map versus for

dHarry has asked for the wisdom of the Perl Monks concerning the following question:

Dear Monks

I was contemplating over the 256 commandments from Damian’s "Perl Best Practices" when I encountered:

List Generation

Use map instead of for when generating new lists from old.
[download]

Being more used to other programming languages I often use the non-Perl approach to get the job done. I would typically use a for and probably not even consider the map. So this was an eye-opener for me. I decided to do a little test to see how much the difference is between the two.

(FYI: Perl v5.8.8 built for MSWin32-x86-multi-thread running on a Dell INSPIRON 9400)

I use the example as mentioned by Damian and the Benchmark module to test:

use strict;
use warnings;
use Benchmark qw(:all);

my @results;
my $count = -5;

# Populate list with 10 mio numbers
for (my $i=0; $i<1000_000; $i++) {
    push @results, $i;
}

cmpthese (
    $count,
    {
        for => "test_for;",
        map => "test_map;",
    }
    );

timethese($count, {
       for => "test_for;",
        map => "test_map;",
                  }
);


sub test_for {
    my @sqrt_results;
    for my $result (@results) {
        push @sqrt_results , sqrt($result);
    }
}

sub test_map {
    my @sqrt_results = map { sqrt $_ } @results;
}
[download]

First the comparison:

$count=-1
(warning: too few iterations for a reliable count)
      Rate  for  map
for 2.67/s   -- -10%
map 2.98/s  12%   --

$count=-5
      Rate  map  for
map 3.05/s   -- -16%
for 3.61/s  18%   --

$count=-10
      Rate for map
for 2.73/s  -- -8%
map 2.95/s  8%  --
[download]

Hmmm, not really impressive this “gain” of using map over for?!

Next some timing:

$count=-1

Benchmark: running for, map for at least 1 CPU seconds...
       for:  1 wallclock secs ( 1.16 usr +  0.00 sys =  1.16 CPU) @  3
+.46/s (n=4)
       map:  2 wallclock secs ( 1.19 usr +  0.00 sys =  1.19 CPU) @  3
+.37/s (n=4)

$count=-5
Benchmark: running for, map for at least 5 CPU seconds...
       for:  6 wallclock secs ( 5.22 usr +  0.00 sys =  5.22 CPU) @  3
+.64/s (n=19)
       map:  6 wallclock secs ( 5.22 usr +  0.00 sys =  5.22 CPU) @  3
+.45/s (n=18)

$count=-10
Benchmark: running for, map for at least 10 CPU seconds...
       for: 10 wallclock secs (10.11 usr +  0.00 sys = 10.11 CPU) @  3
+.46/s (n=35)
       map: 10 wallclock secs (10.03 usr +  0.00 sys = 10.03 CPU) @  3
+.29/s (n=33)
[download]

Am I missing something? Is the example given by Damian a poor example? Should I really favor map over for when I want to generate a new list from another list?

Thanks upfront

Update

Beside the obvious advantages: less code, easier to understand, it is stated that map is normally considerably faster.

Back to Seekers of Perl Wisdom