http://www.perlmonks.org?node_id=1040107


in reply to Re: Find out which subpattern matched in regex
in thread Find out which subpattern matched in regex

Thanks for you comment.

I did a quick benchmark and it turns out that nested loop solution is the fastest in most cases. In all other cases - I believe it's when regular expressions converge into a nice tree-like structure - both named captures and unnamed captures are about the same.

Here's my code sample:

#!/usr/bin/perl -w use 5.010; use strict; use warnings; use Test::More tests => 3; use Benchmark qw(cmpthese); my @reglist = ( qr/food?/, qr/b[a4]rd?/, qr/baz(?:o+ka)?/, 100..999); my @lines = (qw(foobarbaz b4rd perl bazooooka football)); my @expect = ("r0", "r1", "", "r2", "r0"); is_deeply(which_reg_loop(\@reglist, \@lines), \@expect, "which_reg_loo +p") and is_deeply(which_reg_capt(\@reglist, \@lines), \@expect, "which_reg_cap +t") and is_deeply(which_reg_named(\@reglist, \@lines), \@expect, "which_reg_na +med") or die "Results differ, no bench"; @lines = @lines x 1000; cmpthese ( -1, { loop => sub { which_reg_loop(\@reglist, \@lines); }, capt => sub { which_reg_capt(\@reglist, \@lines); }, named => sub { which_reg_named(\@reglist, \@lines); }, }); sub which_reg_loop { my ($reglist, $lines) = @_; my @ret; LINE: foreach my $str (@$lines) { for (my $i = 0; $i < @$reglist; $i++) { $str =~ $reglist->[$i] or next; push @ret, "r$i"; next LINE; }; push @ret, ''; }; return \@ret; }; sub which_reg_capt { my ($reglist, $lines) = @_; my $giant = join "|", map { "($_)" } @$reglist; $giant = qr($giant); my @ret; LINE: foreach (@$lines) { my @hits = $_ =~ $giant; for (my $i = 0; $i < @hits; $i++) { $hits[$i] or next; push @ret, "r$i"; next LINE; }; push @ret, ''; }; return \@ret; }; sub which_reg_named { my ($reglist, $lines) = @_; my $giant = join "|", map { "(?<r$_>$reglist->[$_])" } 0..$#$regli +st; $giant = qr($giant); my @ret = map { $_ =~ $giant ? (keys %+) : '' } @$lines; return \@ret; };