,
and incorrectly convert something like "while(i-->0)" to "while(i0)".
Code health:
- Add unit tests for the preprocessor (run external/builder/test.js).
- Fix broken link to MDN (resolved to DXR).
- Refactor to use STATE_* names instead of magic numbers (the original
meaning of the numbers is preserved, with one exception).
- State 3 has been split in two states, to distinguish between being in
an #if and #else. This is needed to ensure that #else cannot be
started without an #if.
The previous regex was too greedy, it stripped a significant amount of
code when I put another file after core/murmurhash3.js. This was caused
by the fact that the license header in murmurhash3.js does not contain
"Mozilla Foundation", so the regex continued to match until my new file
(which had the standard license header containing "Mozilla Foundation").
It took a while to figure out why adding comments in worker_loader.js
caused the build to fail, because getWorkerSrcFiles did not print an
error message when it failed to parse the file. These issues have been
resolved as follows:
- Leading comments are stripped.
- The trailing comma is removed from the array.
- Errors are detected and useful error messages are printed.
Ordinarily, local files cannot be embedded in a non-local website. Until
this commit, the extension allowed websites to embed local PDF files on
non-local (e.g. http(s)) websites. This unintended feature is now
disabled, to align better with Chrome's existing security policies
(=local file:-URLs cannot be loaded in a tab unless expicitly allowed).
- Use rimraf instead of a custom removeDirSync implementation - rimraf
deals with edge cases like EPERM on Windows.
- Detect when the process exits before it was requested via stop(),
instead of running the cleanup handler.
- Add fallback for process detection when the process exits before it
was requested. On *nix systems, this is done via pkill and pgrep, on
Windows this is done via wmic.
- Add some asserts to check the preconditions of the methods, and output
some status information to aid debugging in case of failure.
I have verified that these changes work on ArchLinux and Windows XP,
using Chrome and Firefox, as follows:
1. node make unittest
2. node make unittest
3. Restart the Firefox process via the task manager as soon as possible.
4. node make unittest
5. Temporary lock a file/directory within the temporary profile
directory until the tests have finished, and then unlock the file
within 10 seconds.
In all cases, the auxilary browser processes are killed, and the
temporary profile directory is wiped.
Basic mathematics would suggest that a double negative should always become positive, but it appears that Adobe Reader simply ignores that case. Hence I think that it makes sense for us to do the same.
Fixes 6218.
After the creation of `PDFViewer`, its `_resetView` method takes care of resetting, among other things, the page number property. Hence we don't need to set `pdfViewer.currentPageNumber = 1;` here any more, and the comment is no longer accurate either.
When the parser finds a stream, it retrieves the Length from the stream
dictionary and advances the lexer to the offset as specified in Length.
If this Length is incorrect, the lexer could end up anywhere.
When the lexer gets in an invalid state, it could throw errors. For
example, in issue 6108, the lexer ends up inside the stream data. This
stream has the ASCIIHexDecode filter, so all characters are made up from
ASCII characters, and the lexer interprets it as a command token. Tokens
cannot be longer than 127 bytes, so eventually 128 bytes are consumed
and the lexer throws "Command token too long" error.
Another possible error is "Illegal character: 41" when the lexer happens
to end up at a ')' due to the length mismatch.
These problems are solved by catching lexer errors and recovering the
parser via the existing stream length detection branch.
*With this patch we're getting very close to fixing 6158.*
The only use-case for `PDFViewerApplication.updateScaleControls` is to try and avoid calling `selectScaleOption` from the `scalechange` event handler in viewer.js.
This will *only* happen when the user has manually changed the scale by using the `<select>` dropdown, which means that in reality this is just a micro optimization. Furthermore, `selectScaleOption` is only skipped for the "named" scale values (e.g. `auto`, `page-actual`, `page-fit`, `page-width`), thus further reducing the value of this code.
Also, since we're updating the scale `<select>` dropdown from an event handler, we're currently depending on the event being dispatched (and handled) completely before the next `scalechange` event. Relying on the execution order of the code in this way, even though it currently works, seems unfortunate since it *could* potentially cause the internal scale value and the UI from getting out of sync.
Xref offsets are relative to the start of the PDF data, not to the start
of the PDF file. This is clear if you look at the other code:
- In the XRef's readXRefTable and processXRefTable methods of XRef, the
offset of a xref entry is set to the bytes as given by a PDF file.
These values are always relative to the start of the PDF file (%PDF-).
- The XRef's readXRef method adds the start offset of the stream to
Xref entry's offset: "stream.pos = startXRef + stream.start".
Clearly, this line assumes that the entry offset excludes the start
offset.
However, when the PDF is parsed in recovery mode, the xref table is
filled with entries whose offset is relative to the start of the stream
rather than the PDF file. This is incorrect, and the fix is to subtract
the start offset of the stream from the entry's byte offset.
The manually created PDF file serves as a regression test. It is a valid
PDF, except:
- The integer to point to the start of the xref table and the %%EOF
trailer are missing. This will activate recovery mode in PDF.js
- Some junk was added before the start of the PDF file. This exposes the
bad offset bug.
The PDF specification (cited below) specifies a maximum length of a name
in bytes as a minimal architectural limit. This means that PDF *writers*
should not create names that exceed 127 bytes.
It does not forbid PDF *readers* to accept such names though. These
names are only used internally to link PDF objects to other objects. For
these use cases, the lengths of the names do not really matter. Hence I
have changed the implementation to not treat long names as errors, but
warnings.
> (7.3.5) The length of a name shall be subject to an implementation
> limit; see Annex C.
>
> (Annex C.2) Table C.1 describes the minimum architectural limits that
> should be accommodated by conforming readers running on 32-bit
> machines. Because conforming readers may be subject to these limits,
> conforming writers producing PDF files should remain within them.
>
> (Table C.1) name 127 "Maximum length of a name, in bytes."
http://adobe.com/content/dam/Adobe/en/devnet/acrobat/pdfs/PDF32000_2008.pdf
*This is the next step towards fixing 6158.*
This patch removes the dependency on the state of the scale `<select>` dropdown from the `resize` event handler, and instead uses the (in `PDFViewer`) stored `currentScaleValue`.
I believe that the way this code is currently written is purely for historical reasons, since originally *only* the numerical scale was stored internally (hence there was no other way to access the scale value).
However, since we now store the scale value, we should use it instead of quering the DOM. This helps ensure that the internally stored scale value is always accurately displayed in the UI (which should be good since, after the creation of `PDFViewer`, the `<select>` DOM element is now updated by an event handler).
For named destinations that are contained in a `Dict`, as opposed to a `NameTree`, we currently iterate through the *entire* dictionary just to fetch *one* destination.
This code appears to simply have been copy-pasted from the `get destinations` method, but in its current form it's quite unnecessary/inefficient since can just get the required destination directly instead.
Doing this helped uncover an issue with the `getDestination` implementation.
Currently if a named destination doesn't exist, the method (in `obj.js`) may return `undefined` which leads to the promise being stuck in a pending state.
*Note:* returning `null` for this case is consistent with other methods, e.g. `getOutline` and `getAttachments`.
*The next step towards fixing issue 6158.*
We can just as well access `pdfViewer.currentScaleValue` directly in `PDFViewerApplication`, instead of having a helper function which just acts as a wrapper for it.
Currently if the zoom level is reset multiple times in a row, i.e. by pressing <kbd>Ctrl</kbd>+<kbd>0</kbd>, the pages can be re-rendered each time even though their size shouldn't change. Whether this happens can depend on the size of the viewer, but documents with pages in landscape mode seem to be very susceptible to this. (An example is: https://wiki.mozilla.org/images/5/55/MobileOpportunity.pdf.)
This can also effect documents with pages in portrait mode, when they are displayed in Presentation Mode.
The reason for this unnecessary re-rendering is that due to limited numerical precision, the new scale value may change in *only* the last decimal place.
