Match Rinku's behaviour for closing punctuation in links

Rinku 2.0.0 (the version we use) will remove the last character of a link if
it's a closing part of a punctuation pair (different types of parentheses and
quotes), unless both of the below are true:

1. The matching pair has different start and end characters.
2. There are equal numbers of both in the matched string (they don't have to be
   balanced).

lib/banzai/filter/autolink_filter.rb

@@ -25,7 +25,7 @@ module Banzai
       # period or comma for punctuation without those characters being included
       # in the generated link.
       #
-      # Rubular: http://rubular.com/r/cxjPyZc7Sb
+      # Rubular: http://rubular.com/r/JzPhi6DCZp
       LINK_PATTERN = %r{([a-z][a-z0-9\+\.-]+://[^\s>]+)(?<!,|\.)}
 
       # Text matching LINK_PATTERN inside these elements will not be linked
@@ -37,23 +37,17 @@ module Banzai
         and contains(., '://')
       ]).freeze
 
+      PUNCTUATION_PAIRS = {
+        "'" => "'",
+        '"' => '"',
+        ')' => '(',
+        ']' => '[',
+        '}' => '{'
+      }.freeze
+
       def call
         return doc if context[:autolink] == false
 
-        text_parse
-      end
-
-      private
-
-      # Return true if any of the UNSAFE_PROTOCOLS strings are included in the URI scheme
-      def contains_unsafe?(scheme)
-        return false unless scheme
-
-        scheme = scheme.strip.downcase
-        Banzai::Filter::SanitizationFilter::UNSAFE_PROTOCOLS.any? { |protocol| scheme.include?(protocol) }
-      end
-
-      def text_parse
         doc.xpath(TEXT_QUERY).each do |node|
           content = node.to_html
 
@@ -69,6 +63,16 @@ module Banzai
         doc
       end
 
+      private
+
+      # Return true if any of the UNSAFE_PROTOCOLS strings are included in the URI scheme
+      def contains_unsafe?(scheme)
+        return false unless scheme
+
+        scheme = scheme.strip.downcase
+        Banzai::Filter::SanitizationFilter::UNSAFE_PROTOCOLS.any? { |protocol| scheme.include?(protocol) }
+      end
+
       def autolink_match(match)
         # start by stripping out dangerous links
         begin
@@ -84,6 +88,22 @@ module Banzai
         match.gsub!(/((?:&[\w#]+;)+)\z/, '')
         dropped = ($1 || '').html_safe
 
+        # To match the behaviour of Rinku, if the matched link ends with a
+        # closing part of a matched pair of punctuation, we remove that trailing
+        # character unless there are an equal number of closing and opening
+        # characters in the link.
+        if match.end_with?(*PUNCTUATION_PAIRS.keys)
+          close_character = match[-1]
+          close_count = match.count(close_character)
+          open_character = PUNCTUATION_PAIRS[close_character]
+          open_count = match.count(open_character)
+
+          if open_count != close_count || open_character == close_character
+            dropped += close_character
+            match = match[0..-2]
+          end
+        end
+
         options = link_options.merge(href: match)
         content_tag(:a, match.html_safe, options) + dropped
       end

spec/lib/banzai/filter/autolink_filter_spec.rb

@@ -4,6 +4,7 @@ describe Banzai::Filter::AutolinkFilter do
   include FilterSpecHelper
 
   let(:link) { 'http://about.gitlab.com/' }
+  let(:quotes) { ['"', "'"] }
 
   it 'does nothing when :autolink is false' do
     exp = act = link
@@ -15,16 +16,6 @@ describe Banzai::Filter::AutolinkFilter do
     expect(filter(act).to_html).to eq exp
   end
 
