I must admit I realized it was less readable as I was posting. I wonder if there's a significant speed difference, since this regexp is likely to be used repeatedly. Either way, the regex could be built dynamically when the module is loaded. That would allow for much more readable code than any of the alternatives we've seen in this thread.

alternations are slower, but tr is faster for simple expressions:

               Rate alternate    single        tr
alternate  631309/s        --      -80%      -86%
single    3198780/s      407%        --      -27%
tr        4411077/s      599%       38%        --

use Benchmark qw(cmpthese);

cmpthese(-1, {
    alternate => \&alternate,
    single => \&single,
    tr => \&tr_op
});

my $s;
BEGIN{ $s = 'this is a string'; }

sub alternate {
    return $s =~ /[a-b]|[c-d]|[e-f]/ ;
}

sub single {
    return $s =~ /[a-f]/;
}

sub tr_op {
    return $s =~ tr/a-f/a-f/;
}
[download]