git: reluctantly give diff-highlight a try

I saw people talking about this a couple of weeks ago[1], but ignored it
as I like to keep things simple, and I don't like the idea of slowing
down the pager.

Nevertheless, in my current sleep-deprived state, I feel like giving it
a shot. I'm bundling a copy of the `diff-highlight` script from Git
2.6.1 because I want this to work everywhere that I use my dotfiles,
without having to worry about `$PATH`, issues. Got it via:

```
cp "$(brew --prefix)/share/git-core/contrib/diff-highlight/diff-highlight" \
  roles/dotfiles/files/.shells/bin
```

Note that I am using  much more subtle color choices than seem to be the
norm. (I don't really want this to shout at me, "HEY! LOOK! I AM A WORD
THAT CHANGED!"; rather, I want diffs to pretty much look like diffs, and
have some subtle cues to spot word differences when I actually need to.

[1]: paulirish/dotfiles@6743b907

roles/dotfiles/files/.gitconfig

@@ -54,6 +54,10 @@
 [color]
 	ui = auto
 
+[color "diff-highlight"]
+	oldHighlight = red 236 dim
+	newHighlight = green 236 dim
+
 [color "interactive"]
 	prompt = blue reverse
 
@@ -62,7 +66,7 @@
 
 [core]
 	excludesfile = ~/.gitignore
-	pager = less -iFMRSX
+	pager = sh -c "~/.shells/bin/diff-highlight | less -iFMRSX" -
 	attributesfile = ~/.gitattributes
 
