in reply to Re (tilly) 3: parsing question
in thread Re: parsing question


> The RE engine does have the power to solve this problem.

Hrm.. please demonstrate. I know you can build a lexer by looping over m/\G$regexp/g, but that doesn't give you the state storage neccessary to balance parens. If you could whip up something to convert the second-level parens in: 

    (((())()))(()(()()))

[download]

to square brackets: 

    ([(())()])([][()()])

[download]

using only regexps.. no variables, no recursion (and for arbitrary strings of nested parens, of course).. I'd love to add the technique to my bag of tricks.

Replies are listed 'Best First'.
Re (tilly) 5: parsing question
by tilly (Archbishop) on Jan 08, 2002 at 06:23 UTC
    You just changed the problem massively. I didn't promise that the RE engine magically has become able to handle arbitrary parsing problems. I never said that no variables were involved. I merely said that the RE engine can handle balanced delimiters.

    The two experimental features which are needed are (??{}) for delayed evaluation and (?>) for telling the engine not to backtrack. A sample script demonstrating the technique: 

    #! /usr/bin/perl
my $braces;
$braces = qr/(\((?:(?>[^\(\)]+)|(??{$braces}))*\))/;

while (<>) {
  if ($_ =~ $braces) {
    print "Matched '$1'\n\n";
  }
  else {
    print "No match\n\n";
  }
}

    [download]
    Run it and start typing in lines. The ones with balanced parentheses will match.
[reply]
[d/l]
(MeowChow) Re4: parsing question
by MeowChow (Vicar) on Jan 08, 2002 at 06:02 UTC
    One example of this sort of thing is here. Look up perlre, and search the text for "postponed". 
       MeowChow                                   
               s aamecha.s a..a\u$&owag.print
[reply]

      Interesting.. thanks.

[reply]

In Section Seekers of Perl Wisdom