Markdown parsing now emits fewer, cleaner nodes (
dbe05f4)
The parser refactor reduces emitted syntax nodes and rewires several Markdown lexer/parser paths, with large snapshot updates across links, autolinks, emphasis, fenced code, and HTML-in-list cases. This should simplify downstream formatting and make the Markdown IR closer to the actual structure being parsed.
HTML blocks in Markdown are parsed as a single content node (f852f65)
HTML block handling was changed so the parser/factory emits one MD_HTML_CONTENT node instead of a broader inline-item list, and the formatter was updated to respect line terminators around HTML/fenced blocks inside lists. That fixes a class of list/HTML formatting edge cases and reduces churn in the Markdown AST.
Markdown formatter adds blockquote, link, and thematic-break support (e324b89)
This formatter release adds substantial handling for blockquotes, inline links/images, link reference definitions, and thematic breaks, with list-marker normalization adjusted to avoid turning valid lists into thematic breaks. In practice, this improves CommonMark compatibility for a wide set of Markdown constructs.
noUnusedVariables stops flagging valid Svelte patterns (
da9b403)
The linter now recognizes Svelte store subscriptions and $bindable() props as intentional uses, avoiding false positives in noUnusedVariables. This is a targeted but important correctness fix for Svelte users.
Other misc changes
- Markdown formatter/parser snapshot updates and generated code refreshes
- Internal embedded-reference plumbing for Svelte linting