README.md

A diff tool that provides line and token level colorization.

Based on the Patience Diff algorithm described by Bram Cohen and then expanded upon by James Coglan in two blogs posts (algorithm and implementation).

Usage

❯ pratdiff --help
Diff files with token level colorization

Usage: 
    pratdiff [OPTIONS] <OLD_FILE> <NEW_FILE>
        Diff two files or directory trees.

    <diff tool> | pratdiff [OPTIONS]
        Rediff an existing diff, adding pratdiff goodness.

        Examples:
          git diff | pratdiff
          diff -u old.txt new.txt | pratdiff --cluster

Arguments:
  [OLD_FILE]  Path to old file, directory tree, or `-` for stdin
  [NEW_FILE]  Path to new file, directory tree, or `-` for stdin

Options:
  -c, --context <NUM>
          Display NUM lines of unchanged context before and after changes [default: 3]
  -v, --verbose-paths
          Print full paths instead of stripping a common prefix
      --color <COLOR>
          [default: auto] [possible values: auto, always, never]
      --cluster
          Group diffs into clusters by change signature
      --ignore-whitespace <IGNORE_WHITESPACE>
          How to treat whitespace while diffing [default: length-changes] [possible values: length-changes, all, no]
      --completions <SHELL>
          The shell to generate the completions for [possible values: bash, elvish, fish, nushell, powershell, zsh]
  -h, --help
          Print help
  -V, --version
          Print version

FAQ

How do I install pratdiff?

Use cargo install pratdiff. You probably want to get cargo from rustup or brew.

How do I enable autocompletions?

The --completions flag takes a shell and outputs a completion script.

eval "$(pratdiff --completions=bash)"

pratdiff --completions=fish | source

Why did you bother doing this?

Cause I wanted a learning project and this seemed like a reasonable one.

Why did you name it pratdiff?

I wanted to insert an "r" into patdiff, and I kind of like "prat" as an oddly out of date insult.

Did you learn anything interesting?

The way that token level diffing uses the same algorithm as the line level diffing is pretty cool in my mind. I didn't think going into it that I would structure it that way and it all kinda just fell out.

Also, I learned that the tiny extension on patience diff I made to use non-unique lines if unique ones fail is a known algorithm called "histogram diff".