Re: Trying to count the captures in a compiled regular expression

[hv]

With respect to (?xism-xism:...), beware of the docs for this (my emphasis):

One or more embedded pattern-match modifiers, to be turned on (or turned off, if preceded by "-") for the remainder of the pattern or the remainder of the enclosing pattern group (if any).

If I understand that correctly, the "only look at openings" approach will fail on something like:

  /(?x:((?-x:)) # (comment)
  )/

or

  /((?-x:)) # (comment)
  /x

The simplistic (?{...}) parsing I referred to is in toke.c:scan_const(); look for the test

            else if (s[2] == '{' /* This should match regcomp.c */
                     || ((s[2] == 'p' || s[2] == '?') && s[3] == '{'))

which simply counts unescaped braces until the opening one is closed - something like:

  our $re_true = qr{(?=)}x;
  our $re_false = qr{(?!)}x;
  our $count;

  /
    # (?{ ... }) or (??{ ... }) or (legacy) (?p{ ... })
    \G \( \? (?: \? \?? | p ) (?= \{ )
    (?{ local $count = 0; })
    (?:
       \{ (?{ local $count = $count + 1 })
    |
       \} (?{ local $count = $count - 1 })
    |
       \\ .
    |
       .
    )+?
    (??{ $count == 0 ? $re_true : $re_false })
  /xgc;

would be fitting, though I suspect there must be a simpler way.

(consider how lucky I am that the regular expression engine is not reentrant...)

Now now, no need for that sort of language.

Hugo

[BooK]

My opinion on (?{ ... }) was that I needed to know how Perl did it to make sure I'd do it right. I'm glad you answered: it proves I was not the only one wanting it to work the hard way. ;-)

Thanks to Roy Johnson below, I guess I'll blend in a does of simplicity in my code. There's still a bug somewhere, though.

Well, at least I can keep the test suite. *sigh*