Merge branch '44587-autolinking-includes-trailing-exclamation-marks' into 'master'

Resolve "Autolinking includes trailing exclamation marks"

Closes #44587

See merge request gitlab-org/gitlab-ce!17965

changelogs/unreleased/44587-autolinking-includes-trailing-exclamation-marks.yml

+---
+title: Don't capture trailing punctuation when autolinking
+merge_request: 17965
+author:
+type: fixed

lib/banzai/filter/autolink_filter.rb

@@ -21,12 +21,13 @@ module Banzai
       #
       # See http://en.wikipedia.org/wiki/URI_scheme
       #
-      # The negative lookbehind ensures that users can paste a URL followed by a
-      # period or comma for punctuation without those characters being included
-      # in the generated link.
+      # The negative lookbehind ensures that users can paste a URL followed by
+      # punctuation without those characters being included in the generated
+      # link. It matches the behaviour of Rinku 2.0.1:
+      # https://github.com/vmg/rinku/blob/v2.0.1/ext/rinku/autolink.c#L65
       #
-      # Rubular: http://rubular.com/r/JzPhi6DCZp
-      LINK_PATTERN = %r{([a-z][a-z0-9\+\.-]+://[^\s>]+)(?<!,|\.)}
+      # Rubular: http://rubular.com/r/nrL3r9yUiq
+      LINK_PATTERN = %r{([a-z][a-z0-9\+\.-]+://[^\s>]+)(?<!\?|!|\.|,|:)}
 
       # Text matching LINK_PATTERN inside these elements will not be linked
       IGNORE_PARENTS = %w(a code kbd pre script style).to_set

spec/lib/banzai/filter/autolink_filter_spec.rb

@@ -122,14 +122,10 @@ describe Banzai::Filter::AutolinkFilter do
     end
 
     it 'does not include trailing punctuation' do
-      doc = filter("See #{link}.")
-      expect(doc.at_css('a').text).to eq link
-
-      doc = filter("See #{link}, ok?")
-      expect(doc.at_css('a').text).to eq link
-
-      doc = filter("See #{link}...")
-      expect(doc.at_css('a').text).to eq link
+      ['.', ', ok?', '...', '?', '!', ': is that ok?'].each do |trailing_punctuation|
+        doc = filter("See #{link}#{trailing_punctuation}")
+        expect(doc.at_css('a').text).to eq link
+      end
     end
 
     it 'includes trailing punctuation when part of a balanced pair' do