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

Although this seems like the eternal question, as there are many variations on this problem posted on PM, some of which include Benchmarks, I've just found something a little curious when it comes to benchmarking two approaches to this.

I'm converting a 2xn array reference, which is the result of a DBI call (selectall_arrayref), into a hashref. At first, I thought the easy way might be to do something like: 
sub AssembleByHash
{
        my ($array_ref) = @_;
        my (%result);

        foreach my $row (@$array_ref)
        {
                $result{$row->[0]} = $row->[1];
        }

        return \%result;
}

[download]
Which I thought would be far to easy, since it didn't harness the Power of Map. So, instead, I supposed that this might be faster: 
sub AssembleByArray
{
        my ($array_ref) = @_;
        my (%result);

        %result = map { @$_ } @$array_ref;

        return \%result;
}

[download]
After all, it's quick and to the point. When applying Benchmark to validate my theory, though, I got the following result using a 10_000 row test-table and 100 iterations: 
Benchmark: timing 10 iterations of AssembleByArray, AssembleByHash...
AssembleByArray: 10 wallclock secs ( 9.64 usr +  0.04 sys =  9.68 CPU)
+ @  1.03/s (n=10)
AssembleByHash:  1 wallclock secs ( 1.31 usr +  0.00 sys =  1.31 CPU) 
+@  7.63/s (n=10)
                  Rate AssembleByArray  AssembleByHash
AssembleByArray 1.03/s              --            -86%
AssembleByHash  7.63/s            639%              --

[download]
So, I can only suppose that the temporary array created by the map call is killing performance. Or, am I missing something?

Replies are listed 'Best First'.
Re: Multi-Dimensional Array to Hash Conversion Optimization
by MeowChow (Vicar) on Apr 03, 2001 at 20:21 UTC
    There is a performance issue in Perl 5.6 which causes it to take quadratic time for returning multiple list values in map calls, though tilly has produced a patch for this.

    For more detail on this issue, check out this thread. 

       MeowChow                                   
               s aamecha.s a..a\u$&owag.print
[reply]
(boo) Re: Multi-Dimensional Array to Hash Conversion Optimization
by boo_radley (Parson) on Apr 03, 2001 at 20:39 UTC
    Huh? why not use 
    $ary_ref = $dbh->selectall_hashref($statement);

    [download]
    don't make things harder than they need to be! :)
[reply]
[d/l]
      Perhaps because of this:      Can't locate object method "selectall_hashref" via package "DBI::db" at test line 5. Update: It would seem that, as boo_radley pointed out below, there is an upgrade to DBI. Jumped the gun there, I guess, if only because I've assumed that such a thing existed before only to find run-time errors. (Heh, oops) Thanks for the explanation! That puts the stuff below into a pseudo-historical context then...

      I'm hoping that some day they will implement one. However, the idea was that I wanted to take an array containing (key,value) pairs inside their own ARRAY-ref and convert back.

      I had cooked up something like this before: 
      sub selectall_hashref_byid
{
        my ($table) = shift;
        my ($select);

        if (scalar(@_) == 1)
        {
                # Single key to fetch, so no HoH

                return { @{$db->selectall_arrayref("SELECT id,? FROM $
+_", $_[0])} };
        }

        if (@_)
        {
                unshift (@_, 'id') unless (@_ && grep ("id", @_));
        }
        else
        {
                @_ = ('*');
        }

        my ($hash, $row, $query);

        $hash = {};

        $query = $db->prepare ("SELECT ".join (',', @_)." FROM $table"
+);
        $query->execute ();

        while (my $row = $query->fetchrow_hashref)
        {
                # Stoke it
                $hash->{$row->{id}} = $row;

                # Deredundantize it
                delete $hash->{$row->{id}}->{id};
        }

        return $hash;
}

      [download]
      The only downside is that the crazy casting that goes on in there doesn't handle errors very well, so it is really flying by the seat of your pants.
[reply]
[d/l]
[select]
        *smacks head*
        not everyone upgrades as rapidly as you do, boo...
        If you can, hie thee unto cpan: 
        Changes in DBI 1.15,    28th March 2001
...
  Added selectall_hashref thanks to Leon Brocard.
...

        [download]
[reply]
[d/l]

Back to Seekers of Perl Wisdom