fix: handle quotes & backticks in JSX in directive scanner

[mogelbrod]

The comment-directive scanner could silently drop lingui-set/lingui-reset directives when a quote ('/") or backtick appeared as JSX text above it. When that happened, directives after the stray character were never registered, so message IDs were generated against the wrong (or a stale) idPrefix.

For example, with idPrefixLeader: ".":

// lingui-set idPrefix="root"
const X = () => <p>'</p>
const Y = () => <p>`</p>

// lingui-set idPrefix="different"
const table = {
  a: msg({ id: '.a', message: `A` }),  // expected "different.a"
}

.a is incorrectly compiled to root.a instead of different.a.

Root cause

scan_source_comments recovers comments from raw source text while skipping over '…', "…", and … regions to avoid false positives.

A quote or backtick in JSX text (<p>'</p>) is content, and not a string literal opener delimeter. The scanner treated it as one and scanned ahead to the next matching delimiter — e.g. the ' in id: '.a' — swallowing every comment in between, including the directive.

Fix

scan_source_comments is now a small context-stack state machine (Code/JsxTag/JsxChildren). Inside JSX text, quotes/backticks/comment markers are treated as literal text; {…} expression containers correctly re-enter code context, so real literals and comments inside them are still handled. This unfortunately also requires adding parsing of JSX and TS type annotations, such as:

• A < opens JSX only in expression position (tracked via the preceding significant byte) followed by a tag-name char or > (fragment). This matches .tsx semantics and prevents TS generics (Record<string, …>) and comparisons (a < b) from being mistaken for JSX. A ,/; inside a tag bails back to code, covering the <T,> generic-arrow form.
• Expression keywords. JSX following keywords like return <div>, export default <App/>, and yield <x> is detected by inspecting the full preceding word, while value identifiers that merely end in keyword letters (myreturn<x) are not.
• skip_string_literal now bails when it hits an unescaped newline, since single/double-quoted JS strings can't span newlines — a cheap, independent guard against a stray quote running away.

[codecov]

Codecov Report

❌ Patch coverage is 97.30769% with 7 lines in your changes missing coverage. Please review.
✅ Project coverage is 95.10%. Comparing base (300cf86) to head (509cff4).

Files with missing lines	Patch %	Lines
...ngui_macro/src/comment_directive/source_scanner.rs	97.30%	7 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #238      +/-   ##
==========================================
+ Coverage   94.95%   95.10%   +0.15%     
==========================================
  Files          10       10              
  Lines        2597     2821     +224     
==========================================
+ Hits         2466     2683     +217     
- Misses        131      138       +7     

Flag	Coverage Δ
unittests	95.10% <97.30%> (+0.15%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
...ngui_macro/src/comment_directive/source_scanner.rs	98.44% <97.30%> (-1.56%)	⬇️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

This PR fixes the Lingui directive comment scanner so it no longer “runs away” when it encounters quotes ('/") or backticks (`) inside JSX text nodes, which previously caused lingui-set / lingui-reset directives to be silently skipped and the wrong idPrefix to be applied.

Changes:

• Reworked scan_source_comments into a small context-stack scanner (Code / JsxTag / JsxChildren) so JSX text treats quotes/backticks and ////* as literal text.
• Added regression coverage: a new Lingui macro snapshot + fixture for JSX containing ' and ` above directives.
• Strengthened string scanning by bailing out of single/double-quoted string skipping on unescaped newlines.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 2 comments.

File	Description
crates/lingui_macro/src/comment_directive/source_scanner.rs	Implements JSX-aware comment scanning and adds targeted unit tests for the previously failing cases.
crates/lingui_macro/tests/lingui_directive.rs	Adds a regression test case exercising directives following JSX text with quotes/backticks.
crates/lingui_macro/tests/snapshots/lingui_directive__jsx_trans_with_unclosed_quotes.snap	New snapshot verifying idPrefix is applied correctly after JSX text containing ' and `.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[timofei-iatsenko]

I'm thiniking that is going way too far, what if we will simplify this scanner to a hardcoded search of a keywords? Just looking for the lingui-set / lingui-reset and once this sequence is located - read everethying following it while it conforms the schema key="value"

I think we can ignore cases when someone could put a lingui directive inside of a string literal, this is an extreme edge case. WDYT?

[mogelbrod]

I'm tarting to feel that way too. How about a regex like the following to limit results to those that appear to be within comments?

const re = /\/(?:\/|\*\*?)\s*lingui-(set|reset)[ ]+([^\n]*)/

[mogelbrod]

Submitted the alternative implementation in #240

[timofei-iatsenko]

closed in favor of #242