Re: Did the inefficiency of /i get fixed?

[Belgarion]

To amplify what sgifford said: it would appear that the longer the string being matched, the bigger the difference between the two methods. For example, using sgifford's script, but with a $s of one character, the numbers on my machine are:

             Rate brackets  slash i
brackets 719424/s       --      -7%
slash i  775194/s       8%       --

While, when $s is 16KB long, the difference is more pronounced, with the following:

            Rate brackets  slash i
brackets  3508/s       --     -66%
slash i  10434/s     197%       --

Now the interesting question: sgifford's regex is very simple. What happens when a more complicated test is used. When using the following test code:

#!/usr/bin/perl -w
use strict;
use Benchmark 'cmpthese';

our $s = "HELLOWORLD" x 1024;
cmpthese(1_000_000, {
                     'slash_i' => sub { $s =~ m/hello/i; $s =~ m/jello
+/i },
                     'brackets' => sub { $s =~ m/[Hh][Ee][Ll][Ll][Oo]/
+;
                                         $s =~ m/[Jj][Ee][Ll][Ll][Oo]/
+; },
                    });

produce the following results:

            Rate brackets  slash_i
brackets  5633/s       --     -66%
slash_i  16683/s     196%       --

It would appear that the length of the string amplifies the difference more than the complexity of the regular expression. (At least when comparing /[Tt]/ with /t/i.)