gfm.rb

require 'digest/md5'

def gfm(text)
  # Extract pre blocks
  extractions = {}
  text.gsub!(%r{<pre>.*?</pre>}m) do |match|
    md5 = Digest::MD5.hexdigest(match)
    extractions[md5] = match
    "{gfm-extraction-#{md5}}"
  end

  # prevent foo_bar_baz from ending up with an italic word in the middle
  text.gsub!(/(^(?! {4}|\t)\w+_\w+_\w[\w_]*)/) do |x|
    x.gsub('_', '\_') if x.split('').sort.to_s[0..1] == '__'
  end

  # in very clear cases, let newlines become <br /> tags
  text.gsub!(/^[\w\<][^\n]*\n+/) do |x|
    x =~ /\n{2}/ ? x : (x.strip!; x << "  \n")
  end

  # Insert pre block extractions
  text.gsub!(/\{gfm-extraction-([0-9a-f]{32})\}/) do
    "\n\n" + extractions[$1]
  end

  text
end

if $0 == __FILE__
  require 'test/unit'
  require 'shoulda'

  class GFMTest < Test::Unit::TestCase
    context "GFM" do
      should "not touch single underscores inside words" do
        assert_equal "foo_bar", gfm("foo_bar")
      end

      should "not touch underscores in code blocks" do
        assert_equal "    foo_bar_baz", gfm("    foo_bar_baz")
      end

      should "not touch underscores in pre blocks" do
        assert_equal "\n\n<pre>\nfoo_bar_baz\n</pre>", gfm("<pre>\nfoo_bar_baz\n</pre>")
      end

      should "not treat pre blocks with pre-text differently" do
        a = "\n\n<pre>\nthis is `a\\_test` and this\\_too\n</pre>"
        b = "hmm<pre>\nthis is `a\\_test` and this\\_too\n</pre>"
        assert_equal gfm(a)[2..-1], gfm(b)[3..-1]
      end

      should "escape two or more underscores inside words" do
        assert_equal "foo\\_bar\\_baz", gfm("foo_bar_baz")
      end

      should "turn newlines into br tags in simple cases" do
        assert_equal "foo  \nbar", gfm("foo\nbar")
      end

      should "convert newlines in all groups" do
        assert_equal "apple  \npear  \norange\n\nruby  \npython  \nerlang",
                     gfm("apple\npear\norange\n\nruby\npython\nerlang")
      end

      should "convert newlines in even long groups" do
        assert_equal "apple  \npear  \norange  \nbanana\n\nruby  \npython  \nerlang",
                     gfm("apple\npear\norange\nbanana\n\nruby\npython\nerlang")
      end

      should "not convert newlines in lists" do
        assert_equal "# foo\n# bar", gfm("# foo\n# bar")
        assert_equal "* foo\n* bar", gfm("* foo\n* bar")
      end
    end
  end
end

Thanks for making this code available - I was pulling my hair out trying to figure out how to accomplish something like this on my site.

Thanks for sharing! I found one Ruby 1.9 compatibility issue that's worth mentioning. Array#to_s is equivalent to Array#inspect now.
http://eigenclass.org/hiki/Changes+in+Ruby+1.9#l83

The fix is simple -- just change the to_s on line 14 to join and this fine code snippet will run for both Ruby 1.8 and 1.9.

The code [\#1](http://github.com) without the backslash breaks GH's markdown parser. This generates [#1](http://github.com).

Edit: it looks ok now it's posted here. Maybe it's just in the issue tracker / JS previews; try it in the preview box below.

If you have a pre-formatted code block that contains the string '#1', it gets expanded to a link.

I agree with @viking. Here is an example of the problem:
https://github.com/tenderlove/nokogiri/issues/#issue/405/comment/725920

Why is there a '?' in here.....< pre >.*?< / pre > Does the ruby regex have some weird bug I don't know about.
Normally .+? == .★

The ? prevents .* from matching </pre>. And .+? is not the same as .*. See reluctant matching.

Thanks.

is there a port of this to PHP?

Coffeescript port is here: https://gist.github.com/2349475

PHP port here: https://gist.github.com/3194002

There's an error in the line break handling code. Take for example this input:

**foo**
bar
baz

On GitHub, this is rendered as expected, because GitHub actually uses redcarpet for rendering GFM on the site:

foo
bar
baz

But using the code here in gfm.rb, it renders as:

foo bar
baz

This is happening because the regex used in gfm.rb is incorrect; it does not add a line break when the line begins with a * (emphasized / strong text) or a > (blockquote). It can rectified as follows:

--- gfm.rb
+++ gfm.2.rb
@@ -15,6 +15,6 @@
   end

   # in very clear cases, let newlines become <br /> tags
-  text.gsub!(/^[\w\<][^\n]*\n+/) do |x|
+  text.gsub!(/^[\w\<\>\*][^\n]*\n+/) do |x|
     x =~ /\n{2}/ ? x : (x.strip!; x << "  \n")
   end

Very good Work