Count capturing parentheses in a compiled regexp

[BooK]

This subroutines count the number of capturing parentheses in a compiled regular expression. Since the regexp is compiled, we know it's correct, and thus we only have to count the opening parentheses.

Update: added /s in the while condition.

Update: This other discussion led to a much better solution, so I commented out the original code.

#sub captures {
#    local $_ = shift;
#    croak "$_ is not a compiled regexp" unless ref eq 'Regexp';
#    my $n = 0;
#    while( /\G(?=.)/gcs ) {
#      /\G[^\\(]*/gc;     # ignore uninteresting stuff
#      /\G(?:\\.)*/gc;    # ignore backslashed stuff
#      /\G\(\?/gc;        # ignore special regexps
#      /\G\(/gc && $n++;  # a capturing (, count it!
#    }
#    $n;
#}

sub captures { ( @_ = '' =~ /(@{[shift]})??/ ) - 1; }

[hv]

Nice snippet, but a couple of problems: the outer lookahead needs to be //s, else eg:

  qr{(
    x
  )}x;

will fail.

Also, this will find parens in embedded code and comments and treat as captures. If that doesn't seem worth worrying about it'd be enough to add a caveat I guess, else I think you can mimic perl's simplistic parsing reasonable easily for the code (just count to the balancing close-brace). Comments may actually be the trickiest, since you'll need to know when //x is in force:

  qr{
    (?x: # (comment)
    )
    (?-x: # (capture) )
  }

Oops, another one: parens in [ ... ] should be ignored too; I'm not sure how easy those would be to parse, since not every ] closes the selection.

Hugo

[BooK]

Thanks a lot for finding these shortcomings in my code. :-) I'll submit updated versions as I correct them.

[japhy]

Once I get Regexp::Parser working (it's the update to YAPE::Regex), you'll be able to do this via:

use Regexp::Parser;  # pushes itself to @Regexp::ISA
print qr/.../->nparens;

[japhy]

In the spirit of "match the regex to get the number of parens in it", here's another way:

sub nparens { "" =~ /|$_[0]/ and $#+ }