Changelog

v0.4.2 — 2026-09-18

Bug Fixes

Render aligned blocks as Markdown islands and indent nested lists relatively (#86)

Full Changelog: https://github.com/discourse/markbridge/compare/v0.4.1...v0.4.2

v0.4.1 — 2026-09-03

Bug Fixes

Always escape setext heading underlines (#82)

Full Changelog: https://github.com/discourse/markbridge/compare/v0.4.0...v0.4.1

v0.4.0 — 2026-07-28

Features

Add a block flag to AST::Code (#72)
Match AST subclasses by ancestry in rules and tag dispatch (#72)
Add HTML heading parsing and the shared html_mode example (#70)

Bug Fixes

Improve code language detection and drop empty element output (#70)

Full Changelog: https://github.com/discourse/markbridge/compare/v0.3.1...v0.4.0

v0.3.1 — 2026-07-20

Features

Add an AST normalization pass for target-format nesting rules (#67)

Full Changelog: https://github.com/discourse/markbridge/compare/v0.3.0...v0.3.1

v0.3.0 — 2026-07-09

Features

Expose the bare-link judgment as AST::Url#bare? (#62)
Let tag overrides fall back to the stock rendering (#62)
Keep bare and relative URLs intact in UrlTag (#62)

Bug Fixes

Model quote attributions with honest names and types (#62)

Performance

BBCode and MediaWiki now convert roughly twice as fast, HTML about 1.5×, with allocations down 3–9.5× depending on the format. Multibyte text no longer carries a penalty — scanning non-ASCII BBCode was up to 5× slower than ASCII and is now at parity.

Trim TextFormatter per-parse and per-element overhead (#61)
Cut the HTML tree walk’s per-node overhead (#61)
Build the MediaWiki inline tag registry only once (#60)
Convert the MediaWiki inline parser to byte-offset jump scanning (#60)
Skip the line split for single-line text in MarkdownEscaper (#59)
Turn RenderContext into a linked parent chain (#59)
Cut small allocations in the parse and render hot paths (#58)
Convert the BBCode scanner to byte offsets (#58)
Build the default handler registry and tag library only once (#58)

Full Changelog: https://github.com/discourse/markbridge/compare/v0.2.1...v0.3.0

v0.2.1 — 2026-07-06

Removed

Removed the Markbridge::Processors::DiscourseMarkdown module — the scanner and the mention/poll/event/upload detectors. This code is moving into the Discourse migrations tooling. (#56)

Full Changelog: https://github.com/discourse/markbridge/compare/v0.2.0...v0.2.1

v0.2.0 — 2026-06-11

Breaking Changes

The migration API was redesigned around structured result types (#30). UPGRADING.md has before/after examples for every change.

parse_* methods now return a Parse and *_to_markdown / convert return a Conversion instead of raw Strings — use Conversion#markdown (or #to_s) for the output (#30)
Global configuration is gone: Markbridge::Configuration, .configure, .reset_defaults!, and all default_* accessors were removed — build a reusable renderer with Markbridge.discourse_renderer(...) and pass it via renderer: (#30)
tags:, tag_library:, escaper:, and escape_hard_line_breaks: were removed from the per-call signatures; *_to_markdown / convert accept only handlers:, renderer:, and raise_on_error: (#30)
Proc/lambda handlers are no longer supported — handlers must be objects responding to #process(...) (#30)
MediaWiki parser kwarg renamed: inline_tag_registry: → handlers: (#30)
TextFormatter handlers must accept a processor: keyword (#30)

Features

Markbridge.discourse_renderer(...) factory builds reusable Renderer instances — accepts tags:, tag_library:, unregister:, escaper:, escape:, escape_hard_line_breaks:, allow:, postprocessor:, strip_trailing_invisibles: (#30)
Markbridge.convert(format:) dispatcher and polymorphic Markbridge.render(Parse | AST::Node) (#30)
AST mutation between parse and render: block form on every *_to_markdown / convert, plus AST::Element#each_descendant, #descendants(klass = nil), and #replace_child(old, new) (#30)
raise_on_error: false collects render errors on Conversion#errors for per-row failure isolation in batch migrations (#30)
Selective Markdown escaping via MarkdownEscaper#allow: (:bullet_list, :ordered_list, :atx_heading, :block_quote, alias :lists), and IdentityEscaper / discourse_renderer(escape: false) for already-trusted Markdown (#30)
Handler registries gain #overlay(name) { |previous| ... } for wrapping default handlers (BBCode, HTML, TextFormatter) (#30)
TagLibrary#unregister — unregistered AST classes auto-pass-through to their rendered children (#30)
HTML and TextFormatter parsers accept pre-parsed Nokogiri input (DocumentFragment, Document, Element) in addition to Strings (#30)
AST::Details + DetailsTag for Discourse [details=…]…[/details] collapsible sections (#30)
Optional strip_trailing_invisibles config flag (#39)
Collapse whitespace in HTML per spec (#36)
Map inline span styles to AST formatting nodes (#34)