The general pattern for an input next to an icon-only button has been
added to the pattern library in the frontend-shared package and is
implemented by the TextInputWithButton component. We don't need custom
treatment in this project anymore.

Use `classes` to extend, not obliterate, component styles, obviating
the need to use complex mixins.

Use `classes` to extend, not obliterate, component styles and obviate
the need for complex mixins.

Use `classes` to extend, not obliterate, components styles and obviate
the need to use complex mixins.

Use `classes` to extend, not obliterate, component styles, and obviate
the need to use complex mixins.

This is functionally the same, just a little cleaner.

The rendering of SVG highlights in PDFs triggered an excessive number of
style recalculations and forced reflows due to interleaving DOM changes
with DOM API calls that require current layout information.

This commit makes several changes to reduce reflows:

 1. Create all the `<hypothesis-highlight>` elements in the text layer
    before creating any of the associated SVG elements.

    This avoids interleaving creating `<hypothesis-highlight>` elements
    with measuring them in order to position the associated SVG
    `<rect>`.

 2. Change `drawHighlightsAbovePdfCanvas` to accept a list of highlight
    elements instead of a single element. By assuming all highlights for
    a single annotation are on the same page, we can avoid some repeated
    work and add all the `<rect>` elements to the parent `<svg>` in one
    batch.

These changes only reduce the thrash when creating highlights for a
single annotation. To eliminate the remaining thrash we'd need to use a
combination of optimizing DOM updates across multiple `highlightRange`
calls and/or reducing the work for PDF pages which are rendered but not
visible on-screen.

Combine two separate if blocks into one.

See https://esbench.com/bench/613f1ca06c89f600a570114b for some related
benchmarks.

To reduce the chances of mis-anchoring quotes, check for a context match
before stopping early. This helps when the quote is a word or phrase
that is common in the document and the position hint is wrong, which can
occur if the quote selector was created when using a different PDF
viewer or version of PDF.js, which exacts text differently.

For quotes captured at the start or end of a page we expect a mismatch
in the prefix and suffix respectively because quote selectors can currently
capture parts of the surrounding pages / DOM in `describe`, but when
anchoring we only search for matches in one page of text at a time.

We have a basic test of quote matching from the perspective of the
`anchor` caller, but we really need to check the data passed to the
quote matching logic in more detail.

As described in #3705, stopping a quote search as soon as we find an
approximate match could result in quotes anchoring to suboptimal matches
if the position selector is no longer accurate.

We want to balance finding the best match against making the search
efficient in long documents. The approach taken in this commit is to
continue to search pages in priority order, but instead of returning the
first "good enough" match (below `matchQuote`'s error threshold), return
the first exact match ignoring whitespace differences, or the nearest
match in the whole document otherwise.

We ignore whitespace differences because this is the main change between
the current version of PDF.js that we include in Hypothesis products, and
newer versions that we are trying to migrate to (see #3687). It is also
likely to be the main source of text differences between PDF.js and
other viewers.

 - Change `anchorQuote` to ignore whitespace differences when searching
   for approximate matches
 - Change `anchorQuote` to keep searching for the best match if it does
   not find a whitespace-insensitive exact match in the current page
 - Add tests for whitespace changes

Fixes https://github.com/hypothesis/client/issues/3705

Add tests to cover the cases where the position is on a page before the
quote and where the position is negative.

Refactor quote anchoring logic to use async/await instead of Promise
chains to make control flow easier to follow and change.

 - Convert `getPageOffset` and `findPage` helper functions to use
   async/await rather than Promise chains.

 - Combine the `prioritizePages` and `findInPages` functions into a
   single `anchorQuote` function.

   Combining the two functions allows some redundant work to be
   eliminated. `findInPages` used to call `getPageOffset` for the
   current page on each iteration which would in turn fetch the text of
   every page until it reached the given offset. `prioritizePages` would
   do a very similar loop in order to enumerate pages close to the given
   offset.

 - Replace use of `TextQuoteAnchor.fromSelector` with direct use of
   `matchQuote`. `fromSelector` wasn't providing any value over just
   calling `matchQuote` directly, and it required a dummy DOM element to
   be created and have its text populated.

In scss files, `@use "..."` has been replaced by `@use '...'" (single
quotes instead of double).

Bumps [prettier](https://github.com/prettier/prettier) from 2.3.2 to 2.4.0.
- [Release notes](https://github.com/prettier/prettier/releases)
- [Changelog](https://github.com/prettier/prettier/blob/main/CHANGELOG.md)
- [Commits](https://github.com/prettier/prettier/compare/2.3.2...2.4.0)

---
updated-dependencies:
- dependency-name: prettier
  dependency-type: direct:development
  update-type: version-update:semver-minor
...
Signed-off-by: dependabot[bot] <support@github.com>

Bumps [typescript](https://github.com/Microsoft/TypeScript) from 4.4.2 to 4.4.3.
- [Release notes](https://github.com/Microsoft/TypeScript/releases)
- [Commits](https://github.com/Microsoft/TypeScript/compare/v4.4.2...v4.4.3)

---
updated-dependencies:
- dependency-name: typescript
  dependency-type: direct:development
  update-type: version-update:semver-patch
...
Signed-off-by: dependabot[bot] <support@github.com>

Bumps [@babel/preset-env](https://github.com/babel/babel/tree/HEAD/packages/babel-preset-env) from 7.15.4 to 7.15.6.
- [Release notes](https://github.com/babel/babel/releases)
- [Changelog](https://github.com/babel/babel/blob/main/CHANGELOG.md)
- [Commits](https://github.com/babel/babel/commits/v7.15.6/packages/babel-preset-env)