This is similar to the already existing, separate, CSS file used for the `textLayer`, and it's necessary in order for the `pageviewer` components example to actually show annotations correctly.
Considering that most methods of `SimpleLinkService` are complete stubs, or practically "useless" considering what they return, we can actually simplify it even more.
*Note:* This depends on the previous patch, that did a small amount of refactoring of `PDFViewer_scrollPageIntoView`, since `PDFViewer.linkService.page` is no longer accessed.
----------
Currently the `pageviewer` components example doesn't work correctly (an error is printed in the console), since no `linkService` is present when the `AnnotationsLayerBuilder` is created.
*Note:* Given that this uses the `SimpleLinkService`, clicking on e.g. internal links won't actually do anything. However, given that internal links (and similar features) are pretty much useless when only *one* page is loaded the `pageviewer` example, I don't think that really matters.
Also, using the complete `PDFLinkService` would require a `PDFViewer` instance. That would significantly complicate the example, thus making it both less clear and less self contained.
Currently in `PDFViewer_scrollPageIntoView`, we're accessing a number of properties in an indirect and overly complicated way. In particular, using `this.linkService.page` is a *very* roundabout way to access `this.currentPageNumber`.
The reason for this appears to be entirely historical, since prior to PR 5361 the code was placed in `PDFPageView` (or `PageView` as it was called at the time).
This patch is the the first step towards to addressing issue 6158, which will be done by refactoring the code for setting/getting the current scale in `viewer.js`.
As explained in
https://github.com/mozilla/pdf.js/issues/6174#issuecomment-118502802.
To verify that this patch works:
1. Build the Chrome extension (node make chromium)
2. Load the Chrome extension (at chrome://extensions)
3. Visit https://robwu.nl/pdfjs/issue6174/.
4. Verify that PDF.js is not used to load the PDF. Either Chrome's
default PDF Viewer is used, or the PDF is offered as a file download.
Before this patch, zooming in/out via the scroll wheel caused the page
to be zoomed relative to the upper-left corner of the page, i.e. the
upper-left corner of the page stays at a fixed position.
After this patch, the page is zoomed relative to the cursor position,
i.e. after zooming in/out, the part under the cursor 'has not moved'.
This only applies when the page does not fit in the viewport, because
pages smaller than the viewpoer are always centered.
In `viewer.js` we have code that ensures that the document is re-rendered on zooming and rotation. However, for `components` based viewers this might not work correctly, since there's currently no code which handles that.
Note that there is a good chance that this "just works" in many `components` viewers already, thanks to the `watchScroll` function.
The explanation is that zooming or rotation, most of the time, causes the viewer to change its (scrollable) size, thus triggering `PDFViewer_update` through `PDFViewer_scrollUpdate`.
However, in general there's no guarantee that this will actually work (since zooming and rotation doesn't necessarily change the size of the viewer for all documents), and requiring every viewer `components` implementer to provide methods for this doesn't seem like a great idea.
The patch adds a test to ensure that `getStats` returns the expected result after the page has been parsed, and cleans up the existing test a bit.
Also, since I'm touching the file anyway, I'm making a small adjustment of the `getDownloadInfo` test. (I have no idea why I didn't just write it like this initially.)
The reason that this code can be moved is that the `if (this.loading && page === 1)` check, in the `pagechange` event handler in viewer.js, is never satisfied since `this.loading` is not defined in that scope.
This *could* be considered a regression from PR 5295, since prior to that `this.loading` was using the `PDFViewerApplication` scope (or `PDFView` as it were).
However, I don't think that we need to fix that since we've been shipping this code in no less than *three* Firefox releases (uplifted in https://bugzilla.mozilla.org/show_bug.cgi?id=1084158), without breaking the world.
An explanation of why the `pagechange` code works, despite `this.loading === undefined`, is that `set currentPageNumber` (in `PDFViewer`) returns early whenever `this.pdfDocument` isn't set. This check is, for all intents and purposes, functionally equivalent to checking `PDFViewerApplication.loading`.
Hence we can move the page switching code into `PDFViewer`, and also remove `PDFViewerApplication.loading` since it's not used any more.
(The `this.loading` property was added in PR 686, which was before the current viewer even existed.)
*Note:* The changes in this patch should also be beneficial to the viewer `components`, since requiring every implementer to provide their own `pagechange` event handler just to get `PDFViewer.currentPageNumber` to actually work seems like an unnecessary complication.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1170063.
The bug only mentions the <kbd>Meta</kbd> key, but given that a similar situation can occur for <kbd>Ctrl</kbd>, it seemed reasonable to also handle that case in the same patch.
The only possible caveat with the patch is that because of the use of `shadow`, things won't work perfectly if either of the prefs are changed *while* the viewer is active. In this case a reload is required in order for it to work correctly, but given that the issue this patch fixes should be quite rare anyway, that seems OK.
Recently I've landed a number patches which fixed issues with ColorSpaces. In most of these cases the cause of the failures were, either partially or entirely, related to the fact that we didn't resolve indirect objects (i.e. the code was missing `xref.fetchIfRef(...)`).
The purpose of this patch is to fix the few remaining cases where indirect objects *could* potentially cause failures.
Given that we have seen how this causes failures in practice, I thus think that it makes sense to try and avoid further issues, instead of waiting for users to file even more bugs for this part of the code-base.
This patch adds:
- Unit tests for the annotation border style class
- Regression test (self-made) for the annotation border style class
- Documentation generation using JSDoc
Currently in the tests which check that incorrect passwords are rejected, we don't ensure that the exceptions thrown are the ones we expect. This patch improves the current situation, so that we actually can be sure that the code "fails" in the correct way.
*Note:* This patch also fixes some cases of weird indentation in the file.
Currently non-embedded ArialBlack fonts are not rendered bold enough, compared to e.g. Adobe Reader.
The issue is that we set the font weight to `bolder`, but since that is actually relative to the font weight of the parent, the result is that there's no practical difference from just using `bold`.
This patch attempts to address that, by explicitly setting the font weight to the maximum value instead (see https://developer.mozilla.org/en-US/docs/Web/CSS/font-weight).
*Note:* I expect one test "failure" in `issue5801`, which in this case is an improvement, since that PDF file uses ArialBlack.
Fixes 6068.
The most notable issue with the font in question is that the `differences` array contains lots of strange entries (of the type `uniXXXX`, instead of proper glyph names).
The Chrome extension is stable and has been around for a few years, so
the ancient note about the extension being experimental can be removed.
Also be a bit more specific about who maintains the Chrome extension,
to make it easier for external contributors to find who is responsible
for the development and maintenance of the PDF.js Chrome extension.
Currently `PDFLinkService` requires access to a `PDFHistory` instance in order for it to work correctly (and to avoid errors). If we want `PDFLinkService` to be more useful in custom viewers, I don't think that we actually want to force it to have a `PDFHistory` instance.
Hence this patch, which contains a very simply approach to make `PDFHistory` optional.
For some reason, https://bugzilla.mozilla.org/show_bug.cgi?id=1167053 changed methods of `NetUtil` yet *again*. This patch thus attempts to handle those changes, while keeping the addon backwards compatible.
I've tested this using all current Firefox versions (Nightly, Aurora/DevEdition, Beta, Release, ESR), and things still appears to work correctly.
The 'Version' field of the most recent document catalog, if present, is
intended to supersede the value in the file prologue.
This is significant for incrementally-built PDF documents and generators that
emit a low version in the prologue and later apply a format version based on
PDF features used, such as Apple's CoreGraphics/Quartz PDF backend.
Fixes the internal version variable, as well as the PDFFormatVersion reported
by the API and consumed by viewers.
This adds a checkbox with which one can disable scrolling, for example to look back at output during testing. Note that the styles are inline because the test runner removes all <style> elements for each test.
- Improve variable names and move variables to the top of the method
- Use constants where possible
- Only print the delay text if there is an atual delay set
- Simplify continuation logic in _nextTask
- Do not pass anything to clearCanvas: we can get the context from this.canvas there.
