A tool to extract plain text from Microsoft .docx (OOXML) documents, preserving basic formatting cues such as list numbering, indentation, and hyperlink annotations.

This is a Rust rewrite of docx2txt 1.4 (http://docx2txt.sourceforge.net/) by Sandeep Kumar. The original is a Perl script; this rewrite exists so that Git for Windows can drop its Perl dependency while retaining git diff support for .docx files.

Functional. Produces byte-identical output to the Perl original on all test fixtures (both hand-crafted and Word-saved).

docx2txt < input.docx
docx2txt < input.docx > output.txt
cat input.docx | docx2txt

Input is always read from stdin, output is always written to stdout. This is the only interface Git needs for its textconv mechanism.

• Main document body text
• Hyperlink URLs (optionally appended as [HYPERLINK: url])
• List markers (bullet, decimal, letter, roman) with indentation
• Paragraph borders as horizontal rules
• Caps formatting (uppercased in output)
• Superscript cross-references (wrapped in [...])

For the full list of OOXML constructs handled, see AGENTS.md in the parent directory.

• Headers and footers
• Footnotes and endnotes
• Images
• Tables as structured data
• Comments and tracked changes (deleted text is stripped)

cargo build --release

The binary is at target/release/docx2txt.

cargo test

The test suite includes both Word-saved fixtures (committed as binary files in tests/fixtures/) and programmatically generated fixtures that exercise every regex in the conversion pipeline.

GNU General Public License version 3 or later, matching the original docx2txt.