---
updated-dependencies:
- dependency-name: "@babel/preset-env"
  dependency-type: direct:development
  update-type: version-update:semver-patch
...
Signed-off-by: dependabot[bot] <support@github.com>

Bumps [sass](https://github.com/sass/dart-sass) from 1.39.0 to 1.39.2.
- [Release notes](https://github.com/sass/dart-sass/releases)
- [Changelog](https://github.com/sass/dart-sass/blob/main/CHANGELOG.md)
- [Commits](https://github.com/sass/dart-sass/compare/1.39.0...1.39.2)

---
updated-dependencies:
- dependency-name: sass
  dependency-type: direct:development
  update-type: version-update:semver-patch
...
Signed-off-by: dependabot[bot] <support@github.com>

Bumps [aws-sdk](https://github.com/aws/aws-sdk-js) from 2.983.0 to 2.987.0.
- [Release notes](https://github.com/aws/aws-sdk-js/releases)
- [Changelog](https://github.com/aws/aws-sdk-js/blob/master/CHANGELOG.md)
- [Commits](https://github.com/aws/aws-sdk-js/compare/v2.983.0...v2.987.0)

---
updated-dependencies:
- dependency-name: aws-sdk
  dependency-type: direct:development
  update-type: version-update:semver-minor
...
Signed-off-by: dependabot[bot] <support@github.com>

Bumps [core-js](https://github.com/zloirock/core-js) from 3.17.2 to 3.17.3.
- [Release notes](https://github.com/zloirock/core-js/releases)
- [Changelog](https://github.com/zloirock/core-js/blob/master/CHANGELOG.md)
- [Commits](https://github.com/zloirock/core-js/compare/v3.17.2...v3.17.3)

---
updated-dependencies:
- dependency-name: core-js
  dependency-type: direct:development
  update-type: version-update:semver-patch
...
Signed-off-by: dependabot[bot] <support@github.com>

Bumps [commander](https://github.com/tj/commander.js) from 8.1.0 to 8.2.0.
- [Release notes](https://github.com/tj/commander.js/releases)
- [Changelog](https://github.com/tj/commander.js/blob/master/CHANGELOG.md)
- [Commits](https://github.com/tj/commander.js/compare/v8.1.0...v8.2.0)

---
updated-dependencies:
- dependency-name: commander
  dependency-type: direct:development
  update-type: version-update:semver-minor
...
Signed-off-by: dependabot[bot] <support@github.com>

Bumps [@hypothesis/frontend-shared](https://github.com/hypothesis/frontend-shared) from 3.8.1 to 3.9.0.
- [Release notes](https://github.com/hypothesis/frontend-shared/releases)
- [Changelog](https://github.com/hypothesis/frontend-shared/blob/main/CHANGELOG.md)
- [Commits](https://github.com/hypothesis/frontend-shared/compare/v3.8.1...v3.9.0)

---
updated-dependencies:
- dependency-name: "@hypothesis/frontend-shared"
  dependency-type: direct:development
  update-type: version-update:semver-minor
...
Signed-off-by: dependabot[bot] <support@github.com>

In https://github.com/mozilla/pdf.js/pull/13661 the API for retrieving
PDF fingerprints (aka. File Identifiers) changed. In generic builds of
PDF.js the old API remains available, but not in the non-generic one
that Firefox's built-in viewer uses.

This commit makes Hypothesis use the new API if available or the old API
otherwise. The fingerprint value should be the same in both cases.

Fixes #3673

This is the result of running scripts/update-pdfjs.

Use async/await to simplify a test.

* rename `embedElement` to `bootScript`
* remove unused class

* Created a helper function to test the addition of the Hypothesis
  script
* Move all the helper functions before the `beforeEach`

`FrameObserver` could be disconnected while waiting for the loading of
the iframe. We add a mechanism to avoid this race condition.

Move documentation to the FrameObserver class.

As suggested here: https://github.com/hypothesis/client/pull/3670#discussion_r686665805

`_removeFrame` must be called in these three scenarios:

- `enable-annotation` attribute is removed from iframe
- `src` attribute is modified in the iframe
- iframe is deleted

When the iframe is deleted there are two possible paths for the execution of
the `_removeFrame`:

1. First and faster execution of `_removeFrame`: `iframe.remove()`
   triggers the `unload` event which calls the `_removeFrame`

2. Second and delayed execution of `_removeFrame`: `iframe.remove()`
   triggers the `MutationObserver` (debounced by 40 ms). This could
   cause the `_removeFrame` to be fired if the first path would not
   remove the iframe from the list of `_handledFrames`.

I moved the addition and deletion of the iframes to `_handledFrames` as
earlier as possible in the `_addFrame` and `_removeFrame` methods to
avoid racing conditions. A consequence of this is that the `_addFrame`
is executed only once per iframe. If it fails (for example, because the
iframe is from a different origin) it is not constantly retried.

The `FrameObserver` unit test now exercise the whole class, not just the
`findFrames` function.

On a future commit, we will inline the `findFrames functionality.

`onDocumentReady` returns now a Promise instead of calling a callback.

The name of `isDocumentReady` gives the impressiong that the function
would return a boolean.  `onDocumentReady`, on the other hand, expresses
more clearly that the execution of the callback will be triggered after
the iframe's documen is ready (loaded and parsed).

On a follow up commit I change the `onDocumentReady` to return a Promise
instead of executing a callback.

`Set` is a better data structure because it has a faster search: `Set.has` has a O(1) while `Array.includes` has a O(n).