docx_to_markdown 0.2.0

Convert Word .docx (OOXML) files to GitHub-Flavored or CommonMark Markdown in pure Dart, with images, tables, lists, footnotes, math, and extensible hooks.

Changelog

0.2.0

This release changes the Markdown produced for documents that already converted; review the notes below before upgrading pinned output.

• Fixed emphasis that Word pads with whitespace rendering as literal asterisks: a bold "1. " produced **1. **, which no CommonMark or GFM parser treats as bold. Whitespace is now hoisted outside the delimiters (**1.** ), and a span holding only whitespace drops its delimiters entirely. Affects bold, italic, strikethrough, and ++underline++.
• Stopped re-emitting a paragraph style's own typography as inline emphasis once that style has become a Markdown block marker: a Heading1 whose style carries <w:b/> now yields # Title instead of # **Title**. Quote styles and definition terms are affected the same way. Bold and italic that the author applied directly, or through a character style, are unchanged. Wrap HeadingBlock.inlines in a StrongInline from transformBlock to restore the previous rendering.
• Unified monospace detection across runs and styles, which previously consulted separate font lists, so Liberation Mono was inline code through a character style but plain text through a direct w:rFonts. The monospaceFonts default now covers the families modern exports use - Liberation Mono, DejaVu Sans Mono, JetBrains Mono, Fira Code, IBM Plex Mono, Cascadia Code, SF Mono, Roboto Mono, Inconsolata, Hack, Iosevka, and more - and matches on whole words, so Monotype Corsiva is no longer taken for a monospace face. Expect more text to render as inline code.
• Fixed merged-cell tables, and every table under TableMode.htmlOnly, losing all inline formatting: a second HTML renderer inside the parser flattened cells to plain text, discarding bold, italic, links, and images, emitted <td> for header cells, and appended a stray <br/> to every cell. The parser now always produces a TableBlock carrying real colSpan and rowSpan, so transformBlock receives a TableBlock where it previously received an HtmlBlock for those tables.
• Stopped treating a monospace document body font as a code signal, which turned entire documents into inline code; detection now requires contrast with the document's default font.
• Stopped letting the complex-script font slot (w:cs) trigger code detection, so an Arabic fallback of Courier New no longer renders a Calibri paragraph as inline code.
• Dropped tables whose cells carry no content, such as the 1x1 shaded table Word uses as a decorative rule, which previously emitted an empty <table>. A cell counts as content when it holds text, an image, a link, a bookmark anchor, a footnote or endnote reference, a comment, math, a page break, or a non-empty nested table. Set preserveEmptyTables: true to keep them; each drop reports a new table.emptyDropped warning.
• Fixed line-start escaping emitting a visible backslash: a paragraph beginning 1. text was escaped as \1., but CommonMark honours a backslash only before ASCII punctuation. The delimiter now carries the escape (1\. text), and headings are no longer escaped at all.
• Stopped reporting OOXML property and marker elements - w:tcPr, w:tblPr, w:trPr, w:proofErr, and the range markers - as unsupported content. A real Word document raised thousands of spurious unsupported.block warnings, and with emitWarningsAsHtmlComments: true those became thousands of comment lines in the output.
• Added preserveEmptyTables and monospaceMatcher to DocxToMarkdownConfig, hasRenderableContent to Block, Inline, TableCell, and TableGrid, and exported kDefaultMonospaceFonts, MonospaceFontMatcher, and normalizeFontName.
• Added a browser-based converter demo at omar-hanafy.github.io/docx_to_markdown.

0.1.0

• Updated the minimum Dart SDK requirement to 3.12.
• Added document metadata parsing with optional Pandoc YAML front matter output.
• Added definition-list parsing and configurable HTML, Pandoc, or paragraph rendering.
• Added web-safe image export via imageAssetSink, plus explicit image opt-out with extractImages: false.
• Added rendering controls for highlights, text color, page breaks, tracked changes, source ordered-list markers, metadata, and image dimensions.
• Expanded DOCX fidelity for comments, endnotes, headers and footers, task lists, internal bookmarks, VML images, text boxes, horizontal rules, complex tables, nested lists, and table-cell footnotes.
• Improved formatting fidelity for style-inherited run properties, character styles, complex-script bold and italic, underline/color/highlight composition, same-run images and notes, and explicit shading resets.
• Improved OMML math fallback rendering for common structures, including superscripts, subscripts, radicals, n-ary operators, matrices, delimiters, and accents.
• Improved hooks and integration APIs with full IR exports and richer warning, link, image, and OMML context.
• Improved package validation and relationship handling for malformed DOCX input, non-DOCX ZIPs, media, links, and metadata parts.

0.0.1

• Initial release of docx_to_markdown.
• Supports parsing DOCX files into a structured Intermediate Representation (IR).
• Renders Markdown (GFM or CommonMark) with configurable options.
• Features:
  • Paragraphs, Headings (1-6), Blockquotes.
  • Lists (ordered, unordered, nested, mixed).
  • Tables (Markdown pipes or HTML fallback for merged cells).
  • Images (extraction, resizing syntax support).
  • Text formatting (bold, italic, strikethrough, underline, sub/superscript).
  • Links and Footnotes.
  • Math equations (OMML to LaTeX hooks).
  • Code blocks (fenced, with language detection).
• Configurable hooks for custom transformations.