My results, Linux on an IBM Netfinity (Intel)

Benchmark: running regexp, transl, each for at least 10 CPU seconds...
    regexp: 10 wallclock secs (10.59 usr +  0.00 sys = 10.59 CPU) @ 37
+274.69/s (n=394739)
    transl: 13 wallclock secs (10.46 usr +  0.00 sys = 10.46 CPU) @ 31
+3981.07/s (n=3284242)
           Rate regexp transl
regexp  37275/s     --   -88%
transl 313981/s   742%     --
[download]

That is a significant differential there for this simple task. A full regexp engine is a big thing to throw at a lightweight string scan. My benchmark code follows:

use strict;
use Benchmark qw(cmpthese);
use vars qw( $x );

$x = 'This-is-a-test-string-I-just-typed-in-for-fun';

cmpthese (-10, {
    'transl' => '$x =~ tr/-/_/; $x =~ tr/_/-/;',
    'regexp' => '$x =~ s/-/_/g; $x =~ s/_/-/g;',
} );
[download]

Oh yeah, I sure am happy Benchmark exists too. =)

Doh! Update: that assignment was:
$x = 'This_is_a_test_string_I_just_typed_in_for_fun';
It wasn't result impacting, just stupid since it no-ops half my test. Interestingly, if I change the string to one with spaces rather than the '-' or '_' I wind up with regexp being 50-60% faster at doing nothing but scanning with no changes...

--
$you = new YOU;
honk() if $you->love(perl)

Hey Extremely, Thanks for taking a look at the code and letting me know where the benchmark may have messed up:

#!/usr/local/bin/perl -w
  use strict;
  use Benchmark;
  my $count =500000;
  ## Method number one
  sub One {
     my $data = 'for bar baz';
    my($outstring);
     ($outstring = $data) =~ tr/-/_/; 
  }

  ## Method number two
  sub Two {
     my $data = 'for bar baz';
     my($outstring) = $data;

     $outstring =~ s/-/_/g;
  }

  ## We'll test each one, with simple labels
  timethese (
    $count,
    {'Method One TR' => '&One',
     'Method Two S' => '&Two',
     }
  );

  exit;
[download]


Claude

Your test strings don't have any '-' hyphens in them. =) And you change assignment forms between the two subs which honestly shouldn't have much effect but is still questionable practice when benchmarking since you want the code identical in nature except for the key point you are testing...

Also, since the $data is set in the sub it will be reset every pass. As well, you can just set up the benchmark like this and avoid having perl eval a sub call in a string:

timethese ( $count,
  {
    'Method One S' => sub {
      my $data = 'foo-bar-baz';
      $data =~ s/-/_/g;
    }
    'Method Two TR' => sub {
      my $data = 'foo-bar-baz';
      $data =~ tr/-/_/;
    }
  }
);
[download]

Not a big deal but it may save you some typing in the future. Your way has the benefit of being easy to run the subs once and test their output, tho...

BTW, I moronically goofed my test string too... =)

--
$you = new YOU;
honk() if $you->love(perl)

#!/usr/local/bin/perl -w
use strict;
use Benchmark;
my $count = 900000;

## Method number two
sub One {
   my $data='for-bar-baz';
   $data =~tr/-/_/;
}
## Method number Two
sub Two {
   my $data='for-bar-baz';
   $data =~s/-/_/g;
}
## We'll test each one, with simple labels
timethese (
  $count,
  {'Method One TR' => '&One',
   'Method Two s'=> '&Two'
   }
);
exit;
[download]

Benchmark: timing 500000 iterations of Method One TR, Method Two s...
Method One TR:  2 wallclock secs ( 1.87 usr +  0.00 sys =  1.87 CPU) @
+ 267379.68/s (n=500000)
Method Two s:  5 wallclock secs ( 4.84 usr +  0.00 sys =  4.84 CPU) @ 
+103305.79/s (n=500000)
[download]