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.
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.
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.
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.
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.
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).
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.
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.
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.
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.
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.
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.
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.)
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
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.
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.
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
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 :-)
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.
Since calling `getDocument` with a `PDFDataRangeTransport` argument will always unconditionally override a manually provided `length` argument, see a1a667809f/src/display/api.js (L390-L394), this patch should thus be safe.
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.
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.
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`.
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.
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.
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.
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).
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).
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.
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.
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.
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 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.
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.
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.
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.
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.
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.
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.
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).
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 :-)
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).
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.
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.
