map in void context is no longer a sin, but still it can't be used lightly instead of for, which is 25% faster than map.

It has been a great improvement from previous versions, though.

use Benchmark;

print "Perl $]\n";

timethese (10000,{
    "map"=> sub { map { $_ } ( 1 .. 1000 ) },
    "for"=> sub { $_ for ( 1 .. 1000 ) }
}
);

__END__

Perl 5.006001
Benchmark: timing 10000 iterations of for, map...
       for:  3 wallclock secs ( 2.81 usr +  0.00 sys =  2.81 CPU)
       map:  5 wallclock secs ( 5.16 usr +  0.01 sys =  5.17 CPU)

Perl 5.008
Benchmark: timing 10000 iterations of for, map...
       for:  3 wallclock secs ( 2.19 usr +  0.01 sys =  2.20 CPU)
       map:  4 wallclock secs ( 4.65 usr +  0.02 sys =  4.67 CPU)

Perl 5.008001
Benchmark: timing 10000 iterations of for, map...
       for:  2 wallclock secs ( 1.72 usr +  0.00 sys =  1.72 CPU)
       map:  2 wallclock secs ( 2.27 usr +  0.02 sys =  2.29 CPU)

Perl 5.008001
Benchmark: timing 50000 iterations of for, map...
       for:  9 wallclock secs ( 8.71 usr +  0.01 sys =  8.72 CPU)
       map: 12 wallclock secs (11.53 usr +  0.04 sys = 11.57 CPU)
[download]

Well, now you are comparing map/block with for/expression. Blocks are more expensive than expressions. And while this doesn't mean for isn't slightly faster than map, it does skew the results. Here's another benchmark, where we do a very simple thing in the block/expression:

#!/usr/bin/perl

use strict;
use warnings;

use Benchmark 'cmpthese';

cmpthese -2 => {
    map_block   =>  sub {my $s = 0; map {$s +=$_} 1 .. 1000;    $s},
    for_block   =>  sub {my $s = 0; for (1 .. 1000) {$s += $_}; $s},
    map_expr    =>  sub {my $s = 0; map $s += $_ => 1 .. 1000;  $s},
    for_expr    =>  sub {my $s = 0; $s += $_ for 1 .. 1000;     $s},
}
__END__
            Rate map_block  map_expr for_block  for_expr
map_block 3224/s        --       -4%      -10%      -17%
map_expr  3345/s        4%        --       -6%      -14%
for_block 3572/s       11%        7%        --       -8%
for_expr  3900/s       21%       17%        9%        --
[download]

map is slower than for, but the difference is in the same order as the difference between a for/expression and a for/block.

Abigail

Abigail