- Remove innerHTML and replace it with textContent. Add a comment why insertAdjancentHTML is so important for performance and runtime.
- Merge _quit and _sendQuitRequest.
- Extract NullTextLayerBuilder and SimpleTextLayerBuilder from the driver code and create classes for them.
- Move options to the constructor and pass in HTML elements instead of getting them in the driver.
- Remove unused masterMode variable.
- Minor method name updates.
The rest is only moving/indenting existing code and adding 'this'.
Reuse the existing getOutputScale() machinery to render and/or rescale
thumbnail images to the native display resolution. Fixes blurry thumbnails on
HiDPI displays making it a lot easier to make out the content.
During A/B testing we found a significant preference (increased engagement
time, reduced downloads of original PDF for local viewing) for the version that
doesn't animate loading thumbnails.
Profiling shows that the transition effects hit at a pessimal time when the
compositor is busy rendering the main PDF page view, and rendering thumbnail
previews, causing the main scroll view to stutter. With the transition removed
scrolling is perceptibly smoother in Chrome and Firefox. We also hypothesize
that the transition effects added to a perception of slowness but presumably
the aforementioned stutter is the primary issue here.
- Remove a hack for Chrome on Windows because we do not run Chrome on the Windows bot anymore, so it added unneeded complexity.
- Merge nextTask and continueNextTask functions. nextTask did nothing more than calling cleanup. This change also allows us to remove the callback for that function.
- Remove unnecessary one-line functions.
Having recently spent some time staring at the PasswordPrompt, while fixing issue 6010, I felt that the current border style does not really give a good visual indication that the input field actually has focus.
The current appearance was first introduced in PR 3527; but I don't know if having a different border style in Overlay dialogs was intentional, or if it just "happened".
However, given the colour palette used in the viewer UI, I think that using the same border style for all .toolbarFields makes sense.
For passwords where the encoding already is correct, the conversion is a no-op.
Also, since `encodeURIComponent` might throw, we need to make sure that we handle that case too.
Fixes 6010.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=1050040.
With this patch the file is completely readable, but given that the font is broken enough to be rejected by OTS the rendering differs slightly from Adobe Reader.
*Note:* the PDF file is sufficiently broken that even Adobe Reader complains about the font, *and* also about another more general issue.
*This patch addresses an issue I stumbled upon while working on rewriting the history implementation.*
Currently the `updateviewarea` UI event doesn't include the `location`, and in the event handler in viewer.js we instead reach into `PDFViewer` to retrieve it. Not only does this seem conceptually wrong, it also makes the `updateviewarea` event less useful in general since it's currently only possible to access the `location` if you have a reference to the `PDFViewer`.
To me, this simply appears to be an oversight in the implementation, hence this patch which adds `location` to the `updateviewarea` event. It also changes `location` to be a "private" property of `PDFViewer`.
*This is a follow-up to PR 5966.*
Now that https://bugzilla.mozilla.org/show_bug.cgi?id=1151931 has landed, we're unfortunately in a situation where Firefox warns about `unreachable code after return statement` multiple times when using the PDF.js code in `node make server` mode.
I'm really sorry about the churn in this code!
However I don't think it's a good idea to let Firefox continue to printing these warnings in the console, since it increases the risk of missing actual errors.
This patch:
- Simplifies the way fields are passed to the document properties overlay
- Simplifies the way fields are filled internally in the document properties overlay
- Avoids passing a document properties reference to the secondary toolbar
- This patch uncomments a large portion of the Firefox specific code in viewer.js, by using a preprocessor "trick", to enable linting. Doing this actually uncovered some issues, e.g. variables defined multiple times.
- This patch also fixes a spelling error, `adobe pdfwritter` -> `abobe pdfwriter`, in the `KNOWN_GENERATORS` list used when reporting telemetry data in Firefox.
*Please note:* there's still some Firefox specific code that is commented out, but it's usually just one (or two) lines of code. Hence the risk that errors creep in should be much lower, compared to entire code-blocks.
This patch:
- Puts public methods at the top of the file
- Puts private methods below the public methods and marks them with an underscore
- Adds JSDoc comments to the class
- Adds setDocumentAndUrl to avoid having to handle that in `viewer.js`
This patch removes the only remaining CPOW usage from the code-base, and should thus fix https://bugzilla.mozilla.org/show_bug.cgi?id=1071082.
*Note:* This will not fix the "Open With Different Viewer" button in e10s, since clicking it still fails with `frontWindow is null` in the console, but I do believe that that issue is somewhat orthogonal to the current patch.
This patch:
- Passes in a reference to the current PDFThumbnailViewer, which is used to ensure that the current thumbnail becomes visible when exiting PresentationMode.
- Changes the way that the event listeners for the contextmenu items are defined, to avoid passing in a reference to the SecondaryToolbar.
- Ensures that |supportsFullscreen| always returns a boolean.
Currently `supportsFullscreen` will, when the browser supports the fullscreen API, return e.g. `function mozRequestFullScreen()` instead of `true`.
- Simplifies the |click| handler code when PresentationMode is active.
This code has been obsolete ever since PR 2919 landed.
- Removes hack used to workaround a bug in WebKit browsers, which caused |mousemove| events to be fired when the cursor changed.
This was fixed close to a year ago, see http://code.google.com/p/chromium/issues/detail?id=103041.
This patch:
- Adds a couple of utility methods to `PDFViewer` to enable checking `presentationModeState` without cumbersome comparisons.
- Disentangles `PresentationMode` from `PDFHistory` and `HandTool`, by adding event listeners for `presentationmodechanged` to both of them.
- Adds a utility function to `PDFViewerApplication` for requesting PresentationMode.
- Prevents initialization of `PresentationMode` if the browser does not support the fullscreen API.
The previous commit implements a check for ShellJS and otherwise prompts the user to run "npm install". New clones of the codebase will need "npm install" for ShellJS and therefore automatically install jshint. Existing clones of the codebase will also need "npm install" again since ShellJS needs to be installed using NPM as it is not in the "external" folder anymore. Since everyone will get this prompt and install everything automatically, we will never reach this code path anymore.
This patch makes #5908 obsolete and reduces code complexity for the lint target. Thanks to @Snuffleupagus for noticing this!
There is no need to have a copy of ShellJS in the repository as it is also available on NPM. The NPM version is also much newer. This way we do not have to update this anymore and let NPM do that automatically.
After the refactoring in PR 5678, the B2G viewer now prints the following in the console:
```
ReferenceError: DownloadManager is not defined
```
This will obviously not matter once the B2G viewer is replaced with the new components-based one. But until that happens, I think it makes sense to just stub the class to suppress the error.
For consistency, this preference should be placed amongst the others. The current position is purely for "historical" reasons, since it was one of the first preferences that was added to the code.
Since the link is no longer valid, we should remove it from the comment.
(Using the Internet Archive, it seems that the contents of that page was basically identical to the PDF specification.)
In Blink-based browsers, there is a mismatch between document size and
paper size. Even if exactly the same values and unit are used, it is
possible that the printed results takes more pages than expected.
To solve the issue, the page size is set via @page size, and the canvas
and ancestor nodes are assigned a width+height of 100% (=relative to the
page). This change resolves bugs such as blank pages and split pages.
This patch also rearrange the order of the references to the DOM elements of the predefined zoom values, in viewer.js, to correspond with the element order in the current viewer UI. (The previous ordering is a leftover from the previous viewer UI.)
Instead of trying to hack around various browser defects, let's just disable PresentationMode in the affected browsers. This patch:
- Disables PresentationMode in IE11+ when the viewer is embedded; fixes 4711.
Set transformation matrix in (polyfilled) mozPrintCallback when a scale
is applied. Removed _scaleX and _scaleY in favor of _transformMatrix to
emphasize that the caller MUST ensure that the state of the matrix is
correct before `addContextCurrentTransform` is called.
Prior to PR 4259, we *incorrectly* ignored `toUnicode` for Type3 fonts. Since we now handle that correctly, this patch adds a `text` test-case to prevent regressions.
