[Help-source-highlight] Re: Backreferences in source-highlight 2.7

[gnombat]

Lorenzo Bettini wrote:
> gnombat wrote:
> > I'm having trouble with backreferences in source-highlight 2.7.  If I
> > create a language definition file named foo.lang:
> >
> > keyword = `a(.)\1`
> >
> > And then I create a test file example.foo:
> >
> > aaa abb acc a** a@@
> >
> > And then I run source-highlight:
> >
> > source-highlight --lang-def=foo.lang example.foo
> >
> > Then "aaa", "abb" and "acc" are highlighted, but "a**" and "a@@" are
> > not.  Shouldn't the dot match any character?
>
> yes you're right, so this must be a bug...
>
> I'll try to work on it tomorrow
>
> thanks for the feedback!
> cheers
>         Lorenzo

I think I figured out what is going on: is_to_isolate from 
regexpstatebuilder.cpp is treating the regular expression as an 
alphanumerical string because `a(.)\1` starts and ends with a letter or 
number.

/**
 * An expression is isolated basically if it is an alphanumerical
 * string
 * TODO check whether this is actually correct in principle
 * @param s
 * @return
 */
bool is_to_isolate(const string &s) {
  if (s.size()) {
    if ((isalnum(s[0]) || s[0] == '_') && (isalnum(s[s.size()-1]) || 
s[s.size()-1] == '_'))
      return true;
  }

  return false;
}

Then the whole regular expression gets surrounded with word boundary 
characters.  This causes the match to fail with "a**" and "a@@".