Benchmarking perl

I had a little code that worked perfectly for what it had to do (process every line in a file, and get a list of unique lines (after being processed)).

The 'process' I had to run on every line was a simple substitution, so I thought of Algorithm::Loops. I used it and worked really well, but then I thought about performance, and tried some other approaches:


Well I could clearly see that while version is the winner, and it really makes a big difference. I suppose this is because Filter takes <$fh> as an array (in list context), and that is slower than the iterator version (just guessing).

Here's the code:

	#!/usr/bin/perl
	use Benchmark qw(:all);
	use Algorithm::Loops qw(Filter);
	system('rm /tmp/footable.txt');
	system('seq 1 2000 >>/tmp/footable.txt') for (1..4);
	system('sed -i s/$/_i02_c00\ foo/ /tmp/footable.txt');
	cmpthese (50, {
	'filterChomp' =>
	'open my $fh, "/tmp/footable.txt";
	my %hash = map {
	$_ => 1
	} Filter {s/_i\d+_c\d+ .*$//} Filter{chomp} <$fh>;
	close $fh; ' ,
	'regexBarraN' =>
	'open my $fh, "/tmp/footable.txt";
	my %hash = map {
	$_ => 1
	} Filter {s/_i\d+_c\d+ .*\n//} <$fh>;
	close $fh;' ,
	'match' =>
	' open my $fh, "/tmp/footable.txt";
	my %hash = map {
	$_ => 1
	} Filter {m/^([^ ]*)_i\d+_c\d+/} <$fh>;
	close $fh; ' ,
	'while' =>
	' open my $fh, "/tmp/footable.txt";
	my %hash;
	while (<$fh>) {
	m/^([^ ]*)_i\d+_c\d+ .*/;
	$hash{$1}=1;
	}
	close $fh;' ,
	});

view raw gistfile1.pl hosted with ❤ by GitHub

And the results:

            Rate filterChomp       match regexBarraN       while
filterChomp 9.49/s          --        -19%        -24%        -64%
match       11.8/s         24%          --         -6%        -55%
regexBarraN 12.5/s         32%          6%          --        -52%
while       26.0/s        174%        121%        108%          --

Well, appart from keeping the while version for my app, now I have a template for bencharking perl codes. Following the same technique than in one of my other perl posts, I've written a template file to be easy to paste in a .pl file from vim.