Configure Word-Level Diff for Prose

Word-Level Diffs for Documentation

Standard line-level diffs are designed for code where each line is a distinct statement. For prose — Markdown, plain text, LaTeX — a single word change in a paragraph shows the entire paragraph as modified. Word-level diffs solve this by highlighting individual word changes.

Setting Up in .gitattributes

*.md   text diff=markdown
*.txt  text diff=text
*.tex  text diff=tex
*.rst  text diff=rst
*.adoc text diff=asciidoc

The diff=markdown driver tells Git to use Markdown-aware hunk headers (showing the nearest heading), and enables better word-boundary detection.

Using git diff --word-diff

With the attributes set, you can use:

# Color-coded word diff
git diff --word-diff

# Porcelain format (for scripts)
git diff --word-diff=porcelain

# Plain format with delimiters
git diff --word-diff=plain

Output Example

Standard diff:

-The quick brown fox jumps over the lazy dog.
+The quick red fox leaps over the lazy cat.

Word diff:

The quick [-brown-]{+red+} fox [-jumps-]{+leaps+} over the lazy [-dog.-]{+cat.+}

Custom Word Regex

You can define what constitutes a "word" in .gitconfig:

[diff "markdown"]
  wordRegex = [[:alnum:]_]+|[^[:space:]]

This treats each alphanumeric sequence and each non-space character as a word, producing finer-grained diffs for Markdown.

Automatic Word Diff for Specific Files

To always show word-level diffs for certain files without the --word-diff flag, add to .gitconfig:

[diff "prose"]
  wordRegex = [[:alnum:]_]+|[^[:space:]]
  command = "git diff --word-diff"

Then in .gitattributes:

*.md diff=prose