Head-to-Head: perl’s eval, Exception.pm, Exception/Class.pm and Error.pm

I wanted to use a more OO approach to error handling in perl instead of just plain old eval , so I did some looking around. I wanted it to be easy, so I looked at these modules (UPDATE: Exception::Class included here thanks to William Sutton):

• Exception
• Error
• Exception::Class

I also found this article on perl exception handling: Object Oriented Exception Handling in Perl. That sounded great and all, but what would I be giving up?

I don’t know that much about perl internals, but I thought that code installed as a $SIG{__DIE__} signal handler would be faster than eval handling. I was wrong. Check this out:

I have these three small test scripts (indentation screwed up with the post):

eval.pl:

#!/usr/bin/perl -wuse strict;

use Benchmark qw(:all);;use constant MAX => 150000;my $code = sub {  eval {    die "this is an exception";  };

  if($@)  {    #continue;  }  else  {

  }};

timethis(MAX, $code);

exception_class.pl

#!/usr/bin/perl -wuse strict;

use Exception::Class ('TestException');use Benchmark qw(:all);

use constant MAX => 150000;

my $code = sub {  eval { TextException->throw(error => "this is an exception"); };  if (UNIVERSAL::isa($@, 'TestException'))  {  }};

timethis(MAX, $code);

error.pl:

#!/usr/bin/perl -wuse strict;

use Error qw(:try);use Benchmark qw(:all);;

use constant MAX => 150000;

my $code = sub {  try {    die "this is an exception";  }  catch Error with  {    #continue;  }  finally  {

  };};

timethis(MAX, $code);

exception.pl:

#!/usr/bin/perl -wuse strict;

use Exception qw(:all);use Benchmark qw(:all);;

use constant MAX => 150000;

my $code = sub {  try {    die "this is an exception";  }  except  {    #continue;  }  finally  {

  }};

timethis(MAX, $code);

How did these compare in execution speed (UPDATES: tests reran to include comparable exception_class.pl info)?

• ./exception.pl: 23 wallclock secs (22.09 usr + 0.11 sys = 22.20 CPU) @ 6756.76/s (n=150000)
• ./error.pl: 9 wallclock secs ( 7.90 usr + 0.02 sys = 7.92 CPU) @ 18939.39/s (n=150000)
• ./exception_class.pl: 2 wallclock secs ( 1.37 usr + 0.00 sys = 1.37 CPU) @ 109489.05/s (n=150000)
• ./eval.pl: 0 wallclock secs ( 0.41 usr + 0.00 sys = 0.41 CPU) @ 365853.66/s (n=150000)

Holy smokes! That’s a huge difference in speed. I’ll be sticking with eval for now, thanks.