From reading IRC scrollback, it became clear that having to manually call `PDFFindController.resolveFirstPage` in order for the find functionality to work isn't particulary good. Hence this PR, which moves that code into `PDFViewer.setDocument` to make life easier for third-party implementations.
Currently if a font contains a `toUnicode` entry, we always create a new `ToUnicodeMap` in evaluator.js. This is done even for `IdentityV/IdentityH`, despite to possibility to use the much more compact `IdentityToUnicodeMap` representation.
This patch refactors the `IdentityH/IdentityV` cases, to:
- Avoid calling `IdentityCMap.getMap`, since this prevents allocating and iterating through an array with 65536 elements.
- Ensure that the handling of `toUnicode` is actually correct in fonts.js.
We rely on `toUnicode instanceof IdentityToUnicodeMap` in a few places, and currently this does not work correctly for `IdentityH/IdentityV`.
See "Case-insensitive language comparisons per RFC 4646"
https://github.com/fabi1cazenave/webL10n/pull/51
Removed the "patch" in compatibility.js and fixed the bug in webL10n,
because the logic belongs to webL10n instead of PDF.js.
The immediate motivation for this patch is that Chrome 40 converts
navigator.language to lower case: https://crbug.com/454331
We can pass it in using the options object. Note that that this also avoids creating a DownloadManager object for each separate link (instead, having only one is enough).
This includes an optimization to zero the height and width of existing thumbnail canvases, when they are removed and recreated during rotation of the document. (Credit goes to nnethercote, who initially found this in PR 4920.)
This is a very small follow-up to PR 5536, which sets `isStandardFont` to `false` instead of `undefined` (as currently happens for some font names).
Since the patch is so small, I hope it's OK to also fix an unrelated copy-and-paste error in a comment that was added in PR 5260.
In old versions of the viewer, anchor links were used to be able to scroll a page into view. Since we for quite some time have had a `scrollIntoView` function which handles this, the anchor links are no longer necessary and I believe that they can be removed. Doing so will reduce the total number of DOM elements in the viewer.
For most documents this patch will probably not make much of a difference performance wise, but for very long (1000+ pages) documents this could help performance especially on lower powered hardware.
Now that [bug 649849](https://bugzilla.mozilla.org/show_bug.cgi?id=649849) has been fixed, adding support for `-moz-appearance: none`, the arrow is now too close to the text in the zoom box. This is currently only an issue in Nightly, but assuming the patch doesn't get backed out, this will soon affect all versions of Firefox.
The only simple solution I could find seems to be removing `*-appearance: none` rules from the CSS. I haven't been able to find any easier solutions that still looks the same with/without bug 649849.
This change does the following:
* Address TODO to remove getEmptyContainer helper.
* Not set container bg-color. The old code is incorrect, causing it to
not have any effect. It sets color to an array (item.color) rather
css string. Also in most cases it would set it to black background
which is incorrect.
* only add border instructions when there is actually a border
* reduce memory consumption by not creating new 3 element arrays for
annotation colors. In fact according to spec, this would be incorrect,
as the default should be "transparent" for an empty array. Adobe
Reader interprets a missing color array as black however.
Note that only Link annotations were actually setting a border style and
color. While Text annotations might have calculated a border they did
not color it. This behaviour is now controlled by the boolean flag.
According to practical experiments, falling back to "Helvetica" when we encounter a non-embedded "[Century Gothic](http://en.wikipedia.org/wiki/Century_Gothic)" `CIDFontType2` font seems to work well.
(Also, the section on Wikipedia about "Printer ink usage" *might* provide some anecdotal evidence that Century Gothic is a fairly standard sans-serif font.)
Obviously this patch doesn't make "Century Gothic" fonts render perfectly, as is often the case with non-embedded fonts, but all the text is now legible in the referenced issues.
Fixes 4722.
Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=879561.
Fixes two issues:
- #4456 : The first 100 bytes are often not unique as they can be
filled with standard PDF headers - so we use the first 200 KB instead.
(This may be overkill)
- Some documents we encountered have invalid xref ids, which were
always coming out as ‘0000000000000000’ - so we detect that and use the
MD5 instead.
When submitting PR 5276 there wasn't a good PDF file to include in the test suite. However, with https://bugzilla.mozilla.org/show_bug.cgi?id=1108753, we now have a better source for a test file, hence this patch.
This is a tentative patch that adds *very* basic support for non-embedded Wingdings fonts (a Windows version of Dingbats), by falling back to the ZapfDingbats encoding. Obviously this approach will not work perfectly, but in my opinion it seems to work reasonably well in pratice.
Instead of this very simple patch, another option would be to try and include more complete glyph data for Wingdings, e.g. a Unicode map and glyph widths, similar to what was done for ZapfDingbats.
However there is, in my opinion, one important difference between Wingdings and ZapfDingbats: ZapfDingbats is part of the 14 standard fonts, which in previous versions of the PDF specification was assumed to be available in PDF readers. To improve compatibility with older files, it thus makes sense for us to include data for ZapfDingbats.
However Wingdings has never been a standard font in PDF files, hence PDF files using it *should* thus contain all the necessary font data.
Given the above, I thus believe that it should be OK to fall back to ZapfDingbats for now. If non-embedded Wingdings fonts turns out to be *a lot* more common, then we can revisit this later.
Fixes 4301 completely.
Fixes 4837 almost completely. With this patch the bullets are displayed correctly, but the arrows are not of the correct type.
Fixes `artofwar.pdf`, pages 14 and 15.
maskData comes out of maskCtx.getImageData, so is 0..255 clamped, and
the used multiplications will not create fractions needing rounding,
neither would addition.
This checks for both prefs on the understanding that we need to work on older versions of Firefox. If that isn't the case, the first part of the if isn't necessary. This should only land if bug 639134 is resolved - I'd make the patch part of that bug, but AIUI pdfjs's canonical repo is on github, so...
Currently if you manage to e.g. open the console (with <kbd>Ctrl</kbd>+<kbd>Shift</kbd>+<kbd>K</kbd>) before the viewer is initialized, the following will be printed in the console: `TypeError: pdfViewer is null`.
This doesn't cause any actual errors, but nevertheless it seems like something we should avoid.
Followup to PR 5413.
When a search term isn't found, the background color of the findInput is supposed to change (to red). This is currently not working as intended, because the CSS rule is not being applied correctly. (It seems that this broke in PR 2208.)
This patch also changes the background color to match the one used in the native Firefox findbar, since the old color seemed a bit too pink.
As described in #5444, the evaluator will perform identity checking of
paintImageMaskXObjects to decide if it can use
paintImageMaskXObjectRepeat instead of paintImageMaskXObjectGroup.
This can only ever work if the entry is a cache hit. However the
previous caching implementation was doing a lazy caching, which would
only consider a image cache worthy if it is repeated.
Only then the repeated instance would be cached.
As a result of this the sequence of identical images A1 A2 A3 A4 would
be seen as A1 A2 A2 A2 by the evaluator, which prevents using the
"repeat" optimization. Also only the last encountered image is cached,
so A1 B1 A2 B2, would stay A1 B1 A2 B2.
The new implementation drops the "lazy" init of the cache. The threshold
for enabling an image to be cached is rather small, so the potential waste
in storage and adler32 calculation is rather low. It also caches any
eligible image by its adler32.
The two example from above would now be A1 A1 A1 A1 and A1 B1 A1 B1
which not only saves temporary storage, but also prevents computing
identical masks over and over again (which is the main performance impact
of #2618)
This patch makes the image from #5349 appear correctly, the artefacts
for the last packet are fixed in #5426.
This patch also optimizes some "in-checks" and adds a few header parsings.
As of Firefox 35, isContentWindowPrivate should be used for DOM windows instead of isWindowPrivate.
