Syntax Highlight in Git Diff

README.md

Hacky syntax highlighting in git diff

Normally git diff would color additions green and deletions red. This is cool, but it would be even cooler if it adds syntax highlighting to those lines. This is a git pager that does so.

It parses the diff output and picks up the SHAs of files with additions and deletions. It uses CodeRay to highlight each file and then it extracts the lines that are shown in the diff. It then uses term/ansicolor to make a gradient from the CodeRay color and the diff color (red for deletion, green for addition) and uses it to replace the original.

I tried using rugged instead of shelling out to git show – it was faster overall, but it did incur a noticeable start up time.

Check out the image below for a demo.

demo.png

git-diff-syntax-highlight.rb

#!/usr/bin/env ruby
require 'coderay'
require 'term/ansicolor'

# --- Adjusting colors -------------------------------------------------------

RGB = Term::ANSIColor::RGBTriple
Attribute = Term::ANSIColor::Attribute
Color = Term::ANSIColor

GREEN = RGB.new(0, 256, 0)
RED   = RGB.new(256, 0, 0)

ADJUSTMENTS = {}
GRADIENTS   = {}

def adjust(text, target)
  open     = adjust_seq("\e[37m", target)
  adjusted = text.chomp.gsub(/\e\[[0-9;]+m/) { |seq| adjust_seq(seq, target) }

  "#{open}#{adjusted}"
end

def adjust_seq(seq, target)
  key = [seq, target]

  ADJUSTMENTS[key] ||=
    begin
      escapes = []

      seq[/\d+(;\d+)*/].split(';').map(&:to_i).each do |number|
        case number
        when 0      then escapes << ["\e[0m", gradient_to(255, target)]
        when 1      then escapes << "\e[1m"
        when 4      then escapes << "\e[4m"
        when 30..37 then escapes << gradient_to(number - 30, target)
        when 40..47 then escapes << gradient_to(number - 30, target, :background)
        else             escapes << seq
        end
      end

      escapes.join('')
    end
end


def gradient_to(num, target, background = false)
  key = [num, target, background]

  GRADIENTS[key] ||=
    begin
      triple = triple(num)
      target = triple.gradient_to(target)[8]
      html   = target.html
      method = background ? "on_#{html}" : html

      Attribute[method].apply
    end
end

def triple(num)
  Attribute[num].to_rgb_triple
end

# --- Parsing git diffs ------------------------------------------------------

FORMATS = {
  'Gemfile' => :ruby,
  'rb' => :ruby,
  'c' => :c,
  'h' => :c,
  'cpp' => :cpp,
  'cxx' => :cpp,
  'hpp' => :cpp,
  'clj' => :clojure,
  'css' => :css,
  'erb' => :erb,
  'go' => :go,
  'java' => :java,
  'js' => :javascript,
  'json' => :json,
  'php' => :php,
  'lua' => :lua,
  'py' => :python,
  'sass' => :sass,
  'scss' => :scss,
  'sql' => :sql,
  'xml' => :xml,
  'yml' => :yaml,
  'yaml' => :yaml,
}

CACHED_OBJECTS = {}

def show(sha, path)
  return [] if sha =~ /^0+$/

  cached = CACHED_OBJECTS.delete sha
  return cached if cached

  format = FORMATS[path[/(\w+)$/, 1]]
  code = `git show #{sha} 2> /dev/null`
  code = File.read(path) if code.empty?
  code = CodeRay.scan(code, format).terminal if format
  CACHED_OBJECTS[sha] = code.lines
end

def process
  new       = nil
  old       = nil
  remaining = nil
  old_file  = nil
  new_file  = nil
  old_hash  = nil
  new_hash  = nil

  while gets
    stripped = $_.gsub(/(\e\[[0-9;]*m)*/, '')
    case stripped
    when /^index (\w+)..(\w+)/
      old_hash = $1
      new_hash = $2
      puts $_
    # @@ -1,4 +1,4 @@
    when /^@@ -(\d+)(?:,(\d+))? \+(\d+)(?:,(\d+))? @@/
      old_start, old_length = $1.to_i - 1, ($2 || 1).to_i
      new_start, new_length = $3.to_i - 1, ($4 || 1).to_i

      remaining = old_length + new_length
      puts $_
    when /^\+\+\+ (.*)$/
      new_file = $1.sub(/^[ab]\//, '')
      old = show old_hash, old_file
      new = show new_hash, new_file
      puts $_
    when /^--- (.*)$/
      old_file = $1.sub(/^[ab]\//, '')
      puts $_
    when /^ /
      puts $_
      next if remaining.nil? || remaining <= 0

      old_start += 1
      new_start += 1
    when /^\+/
      puts adjust("+#{new[new_start]}", GREEN)
      new_start += 1
    when /^-/
      puts adjust("-#{old[old_start]}", RED)
      old_start += 1
    else
      puts $_
    end
  end
end

begin
  process
rescue Errno::EPIPE
  exit 0
end

gitconfig

[core]
	pager = /path/to/diff-syntax-highlight.rb | less

Let's move the discussion here. Here is a fork with syntax highlighting for all of the code, not only the added and removed parts: https://gist.github.com/valo/4a8f4db480c220de215d

The problem is that is becomes difficult to see the added and the remove parts if the syntax highlighting is using red and green.

@skanev, yes, probably a gradient to gray may help. I will experiment with this...

Added gradient to gray and I think it is much better. The unchanged code stays shaded. It looks to me that the original color is quite changed, but I need to doublecheck on daylight conditions :)

I also created a patched version of this based on the gist fork https://gist.github.com/skanev/0eeb943e3111a1df55fd of this fix some missing nil checks that caused exceptions. https://github.com/TysonAndre/git-diff-syntax-highlight

Other notable changes:
I patched coderay's constants to use a custom color scheme in https://github.com/TysonAndre/git-diff-syntax-highlight/blob/master/vim.rb . This may be of interest if the default color scheme is hard to read, or you want the color scheme to match other tools you use.

overrides.each do |key, value|
    CodeRay::Encoders::Terminal::TOKEN_COLORS[key] = value
end