Re: Hash Search is VERY slow

When working with a large file, don't read the whole file into an array. Process the file line by line. Replace

foreach (@list)
[download]

while (<>)
[download]

Also, when populating the hash, Perl can do most of the work for you:

push @{ $ipURL{$ip} }, $url;
[download]

This line can replace the whole if..else... construct. If the key doesn't exist in the hash, it will be created, and the dereference will create an anonymous array to which the url will be pushed. If the key exists, the url is simply pushed to the referenced array.

Whether it's also responsible for the slowdown, I have no idea.

Another place where you can avoid creating a temporary variable is the splitting. You can populate the two variables directly by subscripting the result of split:

my ($ip, $url) = (split /,/)[7, 31];
[download]

I used the following to generate the input:

perl -wE 'for (1..10000) {say join ",", map int rand 1000, 1..35}' > i
+nput
[download]

This was the benchmark:

#!/usr/bin/perl
use strict;
use warnings;

open my $FH, '<', 'input' or die $!;

sub slow {
    my ($fh) = @_;
    seek $fh, 0, 0;
    my @list = <$fh>;
    my $linecounter;
    my %ipURL;
    foreach (@list) {
        $linecounter++;
        my @message=split(',',$_);
        my $ip=$message[7];
        my $url=$message[31];

        if (!(exists $ipURL{$ip})) {
            my @urlList;
            push(@urlList,$url);
            $ipURL{$ip}= \@urlList;
        } else {
            my @urlList=@{$ipURL{$ip}};
            push (@urlList,$url);
            $ipURL{$ip}=\@urlList;
        }
        if (!($linecounter % 50000)) {
                print "Lines: $linecounter\n";
        }
    }
    return \%ipURL
}

sub fast {
    my ($fh) = @_;
    my %ipURL;
    seek $fh, 0, 0;
    while (<$fh>) {
        my ($ip, $url) = (split /,/)[7, 31];
        push @{ $ipURL{$ip} }, $url;
    }
    return \%ipURL
}

use Test::More tests => 1;
is_deeply slow($FH), fast($FH), 'same';

use Benchmark qw{ cmpthese };
cmpthese(-3, {
    slow => sub { slow($FH) },
    fast => sub { fast($FH) },
});
[download]

This is the result on my machine:

1..1
ok 1 - same
       Rate slow fast
slow 20.7/s   -- -26%
fast 27.9/s  35%   --
[download]

map{substr$_->[0],$_->[1]||0,1}[\*||{},3],[[]],[ref qr-1,-,-1],[{}],[sub{}^*ARGV,3]

Comment on Re: Hash Search is VERY slow Select or Download Code