 # ignored by Git older than 1.8.2

roles/dotfiles/files/.shells/bin/diff-highlight

@@ -0,0 +1,218 @@
+#!/usr/bin/perl
+
+use 5.008;
+use warnings FATAL => 'all';
+use strict;
+
+# Highlight by reversing foreground and background. You could do
+# other things like bold or underline if you prefer.
+my @OLD_HIGHLIGHT = (
+	color_config('color.diff-highlight.oldnormal'),
+	color_config('color.diff-highlight.oldhighlight', "\x1b[7m"),
+	color_config('color.diff-highlight.oldreset', "\x1b[27m")
+);
+my @NEW_HIGHLIGHT = (
+	color_config('color.diff-highlight.newnormal', $OLD_HIGHLIGHT[0]),
+	color_config('color.diff-highlight.newhighlight', $OLD_HIGHLIGHT[1]),
+	color_config('color.diff-highlight.newreset', $OLD_HIGHLIGHT[2])
+);
+
+my $RESET = "\x1b[m";
+my $COLOR = qr/\x1b\[[0-9;]*m/;
+my $BORING = qr/$COLOR|\s/;
+
+my @removed;
+my @added;
+my $in_hunk;
+
+# Some scripts may not realize that SIGPIPE is being ignored when launching the
+# pager--for instance scripts written in Python.
+$SIG{PIPE} = 'DEFAULT';
+
+while (<>) {
+	if (!$in_hunk) {
+		print;
+		$in_hunk = /^$COLOR*\@/;
+	}
+	elsif (/^$COLOR*-/) {
+		push @removed, $_;
+	}
+	elsif (/^$COLOR*\+/) {
+		push @added, $_;
+	}
+	else {
+		show_hunk(\@removed, \@added);
+		@removed = ();
+		@added = ();
+
+		print;
+		$in_hunk = /^$COLOR*[\@ ]/;
+	}
+
+	# Most of the time there is enough output to keep things streaming,
+	# but for something like "git log -Sfoo", you can get one early
+	# commit and then many seconds of nothing. We want to show
+	# that one commit as soon as possible.
+	#
+	# Since we can receive arbitrary input, there's no optimal
+	# place to flush. Flushing on a blank line is a heuristic that
+	# happens to match git-log output.
+	if (!length) {
+		local $| = 1;
+	}
+}
+
+# Flush any queued hunk (this can happen when there is no trailing context in
+# the final diff of the input).
+show_hunk(\@removed, \@added);
+
+exit 0;
+
+# Ideally we would feed the default as a human-readable color to
+# git-config as the fallback value. But diff-highlight does
+# not otherwise depend on git at all, and there are reports
+# of it being used in other settings. Let's handle our own
+# fallback, which means we will work even if git can't be run.
+sub color_config {
+	my ($key, $default) = @_;
+	my $s = `git config --get-color $key 2>/dev/null`;
+	return length($s) ? $s : $default;
+}
+
+sub show_hunk {
+	my ($a, $b) = @_;
+
+	# If one side is empty, then there is nothing to compare or highlight.
+	if (!@$a || !@$b) {
+		print @$a, @$b;
+		return;
+	}
+
+	# If we have mismatched numbers of lines on each side, we could try to
+	# be clever and match up similar lines. But for now we are simple and
+	# stupid, and only handle multi-line hunks that remove and add the same
+	# number of lines.
+	if (@$a != @$b) {
+		print @$a, @$b;
+		return;
+	}
+
+	my @queue;
+	for (my $i = 0; $i < @$a; $i++) {
+		my ($rm, $add) = highlight_pair($a->[$i], $b->[$i]);
+		print $rm;
+		push @queue, $add;
+	}
+	print @queue;
+}
+
+sub highlight_pair {
+	my @a = split_line(shift);
+	my @b = split_line(shift);
+
+	# Find common prefix, taking care to skip any ansi
+	# color codes.
+	my $seen_plusminus;
+	my ($pa, $pb) = (0, 0);
+	while ($pa < @a && $pb < @b) {
+		if ($a[$pa] =~ /$COLOR/) {
+			$pa++;
+		}
+		elsif ($b[$pb] =~ /$COLOR/) {
+			$pb++;
+		}
+		elsif ($a[$pa] eq $b[$pb]) {
+			$pa++;
+			$pb++;
+		}
+		elsif (!$seen_plusminus && $a[$pa] eq '-' && $b[$pb] eq '+') {
+			$seen_plusminus = 1;
+			$pa++;
+			$pb++;
+		}
+		else {
+			last;
+		}
+	}
+
+	# Find common suffix, ignoring colors.
+	my ($sa, $sb) = ($#a, $#b);
+	while ($sa >= $pa && $sb >= $pb) {
+		if ($a[$sa] =~ /$COLOR/) {
+			$sa--;
+		}
+		elsif ($b[$sb] =~ /$COLOR/) {
+			$sb--;
+		}
+		elsif ($a[$sa] eq $b[$sb]) {
+			$sa--;
+			$sb--;
+		}
+		else {
+			last;
+		}
+	}
+
+	if (is_pair_interesting(\@a, $pa, $sa, \@b, $pb, $sb)) {
+		return highlight_line(\@a, $pa, $sa, \@OLD_HIGHLIGHT),
+		       highlight_line(\@b, $pb, $sb, \@NEW_HIGHLIGHT);
+	}
+	else {
+		return join('', @a),
+		       join('', @b);
+	}
+}
+
+sub split_line {
+	local $_ = shift;
+	return utf8::decode($_) ?
+		map { utf8::encode($_); $_ }
+			map { /$COLOR/ ? $_ : (split //) }
+			split /($COLOR+)/ :
+		map { /$COLOR/ ? $_ : (split //) }
+		split /($COLOR+)/;
+}
+
+sub highlight_line {
+	my ($line, $prefix, $suffix, $theme) = @_;
+
+	my $start = join('', @{$line}[0..($prefix-1)]);
+	my $mid = join('', @{$line}[$prefix..$suffix]);
+	my $end = join('', @{$line}[($suffix+1)..$#$line]);
+
+	# If we have a "normal" color specified, then take over the whole line.
+	# Otherwise, we try to just manipulate the highlighted bits.
+	if (defined $theme->[0]) {
+		s/$COLOR//g for ($start, $mid, $end);
+		chomp $end;
+		return join('',
+			$theme->[0], $start, $RESET,
+			$theme->[1], $mid, $RESET,
+			$theme->[0], $end, $RESET,
+			"\n"
+		);
+	} else {
+		return join('',
+			$start,
+			$theme->[1], $mid, $theme->[2],
+			$end
+		);
+	}
+}
+
+# Pairs are interesting to highlight only if we are going to end up
+# highlighting a subset (i.e., not the whole line). Otherwise, the highlighting
+# is just useless noise. We can detect this by finding either a matching prefix
+# or suffix (disregarding boring bits like whitespace and colorization).
+sub is_pair_interesting {
+	my ($a, $pa, $sa, $b, $pb, $sb) = @_;
+	my $prefix_a = join('', @$a[0..($pa-1)]);
+	my $prefix_b = join('', @$b[0..($pb-1)]);
+	my $suffix_a = join('', @$a[($sa+1)..$#$a]);
+	my $suffix_b = join('', @$b[($sb+1)..$#$b]);
+
+	return $prefix_a !~ /^$COLOR*-$BORING*$/ ||
+	       $prefix_b !~ /^$COLOR*\+$BORING*$/ ||
+	       $suffix_a !~ /^$BORING*$/ ||
+	       $suffix_b !~ /^$BORING*$/;
+}