Initial support for GitHub extended autolink syntax.

.aspelldict

@@ -1,4 +1,4 @@
-personal_ws-1.1 en 164 
+personal_ws-1.1 en 166 
 CDATA
 CommonMark
 CounterClockwiseContourIntegral
@@ -160,6 +160,8 @@ utf
 wantarray
 wget
 whitespace
+www
+wwww
 xfffd
 xmp
 zA

lib/Markdown/Perl/Inlines.pm

@@ -45,6 +45,10 @@ sub render {
 
   process_styles($that, $tree);
 
+  if ($that->get_use_extended_autolinks) {
+    $tree->map(sub { create_autolinks($that, $_) });
+  }
+
   # At this point we have added the emphasis, strong emphasis, etc. in the tree.
 
   $tree->apply(
@@ -713,4 +717,54 @@ sub delim_characters {
   return join('', uniq @c);
 }
 
+sub create_autolinks {
+  my ($that, $n) = @_;
+  if ($n->{type} ne 'text') {
+    return $n;
+  }
+
+  my @nodes;
+
+  # TODO: technically we should forbid the presence of _ in the last two parts
+  # of the domain, according to the gfm spec.
+  ## no critic (ProhibitComplexRegexes)
+  while (
+    $n->{content} =~ m/
+    (?<prefix> ^ | [ \t\n*_~\(] )              # The link must start after a whitespace or some specific delimiters.
+    (?<url>
+      (?: (?<scheme>https?:\/\/) | www\. )      # It must start by a scheme or the string wwww. 
+      [-_a-zA-Z0-9]+ (?: \. [-_a-zA-Z0-9]+ )*   # Then there must be something that looks like a domain
+      (?: \/ [^ \t\n<]*? )?                     # Some characters are forbidden in the link.
+    )
+    [?!.,:*_~]* (?: [ \t\n<] | $)               # We remove some punctuation from the end of the link.
+   /x
+    ## use critic
+  ) {
+    my $url = $+{url};
+    my $match_start = $LAST_MATCH_START[0] + length($LAST_PAREN_MATCH{prefix});
+    my $match_end = $match_start + length($url);
+    my $has_scheme = exists $LAST_PAREN_MATCH{scheme};
+    if ($url =~ m/\)+$/) {
+      my $nb_final_closing_parens = $LAST_MATCH_END[0] - $LAST_MATCH_START[0];
+      my $open = 0;
+      () = $url =~ m/ \( (?{$open++}) | \) (?{$open--}) /gx;
+      my $remove = min($nb_final_closing_parens, -$open);
+      if ($remove > 0) {
+        $match_end -= $remove;
+        substr $url, -$remove, $remove, '';
+      }
+    }
+    # TODO: handle an HTML entity at the end of the link.
+    if ($match_start > 0) {
+      push @nodes, new_text(substr $n->{content}, 0, $match_start);
+    }
+    my $scheme = $has_scheme ? '' : $that->get_default_extended_autolinks_scheme.'://';
+    push @nodes,
+        new_link($url, type => 'autolink', target => $scheme.$url, debug => 'extended autolink');
+    $n = new_text(substr $n->{content}, $match_end);
+  }
+  push @nodes, $n if length($n->{content}) > 0;
+  return @nodes;
+}
+
 1;

lib/Markdown/Perl/Options.pm

@@ -156,6 +156,14 @@ sub _regex {
   };
 }
 
+sub _word_list {
+  return sub {
+    my @a = ref $_[0] eq 'ARRAY' ? @{$_[0]} : split(/,/, $_[0]);
+    # TODO: validate the values of a.
+    return \@a;
+  };
+}
+
 =pod
 
 =head2 B<use_fenced_code_blocks> I<(boolean, default: true)>
@@ -477,19 +485,39 @@ they will be deactivated in the output.
 
 =cut
 
-# TODO: this is just a "word list" for now, see if this can be shared with other
-# options.
-sub _tag_list {
-  return sub {
-    my @a = ref $_[0] eq 'ARRAY' ? @{$_[0]} : split(/,/, $_[0]);
-    # TODO: validate the values of a.
-    return \@a;
-  };
-}
-
 _make_option(
   disallowed_html_tags => [],
-  _tag_list,
+  _word_list,
   github => [qw(title textarea style xmp iframe noembed noframes script plaintext)]);
 
+=pod
+
+=head2 B<use_extended_autolinks> I<(boolean, default: true)>
+
+Allow some links to be recognised when they appear in plain text. These links
+must start by C<http://>, C<https://>, or C<www.>.
+
+=cut
+
+_make_option(
+  use_extended_autolinks => 1,
+  _boolean, (
+    markdown => 0,
+    cmark => 0
+  ));
+
+=pod
+
+=head2 B<default_extended_autolinks_scheme> I<(enum, default: https)>
+
+Specify which scheme is added to the beginning of extended autolinks when none
+was present initially.
+
+=cut
+
+_make_option(
+  default_extended_autolinks_scheme => 'https',
+  _enum(qw(http https)),
+  github => 'http');
+
 1;

t/901-github-test-suite.t

@@ -10,10 +10,19 @@ use Test2::V0;
 
 # TODO: remove these todos.
 my %opt = (todo => [198 .. 202, 204, 205, 279, 280, 398, 426, 434 .. 436,
-                       473 .. 475, 477, 621 .. 631, 652],
-           # The spec says that some HTML tags are forbidden in the output, but
-           # they still have examples with these tags.
-           bugs => [140 .. 142, 145, 147],
+                       473 .. 475, 477, 626, 628 .. 631, 652],
+           # These are bugs in the GitHub spec, not in our implementation. All
+           # of these have been tested to be buggy in the real cmark-gfm
+           # implementation.
+           bugs => [
+             # The spec says that some HTML tags are forbidden in the output, but
+             # they still have examples with these tags.
+            140 .. 142, 145, 147,
+            # Some things that are not cmark autolinks are matched by the
+            # extended autolinks syntax (but the cmark part of the spec is not
+            # updated).
+            616, 619,
+            ],
            json_file => "${FindBin::Bin}/data/github.tests.json",
            test_url => 'https://github.github.com/gfm/#example-%d',
            spec_tool => "${FindBin::Bin}/../third_party/commonmark-spec/test/spec_tests.py",