Re^5: better array to hash conversion

[tobyink]

The following map variant works out significantly faster than the previously posted map variants, though not as fast as for-loop based variants:

sub variation3ep {
  my $idx = -1;
  my %hash = map(($_ => ++$idx), @array);
}

While the block form of map is often favoured for its clarity, the expression form of map is faster because it avoids the overhead of creating a lexical pad.

This is actually another reason why BrowserUk's variation4 is fast - he uses the for statement modifier rather than a for loop with a block. Compare:

#!/usr/bin/env perl
use strict;
use warnings;
use Benchmark qw(:all);

my @array='aa' .. 'zz';

sub variation4 {
    my $idx = 0;
    my %hash; $hash{ $_ } = $idx++ for @array;
}

sub variation5 {
    my $idx = 0;
    my %hash;
    for (@array) {
        $hash{ $_ } = $idx++;
    }
}

cmpthese(-3, {
    'variation4' => \&variation4,
    'variation5' => \&variation5,
});

__END__
            Rate variation5 variation4
variation5 740/s         --        -4%
variation4 773/s         5%         --

Yes, it's a small difference, but it's pretty consistently observable.

The absolute fastest I've been able to achieve is a small variation on BrowserUk's variation4 using the prefix increment rather than postfix increment:

sub variation6 {
    my $idx = -1;
    my %hash; $hash{ $_ } = ++$idx for @array;
}

It seems to give you about a 6% speed up.