See https://bugzilla.mozilla.org/show_bug.cgi?id=1069059 and
http://hg.mozilla.org/mozilla-central/diff/324798b60ba3/toolkit/modules/PrivateBrowsingUtils.jsm
Without this fix, you will get the following error message when Firefox+PDF.js is started:
>
WARNING: content window passed to PrivateBrowsingUtils.isWindowPrivate. Use isContentWindowPrivate instead (but only for frame scripts).
pbu_isWindowPrivate@resource://gre/modules/PrivateBrowsingUtils.jsm:25:14
ChromeActions.prototype.isInPrivateBrowsing@resource://pdf.js/PdfStreamConverter.jsm:237:12
xhr_onreadystatechange@resource://pdf.js/PdfStreamConverter.jsm:545:30
NetworkManager_requestRange@resource://pdf.js/network.js:95:7
NetworkManager_requestRange@resource://pdf.js/network.js:81:14
RangedChromeActions_requestDataRange@resource://pdf.js/PdfStreamConverter.jsm:596:1
RequestListener.prototype.receive@resource://pdf.js/PdfStreamConverter.jsm:705:5
PdfStreamConverter.prototype.onStartRequest/proxy.onStopRequest/<@resource://pdf.js/PdfStreamConverter.jsm:909:11
FirefoxComClosure/<.request@resource://pdf.js/web/viewer.js:529:14
PdfDataRangeTransport_requestDataRange@resource://pdf.js/web/viewer.js:2977:9
transportDataRange@resource://pdf.js/build/pdf.js:2122:13
messageHandlerComObjOnMessage@resource://pdf.js/build/pdf.js:1219:9
After PR 5263, setting `disableAutoFetch = true` in the generic viewer no longer works correctly, since the entire file loads even with `disableStream = true`.
Since `preferenceSidebarViewOnLoad` is an enumeration value, it seems better to initialize it with the default one instead of a boolean.
`preferencesPdfBugEnabled` uses a superfluous "s", which is a typo I made when I introduced it.
Even after PR 5359, switching to/from Presentation mode is still sluggish and I'm still seeing `too much recursion` printed in the console.
I've managed to track down the issue to line https://github.com/mozilla/pdf.js/blob/master/web/page_view.js#L371. It appears that for this particular case, we actually *need* to set `noScroll = true` when calling the `_setScale` function. (Note that it only applies to *this* line, and not to Presentation mode in general.)
Given the amount of refactoring done to the viewer, I'm not sure what the best way to fix this would be. However I'm submitting this patch as an easy workaround for now, but we should probably refactor this to avoid calling a "private" method directly.
After PR #5282, the text highlighting in the Font Inspector is very hard to see against a white background. That seems to be an oversight from the mentioned PR, hence this patch fixes that.
As a follow-up to PR #4795, this patch improves the `getData` unit test to ensure that the returned data is of the expected type and of the correct length.
As a consequence of merging #5221 it is more likely to have multiple
overlapping selection divs inside the text layer. Because each individual
element gets the selection style applied, the 30%opacity stacks, making a
60% bar visible where the overlap happens.
As proposed by @rocallahan, this can be fixed by making the selection
style solid and setting opacity for the overall layer.
I assume also that this should make the work for the renderer easier, but
was unable to bench it.
Currently when an exception is thrown, we try to reject `workerReadyCapability` with multiple arguments in src/core/api.js. This obviously doesn't work, hence this patch changes that to instead reject with the exception object as is.
In src/core/worker.js the exception is currently (unncessarily) wrapped in an object, so this patch also simplifies that to directly send the exception object instead.
"This commit fixes the issue that a focused element cannot lose
focus when the draggable element is clicked."
https://github.com/Rob--W/grab-to-pan.js/commit/11c368f97cebe19
Steps to reproduce/verify:
1. Enable the hand tool (e.g. by using the H shortcut)
2. Click on the Page number input.
3. Click or drag the PDF canvas.
4. Observe that the page number input is still focused.
Consequently, the hand tool cannot be disabled using the H shortcut.
Currently in RTL locales, the loading indicator is placed such that it is in the way when entring a search term. Hence this patch moves it to the other side of the input field to fix this.
The files issue3115.pdf and issue2337.pdf are identical, the only difference being that the first one is an `eq` test and the second one a `load` test. Hence there is no reason to keep the `load` test, since it's just a subset of the `eq` test.
Remove pageAction logic from extension router, and put it in a
separate file. The pageAction URL parsing logic has been simplified,
and all pageAction-related files have been moved to a separate directory.
This change makes scrolling noticeably smoother on files with many
single-char text divs, such as the one in #1045. The trade-off is that
the visual appearance of text selection in such documents is slightly
worse, because more text divs overlap.
This change also uses `scaleX(N)` instead of `scale(N, 1)`. This might
be marginally more efficient in terms of JS string concatenation.
This patch avoids creating many intermediate strings, when adding dummy width/lsb entries for glyphs where those are missing.
For the relevant PDF files in our test suite, the average number of intermediate strings are well over 1000.
transformOrigin is set to 0% 0% in all cases. This adds extra memory
impact into the dom tree. It also involves the CustomStyles workaround
to determine the correct css rule for the browser.
By setting all vendor and standard variants in css, the rule is applied
without the dom memory overhead and without the minor computation
overhead to set the value.
The scanned, black-and-white document at
https://bugzilla.mozilla.org/show_bug.cgi?id=835380 doesn't benefit from
the critical GRAYSCALE_1BPP optimization because the optimization is
skipped if `needsDecode` is set.
This change addresses that, and reduces both rendering time and memory
usage for that document by almost 10x.
As requested in #5178, this change debounces the scroll events.
The reason for doing so is that browsers can event-storm especially on
scroll, communicating hundreds of subpixel changes.
The main reason for this resulting in poor performance is that on each
scroll event `scrollTop` was queried, which forces layouting.
This change will use `requestAnimationFrame` to make sure the browser can
allocate enough time to other tasks. The delay is however quite small, thus
the reduction in executions is less noticeable. Modern browsers however utilize
`requestAnimationFrame` to smoothen out rendering.
setGStateForKey() is a closure that serves no particularly useful
purpose. This change inlines it at the single call site. This avoids 1.7
MiB of allocations (because closures are objects) for the MTA map
mentioned in https://bugzilla.mozilla.org/show_bug.cgi?id=835380#c17.
The data-font-name attribute of textLayer divs are only used by the Font
Inspector. This change ensures they are only present when the pdfBug
tools are enabled.
With the changes made in PR 5169 to enable quicker display of the File size, `setFileSize()` can now be called during file loading. This function parses *and* localizes the file size, where the l10n is a synchronous operation.
This patch:
* Re-factors `setFileSize()` to *only* update `rawFileSize`, and moves parsing and fetching of the l10n string to a separate method that isn't called until the Document Properties dialog is actually opened.
* Adds a couple of checks to make sure that `rawFileSize` is set to a positive number.
* Removes the `fileName` and `fileSize` properties. Since they are parsed every time the Document Properties dialog is opened, it doesn't seem necessary to cache them.
For the document in #2504, 11% of the ops are `setGState` with a
`gStateObj` that is an empty array, which is a no-op. This is possible
because we ignore various setGState keys (OP, OPM, BG, etc).
This change prevents these ops from being inserted into the operator
list.
This allows the JS engine to do a better job of allocating the right
number of elements for the array, avoiding some resizings. For the PDF
in #2504, this avoids 100s of MiBs of allocations in Firefox.
makeInlineImage() has a "are the next five chars ASCII?" check which is
run after an "EI" sequence has been found. This check involves the
creation of a new object because peekBytes() calls subarray().
Unfortunately, the check is currently run on whitespace chars even when
an "EI" sequence has not yet been found, i.e. when it's not needed. For
the PDF in #2618, there are over 820,000 such checks.
This change reworks the relevant loop so that the check is only done
once an "EI" sequence has been seen. This reduces the number of checks
to 157,000, and speeds up rendering by somewhere between 2% and 7% (the
measurements are noisy).
