For simplicity we will abort /Form XObject parsing *immediately* when encountering a circular reference, rather than letting it continue up until some limit (as e.g. PDFium appears to do), which should be fine since there are never any guarantees if/how *corrupt* PDF documents will render.
In rare cases /Resources are also found in the /Contents stream-dict, in addition to in the /Page dict, hence we need to prefer those when available; see `issue18894.pdf`.
Previously we'd only do this for Type1/CFF fonts, see e.g. PR 6736, since the font-program may update the /FontMatrix.
However, it seems that we should do this unconditionally to account for fonts with non-default /FontMatrix-entries in the font-dictionary (which seem to be pretty rare).
In PDF version 2.0 the handling of Indexed color spaces was clarified as follows:
> The index value should be an integer in the range 0 to hival. If the value is a real number, it shall be rounded to the nearest integer (0.5 values shall be rounded up); if it is outside the range 0 to hival, it shall be adjusted to the nearest value within that range.
Please refer to https://github.com/pdf-association/pdf-differences/tree/main/IndexedColor
After the introduction of `OffscreenCanvas` support we now have *two separate* mask-methods in the `PDFImage` class, and the reason that they were not combined is likely that we need the "raw" bytes when parsing Type3-glyph image masks.
However, that case is easy to support simply by disabling `OffscreenCanvas` usage when parsing Type3-glyphs and that way we're able to reduce some code duplication.
Another slightly strange property of the `PDFImage.createMask` method is that it needs various image-dictionary parameters *manually* provided, which is probably because this is very old code.
That feels slightly unwieldy, and we instead change the method to pass in the image-stream directly and do the necessary data-lookup internally.
A side-effect of this re-factoring is that we now support using the custom `isSingleOpaquePixel` operator in Type3-glyphs, which shouldn't hurt even though it seems extremely unlikely for that to ever happen in Type3-glyphs.
This disallowd the following types of `export` declaration:
- `export class A {}`/`export function A() {}`
- `export default class A {}`/`export default function A() {}`
- `export let A`/`export const A`/`export var A`
While allowing
- `export { A }`
- `export default A`
The current text layer approach based on absolutely positioned
`<span>` elements by default causes flickering with text selection,
and we have browser-specific workarounds to solve that.
In Chrome, the workaround involves moving the `.endOfContent` element to
right after the last element that contains some selected content. This
works well in simple PDFs, but breaks when we have `span.markedContent`
elements. Given a text layer structure like the following, rendered
as four consecutive lines:
```html
<span class="markedContent">
<br>
<span>development enter the construction phase (estimated at around</span>
</span>
<span class="markedContent">
<br>
<span>300 MEUR).</span>
</span>
<span class="markedContent">
<br>
<span>Kreate's EBITA increased to 2.8 MEUR (Q4'23: 2.7 MEUR) and the</span>
</span>
<span class="markedContent">
<br>
<span>margin rose to 3.7% (Q4'23: 3.4%). However, profitability was</span>
</span>
```
when starting to select from inside the first line and dragging down
to the empty space after the second line, Chrome will anchor the
selection at the beginning of either the `<br>` or the `<span>` inside
the last `.markedContent`, depending on whether the selection is in
"per-character mode" (i.e. click and drag) or "per-word mode" (i.e.
double click and drag). This causes us to insert the `.endOfContent`
element in the wrong place (one element too far), which causes one
more line to be selected, which triggers another `"selecctionchange"`
event, which causes us to move `.endOfContent` again, and so on, looping
until when the whole page is selected.
This commit fixes the issue by making sure that when the end of the
selection range points to the _begining_ of an element, we walk back
the dom finding the first non-empty element, and attatch `.endOfContent`
to the end of that.
These `getAll` methods are not used anywhere within the PDF.js code-base, outside of tests, and were mostly added (speculatively) for third-party users.
To still allow access to the same data we instead introduce iterators on these classes, which (slightly) shortens the code and allows us to remove the `objectFromMap` helper function.
A summary of the changes in this patch:
- Replace the `getAll` methods with iterators in the following classes: `AnnotationStorage`, `Metadata`, and `OptionalContentGroup`.
- Change, and also re-name, `AnnotationStorage.prototype.setAll` into a test-only method since it's not used elsewhere.
- Remove the `Metadata.prototype.has` method, since it's only used in tests and can be trivially replaced by calling `Metadata.prototype.get` and checking if the returned value is `null`.
In order to use the PDF.js library in Node.js environments the `process.getBuiltinModule` functionality must be available, which was released in [version `20.16.0`](https://nodejs.org/en/blog/release/v20.16.0), however we've seen repeated issues filed by users on older `20.x` versions.
We want to iterate through the data in the `computeMD5` function, and `Map`s have "nicer" support for that than generic objects.
(Somewhat recently `Map` performance was improved in Firefox, however this also isn't really performance sensitive code.)
Note that we load all wasm-files manually, however the Emscripten Compiler (emcc) unfortunately generates `URL`s for fallback wasm-file loading.
In the PDF.js build-scripts we work-around that by using suitable Webpack-options, however that apparently doesn't work when third-party users re-bundle our code and we thus try to work-around this by adding "ignore comments" to these `URL`s (similar to how we handle `import`-statements).
It seems that the `@napi-rs/canvas` dependency has *basic* canvas-filter support, whereas the "old" `canvas` dependency didn't, hence we no longer need the Node.js-specific checks in the `src/display/canvas.js` file.
Note that I've successfully tested the [`pdf2png` example](https://github.com/mozilla/pdf.js/tree/master/examples/node/pdf2png) with this patch applied and things appear to work as before.
