Java Regex vs. PHP, Dangling meta character ‘?’

Tags: , ,



I’m tagging this with PHP even though it’s a Java question. The regex is copied from a PHP source so I’m hoping some PHPers can help with the question.

I decided to build a simple spam filter, just for fun, and I copied the spam blocklist from MediaWiki: https://meta.wikimedia.org/wiki/Spam_blacklist

Mostly this seems to work, but a few of the patterns fail with a syntax error. I don’t know if this is a typo or if PHP uses a different syntax than Java. Can anyone help me fixing these regex so that they compile?

Here’s the problems:

java.util.regex.PatternSyntaxException: Dangling meta character '?' near index 17
bfacebo(?:o[ob]|?o)k.comb
                 ^
java.util.regex.PatternSyntaxException: Dangling meta character '?' near index 5
b????.tkb
     ^
java.util.regex.PatternSyntaxException: Dangling meta character '?' near index 0
??.xsl.ptb
^
java.util.regex.PatternSyntaxException: Dangling meta character '?' near index 4
b????.shopb
    ^
java.util.regex.PatternSyntaxException: Dangling meta character '?' near index 4
b???.??b
    ^

Here’s the code that compiles them, in case you’re interested. I don’t think it makes a difference though.

   private static synchronized void init() throws IOException {
      
      if( blackListPatterns.get() != null ) return;
      InputStream blacklistfile = SpamBlackList.class.getResourceAsStream( "blacklist.txt" );
      BufferedReader buf = new BufferedReader( new InputStreamReader( blacklistfile, "UTF-8" ) );
      ArrayList<String> blacklist = new ArrayList<>( 12000 );
      for( String line; (line = buf.readLine()) != null; )
         if( !line.isBlank() && line.trim().charAt(0) != '#' )
            blacklist.add( line );
      ArrayList<Pattern> tempPatterns = new ArrayList<>( blacklist.size() );
      for( String pat : blacklist )
         try {
            tempPatterns.add( Pattern.compile( pat ) );
         } catch ( java.util.regex.PatternSyntaxException ex ) {
            System.err.println( ex );  // should log this, low level like FINER
         }
      blackListPatterns = new WeakReference<>( tempPatterns );
   }
   
   private static volatile WeakReference<List<Pattern>> 
           blackListPatterns = new WeakReference( null );

Answer

Your downloaded copy of https://meta.wikimedia.org/wiki/Spam_blacklist (blacklist.txt) is corrupt. The dangling question marks are non-ASCII characters, e.g. bfacebo(?:o[ob]|?o)k.comb is actually bfacebo(?:o[ob]|ıo)k.comb. Note the dotless “ı”.

Download https://meta.wikimedia.org/wiki/Spam_blacklist?action=raw and take into account that it is UTF-8.

And you may want to pass Unicode flag to the regular expressions. Also take into account that:

What is referred to here as regular expressions are not proper regular expressions, but rather subpatterns that are inserted into a hard-coded regular expression. i.e. the subpattern Foo from above would create a regular expression like /^Foo$/usi.

(see https://www.mediawiki.org/wiki/Extension:TitleBlacklist#Block_list).



Source: stackoverflow