In performing code review for an internal project, a question was raised as to the better method to validate email addresses against RFC822. Two techniques from separate projects were examined and benchmarked, one which uses Email::Valid::Loose and the other, Mail::RFC822::Address, the results of which follow.

The question which I have however is as to whether this benchmark valid? Is the benchmark code accurately measuring the relative validation techniques or is this comparison in some manner flawed? I am most interested in this, as the Benchmark results received differ from what I would have expected, particularly given the size and complexity of the $Addr_spec_re regular expression from Email::Valid::Loose.

The benchmark code:

use Benchmark;
use Email::Valid::Loose;
use Mail::RFC822::Address qw( valid );


my $iter = 100000;

my @results = ();

timethese($iter, { 'Mail::RFC822::Address' => <<EOS,
\$result[0] = valid('per\@p');
\$result[1] = valid('');
\$result[2] = valid('email\@domain.com');
\$result[3] = valid('email\@email\@domain.com');
EOS

'Email::Valid::Loose' => <<EOS } );
(\$result[4]) = 'per\@p' =~ /\^($Email::Valid::Loose::Addr_spec_re)\$/
+;
(\$result[5]) = '' =~ /\^($Email::Valid::Loose::Addr_spec_re)\$/;
(\$result[6]) = 'email\@domain.com' =~ /\^($Email::Valid::Loose::Addr_
+spec_re)\$/;
(\$result[7]) = 'email\@email\@domain.com' =~ /\^($Email::Valid::Loose
+::Addr_spec_re)\$/;
EOS
[download]

And the benchmark results:

Benchmark: timing 100000 iterations of Email::Valid::Loose, Mail::RFC8
+22::Address...
Email::Valid::Loose:  6 wallclock secs ( 7.26 usr +  0.00 sys =  7.26 
+CPU) @ 13774.10/s (n=100000)
Mail::RFC822::Address: 28 wallclock secs (27.03 usr +  0.00 sys = 27.0
+3 CPU) @ 3699.59/s (n=100000)
[download]