po(change): avoid losing information before guessing the same...

[git.ikiwiki.info.git] / IkiWiki / Plugin / smiley.pm

diff --git a/IkiWiki/Plugin/smiley.pm b/IkiWiki/Plugin/smiley.pm
index 932c2c4fe928f88e8a140482144290b36c2016bb..2633b1ea14cf6ab247ea858e3049c977db5f54bc 100644 (file)
--- a/IkiWiki/Plugin/smiley.pm
+++ b/IkiWiki/Plugin/smiley.pm
@@ -10,13 +10,33 @@ my $smiley_regexp;
 
 sub import { #{{{
        add_underlay("smiley");
-       hook(type => "filter", id => "smiley", call => \&filter);
+       hook(type => "getsetup", id => "smiley", call => \&getsetup);
+       hook(type => "sanitize", id => "smiley", call => \&sanitize);
 } # }}}
 
+sub getsetup () { #{{{
+       return
+               plugin => {
+                       safe => 1,
+                       # force a rebuild because turning it off
+                       # removes the smileys, which would break links
+                       rebuild => 1,
+               },
+} #}}}
+
 sub build_regexp () { #{{{
        my $list=readfile(srcfile("smileys.mdwn"));
-       while ($list =~ m/^\s*\*\s+\\([^\s]+)\s+\[\[([^]]+)\]\]/mg) {
-               $smileys{$1}=$2;
+       while ($list =~ m/^\s*\*\s+\\\\([^\s]+)\s+\[\[([^]]+)\]\]/mg) {
+               my $smiley=$1;
+               my $file=$2;
+
+               $smileys{$smiley}=$file;
+
+               # Add a version with < and > escaped, since they probably
+               # will be (by markdown) by the time the sanitize hook runs.
+               $smiley=~s/</&lt;/g;
+               $smiley=~s/>/&gt;/g;
+               $smileys{$smiley}=$file;
        }
        
        if (! %smileys) {
@@ -32,15 +52,49 @@ sub build_regexp () { #{{{
        #debug($smiley_regexp);
 } #}}}
 
-sub filter (@) { #{{{
+sub sanitize (@) { #{{{
        my %params=@_;
-       
+
        build_regexp() unless defined $smiley_regexp;
-       $params{content} =~ s{(?:^|(?<=\s))(\\?)$smiley_regexp(?:(?=\s)|$)}{
-               $1 ? $2 : htmllink($params{page}, $params{destpage}, $smileys{$2}, linktext => $2)
-       }egs if length $smiley_regexp;
+       
+       $_=$params{content};
+       return $_ unless length $smiley_regexp;
+                       
+MATCH: while (m{(?:^|(?<=\s|>))(\\?)$smiley_regexp(?:(?=\s|<)|$)}g) {
+               my $escape=$1;
+               my $smiley=$2;
+               my $epos=$-[1];
+               my $spos=$-[2];
+               
+               # Smilies are not allowed inside <pre> or <code>.
+               # For each tag in turn, match forward to find the next <tag>
+               # or </tag> after the smiley.
+               my $pos=pos;
+               foreach my $tag ("pre", "code") {
+                       if (m/<(\/)?\s*$tag\s*>/isg && defined $1) {
+                               # </tag> found first, so the smiley is
+                               # inside the tag, so do not expand it.
+                               next MATCH;
+                       }
+                       # Reset pos back to where it was before this test.
+                       pos=$pos;
+               }
+       
+               if ($escape) {
+                       # Remove escape.
+                       substr($_, $epos, 1)="";
+                       pos=$epos+1;
+               }
+               else {
+                       # Replace the smiley with its expanded value.
+                       substr($_, $spos, length($smiley))=
+                               htmllink($params{page}, $params{destpage},
+                                        $smileys{$smiley}, linktext => $smiley);
+                       pos=$epos+1;
+               }
+       }
 
-       return $params{content};
+       return $_;
 } # }}}
 
 1