Perl Hash vs Python Dictionary - a micro benchmark we want to benchmark the performance of populating a hash, mapping a fixed-length 10-byte key to the same value

bench.dict.insert.py

#!/usr/bin/env python
import timeit
REPEATS= 10

using_forloop_total = timeit.timeit(
    """
    d={}
    for n in range(1000000):
        d["%10d" % n] = "%10d" % n
    """, number=REPEATS)

print "using_forloop: s/iter = %f" % (using_forloop_total/REPEATS)


# using map(), generate a dict from tuples
using_map_total = timeit.timeit(
    """
    d=dict(map(lambda n: ("%10d" % n, "%10d" % n), range(1000000)))
    """, number=REPEATS)

print "using_map: s/iter = %f" % (using_map_total/REPEATS)

bench.hash.insert.pl

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

cmpthese(10, {
    using_forloop => sub {
        my $hash = { };
        foreach (1..1_000_000) {
            $hash->{sprintf("%10d", $_)} = sprintf("%10d", $_);
        }
    },
    using_map => sub {
        # for comparision purposes only, this is very very fast if you
        # already have (most of) the data in memory
        my $hash = map { sprintf("%10d", $_) => sprintf("%10d", $_) } (1..1_000_000);
    },
});

__END__

results.md

##Results:

on a MBP 2012 Core i5

Perl 5.12.4

              s/iter using_forloop     using_map
using_forloop   1.73            --          -75%    = ~578k sets/sec
using_map      0.429          303%            --    = ~2.3M sets/sec, inlined ?

Python 2.7.2

using_forloop: s/iter = 1.659030                    = ~602k sets/sec
using_map: s/iter = 1.812623                        = ~551k sets/sec, lambda call overhead ?