When the binary CMaps were added, some of the relevant unit tests were not changed. This patch updates them, so that we actually test the current implementation.
What's somewhat troubling here is that we currently have CMap unit tests that passes, *despite* not working as intended (the CMap files doesn't load).
The `console.log` statement in evaluator_spec.js is obviously not needed. In obj.js it could have been replaced by `info`, but that seemed unnecessary given the already existing `error`.
readCharCode() returns two values, and currently allocates a length-2
array on every call to do so. This change makes it instead us a
passed-in object which can be reused.
This tiny change reduces the total JS allocations done for the document
in Mozilla bug 992125 by 4.2%.
Since the "Previous/Next Page" buttons are disabled when the first/last page is active, I think that for the sake of consistency the "Go to First/Last Page" should behave the same way.
EvaluatorPreprocessor_read() is called in two cases. For the normal
layer, the args array it produces is used beyond the bounds of the loop
in which EvaluatorPreprocessor_read() is called.
But for the text layer, the args array is used in a very short-term
fashion. This change reworks things so that a single array is repeatedly
used for the text layer. This reduces total JS allocations for the
Spoorkaart map by 11%, and has similar effects on many other PDFs.
Currently the File size field in the Document Properties dialog isn't set until `PDFView.pdfDocument.getDownloadInfo()` is resolved. If the Document Properties dialog is opened while a PDF file is loading with range requests, this leads to the less desirable situation where all fields *except* File size are available.
In cases where `PDFView.open()` is called with the `args` parameter defined, and `args` contains the property `length`, we actually know the File size when the PDF file begins to load. (This is usually the case when ranged loading is used in the Firefox addon/built-in version.)
Hence we can use `args.length` to set the File size immediately when `PDFView.open()` is called, resulting in a better user experience.
After PR 4982, the rendering of the first two pages of http://www.openmagazin.cz/pdf/2011/openMagazin-2011-04.pdf (from issue 215) no longer completes.
The issue is that we cannot have `args === null` in `PartialEvaluator_buildPath`, but *must* use an empty array instead.
In this patch I've also moved the `argsLength` variable definition in `EvaluatorPreprocessor_read`, to make sure that it's always defined.
When loading the PDF from issue #4935, this change reduces peak RSS from
~2400 to ~300 MiB, and improves overall speed by ~81%, from 6336 ms to
1222 ms.
This change improves the text layer creation of "normal" text heavy documents.
It changes 2 things:
* If the text element is not rotated, it will neither calculate nor apply a
textTransform: rotate(0deg) to the text layer element.
* For scaling the text layer div, the context will measure the width of the
text in the given font. For many text documents the font changes seldom.
If the font stays the same, the context does not need to be set to a new font
especially avoiding the temporary creation of the same font string over and
over again.
IdentityCMap uses an array to represent a 16-bit unsigned identity
function. This is very space-inefficient, and some files cause multiple
IdentityCMaps to be instantiated (e.g. the one from #4580 has 74).
This patch make the representation implicit.
When loading the PDF from issue #4580, this change reduces peak RSS from
~370 to ~280 MiB. It also improves overall speed on that PDF by ~30%,
going from 522 ms to 366 ms.
QtWebKit does not support file:-URLs in Web Workers.
To solve this issue, I'm loading file:-URLs in the main thread
for generic builds. file:-URLs load relatively quick, so there
is no need for sophisticated load progress notifications.
cid chars are 16-bit unsigned integers. Currently we convert them to
single-char strings when inserting them into the CMap, and then convert
them back to integers when extracting them from the CMap. This patch
changes CMap so that cid chars stay in integer format throughout, saving
both time and space.
When loading the PDF from issue #4580, this change reduces peak RSS from
~600 to ~370 MiB. It also improves overall speed on that PDF by ~26%,
going from 724 ms to 533 ms.
PR 4259 fixed a large number of font bugs, but none of those where added as test-cases. This was, in my opinion, unfortunate since it increases the risk of regressions in the future when other font bugs are fixed.
This PR simply adds a few more test-cases, to improve our test coverage somewhat.
document.documentElement.style is null in some XML documents.
The previous snippet caused the following error:
Uncaught TypeError: Cannot use 'in' operator to search for 'animation' in null
To fix this bug, `'animation' in document.documentElement.style` has been
replaced with `CSS.supports('animation', '9s')`. This method was introduced
in Chromium 28, but it is not necessary to detect whether this method is
supported because the required createShadowRoot method for embeds is not
available in Chromium 32 and earlier.
This change avoids the element stringification caused by for..in for the
vast majority of CMaps.
When loading the PDF from issue #4580, this change reduces peak RSS from ~650
to ~600 MiB, and improves overall speed by ~20%, from 902 ms to 713 ms. Other
CMap-heavy documents will also see improvements.
This lets the JS engine resize the array elements buffer immediately,
thus avoiding some intermediate resizings. This can save multiple MiBs
of reallocation in text-heavy files.
If a user downgrades from Chromium 35+ to 34, then the PDF Viewer
extension will not work any more because the extension assumes
that certain features were available based on the cached feature
detection results.
To resolve this problem, all feature detection scripts run again
if the browser was downgraded.
In https://github.com/mozilla/pdf.js/commit/b5b94a4af389ef387570cf22662d9fc6bd41417a, i.e. PR #4259, we stopped using cidmaps.js. Despite that, it's still included when PDF.js is built. At almost 0.5 MB (and approx. 7000 lines), this is currently the single largest file in the codebase.
Including such a large file in the builds, when it is not actually used, seems extremely wasteful; hence this patch.
The absence of this property caused the following error in PDFView.download:
Cannot read property 'split' of undefined
This property was unset when the PDF was loaded through a typed array instead
of a URL. With the fix, the file name will default to "document.pdf", as
defined by getPDFFileNameFromURL in web/ui_utils.js
I have a large PDF where this function is called 1.6 million times
during loading. Minimizing the string concatenations reduces the
cumulative allocations done by Firefox within this function from 113 MB
to 48 MB.
The current way that the tabindices are defined, has the unfortunate consequence that adding a single button to the UI might mean having to change almost all tabindices. (This was the case in e.g. PR 4329.)
This patch attempts to avoid that situation happening again, by re-defining the tabindices slightly.
It's currently possible to step through the test results using the <kbd>N</kbd> and <kbd>P</kbd> keys, and you can also switch between test and reference images with the <kbd>T</kbd> key.
However if you want to highlight the differences, that can only be done by clicking. This has always annoyed me somewhat, so this patch adds support for toggling the differences view with the <kbd>D</kbd> key.
In the referenced issue, there is a PDF which uses a fill pattern which does not
have a matrix defined. This causes singularValueDecompose2dScale to fail with
undefined property error when accessing elements of that matrix.
This fix will only use the matrix when it is defined. The output for the PDF in
question now looks identical to chrome and preview with respect to the gradient
fill pattern.
QueueOptimizer is really hard to read. Enough so that it's blocking my
efforts to streamline the representation used for operator lists.
This patch improves its readability in the following ways.
- More descriptive variable names make the sequence checking much clearer,
as do additional comments.
- The addState() functions now return the index of the first op past the
sequence, instead of setting context.currentOperation to the last op of
the sequence.
- The loop in optimize() is clearer.
- The array modification in the fourth addState() function is much clearer
-- we're just removing trios of ops.
- All four |addState| functions are now more consistent with each other.
I used some debug printfs to find documents where these optimizations are
used and then checked that the number of optimized ops was the same before
and after my changes.
DecodeStream currently initializes its |buffer| field to |null|, which
is reasonable, because lots of DecodeStreams never need to instantiate a
buffer. But this requires various special cases in the code.
This patch change it so DecodeStreamClosure has a single empty
Uint8Array which gets shared between all buffers upon initialization.
This avoids the special cases.
DecodeStream.prototype.ensureBuffer() is really hot, and this removes a
test from the fast path. For one 226 page scanned document this sped up
rendering by about 2%.
