Re: ref() and Regexp

[brian_d_foy]

in reply to ref() and Regexp

You shouldn't ever have to think about what string ref() returns. By comparing it to a prototype of the same thing, you insulate yourself from version changes and thinkos (like me forgetting if it's REGEX, REGEXP, Regex, or Regexp):

if( ref $regex eq ref qr// ) { ... }

Or, hide that in a subroutine is_regex() or something similar.

[ikegami]

Or, hide that in a subroutine is_regex() or something similar.

Like this?

BEGIN {
   if (!defined(&re::is_regexp)) {
      package re;
      my $re_class = ref qr//;
      *is_regexp = sub($) {
         local *__ANON__ = 'is_regexp';
         return UNIVERSAL::isa($_[0], $re_class);
      };
   }
}

Using your philosophy to avoid magical constants, it checks if re::is_regexp is defined rather than checking $] for 5.10.

BEGIN and the prototype duplicates the behaviour of 5.10's re::is_regexp.

It uses isa instead of ref so regexes can be made detectable even if there's a need to rebless the regex (by adding Regex to that class's @ISA). It's not an invitation for non-regexes to pretend to be regexes.

[HeatSeekerCannibal]

Whoa!

So I could use this function for a problem like this? XML & Schema VS Config::IniFiles

....crossing fingers...

[Anonymous Monk]

Wouldn't qr/FOO/ be cheaper? (an empty regex is special)

[JadeNB]

an empty regex is special

In a match (see m//), but not for qr//:

$ perl -e 'print qr//;'
(?-xism:)