-  context 'when the input contains no links' do
-    it 'does not parse_html back the rinku returned value' do
-      act = HTML::Pipeline.parse('<p>This text contains no links to autolink</p>')
-
-      expect_any_instance_of(described_class).not_to receive(:parse_html)
-
-      filter(act).to_html
-    end
-  end
-
   context 'Various schemes' do
     it 'autolinks http' do
       doc = filter("See #{link}")
@@ -141,6 +132,45 @@ describe Banzai::Filter::AutolinkFilter do
       expect(doc.at_css('a').text).to eq link
     end
 
+    it 'includes trailing punctuation when part of a balanced pair' do
+      described_class::PUNCTUATION_PAIRS.each do |close, open|
+        next if open.in?(quotes)
+
+        balanced_link = "#{link}#{open}abc#{close}"
+        balanced_actual = filter("See #{balanced_link}...")
+        unbalanced_link = "#{link}#{close}"
+        unbalanced_actual = filter("See #{unbalanced_link}...")
+
+        expect(balanced_actual.at_css('a').text).to eq(balanced_link)
+        expect(unescape(balanced_actual.to_html)).to eq(Rinku.auto_link("See #{balanced_link}..."))
+        expect(unbalanced_actual.at_css('a').text).to eq(link)
+        expect(unescape(unbalanced_actual.to_html)).to eq(Rinku.auto_link("See #{unbalanced_link}..."))
+      end
+    end
+
+    it 'removes trailing quotes' do
+      quotes.each do |quote|
+        balanced_link = "#{link}#{quote}abc#{quote}"
+        balanced_actual = filter("See #{balanced_link}...")
+        unbalanced_link = "#{link}#{quote}"
+        unbalanced_actual = filter("See #{unbalanced_link}...")
+
+        expect(balanced_actual.at_css('a').text).to eq(balanced_link[0...-1])
+        expect(unescape(balanced_actual.to_html)).to eq(Rinku.auto_link("See #{balanced_link}..."))
+        expect(unbalanced_actual.at_css('a').text).to eq(link)
+        expect(unescape(unbalanced_actual.to_html)).to eq(Rinku.auto_link("See #{unbalanced_link}..."))
+      end
+    end
+
+    it 'removes one closing punctuation mark when the punctuation in the link is unbalanced' do
+      complicated_link = "(#{link}(a'b[c'd]))'"
+      expected_complicated_link = %Q{(<a href="#{link}(a'b[c'd]))">#{link}(a'b[c'd]))</a>'}
+      actual = unescape(filter(complicated_link).to_html)
+
+      expect(actual).to eq(Rinku.auto_link(complicated_link))
+      expect(actual).to eq(expected_complicated_link)
+    end
+
     it 'does not include trailing HTML entities' do
       doc = filter("See &lt;&lt;&lt;#{link}&gt;&gt;&gt;")
 
@@ -162,16 +192,27 @@ describe Banzai::Filter::AutolinkFilter do
   end
 
   context 'when the link is inside a tag' do
-    it 'renders text after the link correctly for http' do
-      doc = filter(ERB::Util.html_escape_once("<http://link><another>"))
+    %w[http rdar].each do |protocol|
+      it "renders text after the link correctly for #{protocol}" do
+        doc = filter(ERB::Util.html_escape_once("<#{protocol}://link><another>"))
 
-      expect(doc.children.last.text).to include('<another>')
+        expect(doc.children.last.text).to include('<another>')
+      end
     end
+  end
 
-    it 'renders text after the link correctly for not other protocol' do
-      doc = filter(ERB::Util.html_escape_once("<rdar://link><another>"))
+  # Rinku does not escape these characters in HTML attributes, but content_tag
+  # does. We don't care about that difference for these specs, though.
+  def unescape(html)
+    %w([ ] { }).each do |cgi_escape|
+      html.sub!(CGI.escape(cgi_escape), cgi_escape)
+    end
 
-      expect(doc.children.last.text).to include('<another>')
+    quotes.each do |html_escape|
+      html.sub!(CGI.escape_html(html_escape), html_escape)
+      html.sub!(CGI.escape(html_escape), CGI.escape_html(html_escape))
     end
+
+    html
   end
 end