In src/core/obj.js, we convert a Ref to a string to index into a table like
this: 'R1.0'. This conversion is repeated numerous times.
This patch factors out the conversion into a new function.
Ref.prototype.toString().
This function can be called 100s of 1000s or even millions of times, and the
allocated return object accounts for 10% of all GC thing allocations for some
documents. It's easy to avoid, which reduces stress on the garbage collector,
and this patch does that.
PartialEvaluator.getTextContent() builds up textChunk strings 1 char at a time,
creating many 100s of 1000s of intermediate strings along the way. This patch
make it instead push chars to an array and then join them at the end, as we
have done in numerous other places.
This new function is much faster than ensureRange(pos, pos+1), which is a very
common case.
This speeds up the rendering of some test cases (including the Tracemonkey
paper) by 4--5%.
Currently (at least in Firefox) when the HOME/END keys are pressed, this will trigger unnecessary rendering of pages that lay between the current page and the first/last page. Avoid this by going straight to the first/last page instead.
*Added AES128 Encryption
*Added AES258 Encryption/Decryption
*Added SHA256
*Added SHA512
*Added class to handle 8 byte integers and associated bit operations
*Added SHA384
*Added routines to handle new algorithm and perform PDF2.0 hashing.
Profiling showed that receiveAndExtend is frequently called with the length of
one bit. This happens for example in decodeBaseline.
For a single bit, the loop and shift in receive, as well as the shifts in
receiveAndExtend are overhead.
This shortcut manually calculates the shifts by either returning 1 or -1 from
receiveAndExtend by reading the bit and deciding on the return value.
While it comes with an overhead for each non-one length, the speedup is at about
10% in the hot parse/decode path.
This change makes the 4 conversion loops look the same.
It optimizes access of the array length and access of the property
numComponents, which is known to be constant.
"text/javascript" is not a correct MIME type (the correct one is
"application/javascript") but it's not even needed; all browsers default
to the correct type and treat it as executable JS when type is ommited.
Since not all browsers recognize the "application/javascript" MIME type
the only way to both stay compliant and to support all popular browsers
is to omit the type. It's also shorter this way.
This refactors getData to be more readable and extracts all the color
conversion algorithms to their own functions. The resulting code was then
cleaned up.
This also introduces a flag `forceRGBoutput` to getData, that allows to always
get the data as a `width * height * 3` bytes long RGB buffer
The function `assertWllFormed` was doing nothing different than `assert` which is
available in the same namespace. Removing it will lighten the filesize - albeit
very slightly - and reduce complexity.
Different fonts can point to the same font descriptor
(see https://github.com/mozilla/pdf.js/issues/4339 for details). With this
commit such fonts are treated as aliases if they have also the same encoding
and the same toUnicode map. The according info is stored on the font descriptor.
This change must also ensure that aliases use always the same font name
because translated fonts can get cleared depending on the CLEANUP_TIMEOUT setting.
The issues are used to track both bugs filed by users and specific work items for developers. Try to file one issue per problem observed. Please specify a valid title (e.g. "Glyph spacing is incorrect" instead of "PDF.js does not work") and provide more details about the issue: link to the PDF, location in the PDF, screenshot, browser version, operating system, PDF.js version and JavaScript console warning/error messages. Issues that do not have enough details provided will be closed as invalid/incomplete.
The issue tracking system is designed to record a single technical problem. A bug report is something where a developer/contributor can work on. The GitHub issue tracker is not a good place for general, not well thought out or unworkable ideas. Most likely a discussion-type issue will not be addressed for a long time or closed as invalid. The best place is our dev-pdf-js@lists.mozilla.org mailing list. You can subscribe to it using http://lists.mozilla.org or Google Groups. This way you will reach not only developers. As an alternative, you can join our weekly engineering meeting to discuss new ideas for the project.
If the issue is related to errors produced by a specific PDF, please always include the PDF by providing a URL where contributors can download it. Without a PDF for reproduction, such issues will be closed. We understand that many PDFs contain sensitive information, however having a PDF is essential to resolving the issue and building our regression testing suite. If possible, try creating a reduced example exhibiting the problem but not containing sensitive data. Also small PDFs are best suited for our regression testing. If an important issue only shows on sensitive PDFs, contributors might be willing to accept these PDFs via a secure exchange.
If you are developing a custom solution, first check the examples at https://github.com/mozilla/pdf.js#learning and search existing issues. If this does not help, please prepare a short well-documented example that demonstrates the problem and make it accessible online on your website, JS Bin, GitHub, etc. before opening a new issue or contacting us on the IRC channel -- keep in mind that just code snippets won't help us troubleshoot the problem.
The issue tracking system is designed to record a single technical problem. A bug report is something where a developer/contributor can work on. The GitHub issue tracker is not a good place for general, not well thought out or unworkable ideas. Most likely a discussion-type issue will not be addressed for a long time or closed as invalid. The best place for general discussions is our Matrix room at https://chat.mozilla.org/#/room/#pdfjs:mozilla.org.
Note that the translations for PDF.js in the `l10n` folder are synchronized with the Aurora branch of Mozilla Firefox. This means that we will only accept pull requests that add strings currently missing in the Aurora branch (because it will take at least six weeks before the most recent translations are in the Aurora branch), but keep in mind that the changes will be overwritten when we synchronize again.
If you are developing a custom solution, first check the examples at https://github.com/mozilla/pdf.js#learning and search existing issues. If this does not help, please prepare a short well-documented example that demonstrates the problem and make it accessible online on your website, JS Bin, GitHub, etc. before opening a new issue or contacting us in the Matrix room -- keep in mind that just code snippets won't help us troubleshoot the problem.
Note that the translations for PDF.js in the `l10n` folder are imported from the Nightly channel of Mozilla Firefox, such that we don't have to maintain them ourselves. This means that we will not accept pull requests that add new languages and/or modify existing translations, unless the corresponding changes have been made in Mozilla Firefox first.
This repository is governed by Mozilla's code of conduct and etiquette guidelines.
For more details, please read the
[Mozilla Community Participation Guidelines](https://www.mozilla.org/about/governance/policies/participation/).
## How to Report
For more information on how to report violations of the Community Participation Guidelines, please read our '[How to Report](https://www.mozilla.org/about/governance/policies/participation/reporting/)' page.
<!--
## Project Specific Etiquette
In some cases, there will be additional project etiquette i.e.: (https://bugzilla.mozilla.org/page.cgi?id=etiquette.html).
PDF.js is built into version 19+ of Firefox, however two extensions are still
available that are updated at a different rate:
PDF.js is built into version 19+ of Firefox.
+ [Development Version](http://mozilla.github.io/pdf.js/extensions/firefox/pdf.js.xpi) - This version is updated every time new code is merged into the PDF.js codebase. This should be quite stable but still might break from time to time.
+ [Stable Version](https://addons.mozilla.org/firefox/addon/pdfjs) - After version 24 of Firefox is released we no longer plan to support the stable extension. The stable version will then be considered whatever is built into Firefox.
#### Chrome
#### Chrome and Opera
The Chromium extension is still somewhat experimental but it can be installed two
ways:
+ [Unofficial Version](https://chrome.google.com/webstore/detail/pdf-viewer/oemmndcbldboiebfnladdacbdfmadadm) - *This extension is maintained by a PDF.js contributor.*
+ Build Your Own - Get the code as explained below and issue `node make chromium`. Then open
+ The official extension for Chrome can be installed from the [Chrome Web Store](https://chrome.google.com/webstore/detail/pdf-viewer/oemmndcbldboiebfnladdacbdfmadadm).
*This extension is maintained by [@Rob--W](https://github.com/Rob--W).*
+ Build Your Own - Get the code as explained below and issue `gulp chromium`. Then open
Chrome, go to `Tools > Extension` and load the (unpackaged) extension from the
directory `build/chromium`.
The version of the extension for the Opera browser can be found at the [Opera add-ons catalog](https://addons.opera.com/en/extensions/details/pdf-viewer/).
