html-parse-stringify

First release from the package's new maintenance home. Functionally builds on 3.1.0; see CHANGELOG.md for the full picture including breaking changes and migration notes.

Breaking: exports map (no deep dist/ imports), boolean attributes are null in the AST and stringify bare, doctype parses as a void node, <script>/<style> content is raw text.

Added: allowedTags parse option, named exports, rewritten TypeScript types (discriminated union incl. CommentNode), zero runtime dependencies, CI with a react-i18next integration gate.

Fixed: comments containing >, unquoted attribute values containing =, double-quote escaping in stringify, and parse no longer mutates the caller's options object.

First release from the package's new maintenance home: i18next/html-parse-stringify. Functionally it builds on 3.1.0; the breaking changes below are deliberate and small.

Breaking

• Modern package entry points: exports map with dual ESM/CJS builds and bundled types. Deep imports like html-parse-stringify/dist/... no longer work; import the package root instead.
• Boolean attributes (e.g. disabled) now carry the value null in the AST (was '') and stringify bare: <input disabled/> instead of <input disabled=""/>.
• <!DOCTYPE html> parses as a void node and round-trips correctly (#54). Previously the rest of the document was nested inside it and stringify produced </!DOCTYPE>.
• <script> and <style> content is treated as raw text: the whole content becomes a single text child, and </> inside it no longer produce bogus tag nodes.

Fixed

• Comments containing > (e.g. <!-- a > b -->) are now parsed fully instead of being cut at the first >.
• Unquoted attribute values containing = are no longer truncated at the second = (<div data-x=a=b> now keeps a=b).
• stringify escapes double quotes in attribute values as ", so values parsed from single-quoted or multiline attributes can't produce broken markup.
• parse no longer mutates the caller's options object.

Added

• allowedTags parse option (array of names or predicate): only listed tags are parsed as markup, everything else tag-shaped stays literal text. This is the parser-level version of the escaping that consumers like react-i18next had to implement themselves.
• Named exports: import { parse, stringify } from 'html-parse-stringify' (the default export remains for compatibility).
• Rewritten TypeScript types with a discriminated union (TagNode | TextNode | CommentNode | ComponentNode) (#56). CommentNode was missing entirely before.
• Zero runtime dependencies: the former void-elements dependency is inlined.
• CI (lint, tests, build on Node 18–24) plus an integration job that runs the full react-i18next test suite against this parser before anything ships.