use strict;
use warnings;
$| = 1;

my $SAMPLE = 'AB ABDA DCACCB AAA BSAA CAAB ACS ABA DBA BA DASSABACA A'
+;

local our $first;

$SAMPLE =~ m/
    (?: BA[ABC]{2} (?{ $first = 1; }) |
        CA[CD]{2}  (?{ $first = 2; }) |
        DA[SC]{2}  (?{ $first = 3; })
    ) /x;

print $first;
[download]

And I did a test. This second version runs almost twice as fast as the first one. Very interesting... :-)

Edit: Sorry, I don't want to change the subject but can't help it. lol I did another test on isNumber() which looks like this:

sub isNumber
{
  defined $_[0] or return 0;
  my $N = 1;
  local $SIG{__WARN__} = sub { $N = 0; };
  return ($_[0] & 0) | $N;
}
[download]

and it turns out that a simple regex test on a string to see if it's a number runs orders of magnitude faster than using the above method of trying to catch a warning...

> Sorry, I don't want to change the subject but can't help it.

Best a new thread with full benchmark. Too much guessing needed...

And consider

use Scalar::Util qw(looks_like_number);