## Getting the Code
To get a local copy of the current code, clone it using git:
Next, you need to start a local web server as some browsers don't allow opening
PDF files for a file:// url:
Next, install Node.js via the [official package](https://nodejs.org) or via
[nvm](https://github.com/creationix/nvm). You need to install the gulp package
globally (see also [gulp's getting started](https://github.com/gulpjs/gulp/blob/master/docs/getting-started.md#getting-started)):
$ node make server
$ npm install -g gulp-cli
You can install Node via [nvm](https://github.com/creationix/nvm) or the
[official package](http://nodejs.org). If everything worked out, you can now
serve
If everything worked out, install all dependencies for PDF.js:
$ npm install
Finally, you need to start a local web server as some browsers do not allow opening
PDF files using a `file://` URL. Run:
$ gulp server
and then you can open:
+ http://localhost:8888/web/viewer.html
You can also view all the test pdf files on the right side serving
Please keep in mind that this requires an ES6 compatible browser; refer to [Building PDF.js](https://github.com/mozilla/pdf.js/blob/master/README.md#building-pdfjs) for usage with older browsers.
It is also possible to view all test PDF files on the right side by opening:
+ http://localhost:8888/test/pdfs/?frame
## Building PDF.js
In order to bundle all `src/` files into two productions scripts and build the generic
viewer, issue:
In order to bundle all `src/` files into two production scripts and build the generic
viewer, run:
$ node make generic
$ gulp generic
This will generate `pdf.js` and `pdf.worker.js` in the `build/generic/build/` directory.
Both scripts are needed but only `pdf.js` needs to be included since `pdf.worker.js` will
be loaded by `pdf.js`. If you want to support more browsers than Firefox you'll also need
to include `compatibility.js` from `build/generic/web/`. The PDF.js files are large and
should be minified for production.
be loaded by `pdf.js`. The PDF.js files are large and should be minified for production.
## Using PDF.js in a web application
To use PDF.js in a web application you can choose to use a pre-built version of the library
or to build it from source. We supply pre-built versions for usage with NPM and Bower under
the `pdfjs-dist` name. For more information and examples please refer to the
[wiki page](https://github.com/mozilla/pdf.js/wiki/Setup-pdf.js-in-a-website) on this subject.
## Including via a CDN
PDF.js is hosted on several free CDNs:
- https://www.jsdelivr.com/package/npm/pdfjs-dist
- https://cdnjs.com/libraries/pdf.js
- https://unpkg.com/pdfjs-dist/
## Learning
You can play with the PDF.js API directly from your browser through the live
demos below:
You can play with the PDF.js API directly from your browser using the live demos below:
More examples can be found in the [examples folder](https://github.com/mozilla/pdf.js/tree/master/examples/). Some of them are using the pdfjs-dist package, which can be built and installed in this repo directory via `gulp dist-install` command.
For an introduction to the PDF.js code, check out the presentation by our
We're currently working on better API docs, but the API is well documented in [api.js](https://github.com/mozilla/pdf.js/blob/master/src/display/api.js).
We're currently working on <a href="draft/index.html">better API docs</a>, but the API is well documented in [api.js](https://github.com/mozilla/pdf.js/blob/master/src/display/api.js).
PDF.js heavily relies on the use of [Promises](https://developer.mozilla.org/docs/Web/JavaScript/Reference/Global_Objects/Promise). If promises are new to you, it's recommended you become familiar with them before continuing on.
This tutorial shows how PDF.js can be used as a library in a web browser.
[examples/](https://github.com/mozilla/pdf.js/tree/master/examples) provides more examples, including usage in Node.js (at [examples/node/](https://github.com/mozilla/pdf.js/tree/master/examples/node)).
### Document
The object structure of PDF.js loosely follows the structure of an actual PDF. At the top level there is a document object. From the document, more information and individual pages can be fetched. To get the document:
```js
PDFJS.getDocument('helloworld.pdf')
pdfjsLib.getDocument('helloworld.pdf')
```
Remember though that PDF.js uses promises, so the above will return a promise that is resolved with the document object.
Remember though that PDF.js uses promises, and the above will return a `PDFDocumentLoadingTask` instance that has a `promise` property which is resolved with the document object.
Each PDF page has its own viewport which defines the size in pixels(72DPI) and initial rotation. By default the viewport is scaled to the original size of the PDF, but this can be changed by modifying the viewport. When the viewport is created an initial transformation matrix will also be created that takes into account the desired scale, rotation, and it transforms the coordinate system (the 0,0 point in PDF documents the bottom-left whereas canvas 0,0 is top-left).
Each PDF page has its own viewport which defines the size in pixels(72DPI) and initial rotation. By default the viewport is scaled to the original size of the PDF, but this can be changed by modifying the viewport. When the viewport is created, an initial transformation matrix will also be created that takes into account the desired scale, rotation, and it transforms the coordinate system (the 0,0 point in PDF documents the bottom-left whereas canvas 0,0 is top-left).
```js
varscale=1.5;
varviewport=page.getViewport(scale);
varviewport=page.getViewport({scale:scale,});
varcanvas=document.getElementById('the-canvas');
varcontext=canvas.getContext('2d');
@@ -57,9 +61,29 @@ Alternatively, if you want the canvas to render to a certain pixel size you coul
@@ -21,7 +21,7 @@ Before downloading PDF.js please take a moment to understand the different layer
<tbody>
<tr>
<td>Core</td>
<td>The core layer is where a binary PDF is parsed and interpreted. This layer is the foundation for all subsequent layers. It is not documented here because using it directly is considered an advanced usage and the API is likely to change. For an example of using the core layer see the [PDF Object Browser](https://github.com/brendandahl/pdf.js.utils/tree/master/browser)
<td>The core layer is where a binary PDF is parsed and interpreted. This layer is the foundation for all subsequent layers. It is not documented here because using it directly is considered an advanced usage and the API is likely to change. For an example of using the core layer see the <a href="https://github.com/brendandahl/pdf.js.utils/tree/master/browser">PDF Object Browser</a>.
</td>
</tr>
<tr>
@@ -30,7 +30,7 @@ Before downloading PDF.js please take a moment to understand the different layer
</tr>
<tr>
<td>Viewer</td>
<td>The viewer is built on the display layer and is the UI for PDF viewer in Firefox and the other browser extensions within the project. It can be a good starting point for building your own viewer. *However, we do ask if you plan to embed the viewer in your own site, that it not just be an unmodified version. Please re-skin it or build upon it.*</td>
<td>The viewer is built on the display layer and is the UI for PDF viewer in Firefox and the other browser extensions within the project. It can be a good starting point for building your own viewer. <em>However, we do ask if you plan to embed the viewer in your own site, that it not just be an unmodified version. Please re-skin it or build upon it.</em></td>
</tr>
</tbody>
</table>
@@ -38,77 +38,95 @@ Before downloading PDF.js please take a moment to understand the different layer
## Download
<div class="row">
<div class="col-md-6">
<h3>Pre-built</h3>
<div class="col-md-4">
<h3>Prebuilt</h3>
<p>
Includes the generic build of PDF.js and the viewer.
└── package.json - package definition and dependencies
```
## Trying the Viewer
With the prebuilt or source version open `web/viewer.html` in a browser and the test pdf should load. Note: the worker is not enabled for file:// urls, so use a server. If you're using the source build and have node, you can run `node make server`.
With the prebuilt or source version, open `web/viewer.html` in a browser and the test pdf should load. Note: the worker is not enabled for file:// urls, so use a server. If you're using the source build and have node, you can run `gulp server`.
## More Information
For a further walkthrough of a minimal viewer see the [hello world example](/examples/). More documentation can be found in our [wiki](https://github.com/mozilla/pdf.js/wiki) too.
For a further walkthrough of a minimal viewer, see the hello world example. More documentation can be found in our [wiki](https://github.com/mozilla/pdf.js/wiki) too.
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Tim van der Meij
