I started with the Mastering Perl Algorithms code (I left it in a subroutine), and abandoned it for the ternary when the ternary was faster. Of course, the subroutine entry was the main speed bottleneck.

This is fascinating. In the if/then/else results, we pay a considerably higher cost for the "constant item," "scalar variable" and "scalar assignment" operations. I wonder why that is so... Even the numeric comparison is more expensive in if/then/else than book max.

Also, your grep benchmarked slower than if/then/else. I saw the opposite. Would you post your complete code, to let me compare?

Thanks, you've made a fantastic post!

Russ

My complete code is a CGI that spits out the results inside the tables, so it's a bit ugly. Here I've commented out the CGI line, so it will print out the HTML without the header. You might want to redirect the output to a file, and view it in a browser.

I did wrap each version into a sub, as a max() should usually be used that way.

Also note that this code REQUIRES perl 5.6.0. If you have an old version, you will need to replace "our" with "use vars"

#!/usr/bin/perl
use strict;
#use CGI qw( header ); print header();
use Benchmark;
use Devel::OpProf qw(profile zero_stats stats );
profile(1);    # turn on profiling
our $now = 8; # we'll pretend it's between 8 and 9 PM
our %url = (
        monday => {
        @{[map(($_,1), (1..1000))]}
        }
        );
#start profiling    
$|++;
print "<TABLE>\n";

print "<TR><TD>\n";
my_profile( "grep",\&test_grep ); 
print "</TD><TD>\n";
my_profile( "if/then/else",\&test_if_then ); 
print "</TD><TD>\n";
my_profile( "book max()",\&test_max ); 
print "</TD></TR>\n";

# now lets benchmark
print "<TR><TD COLSPAN=3><PRE>";
timethese(10000, {
    bench_grep => q{
    test_grep();
    },
    bench_if_then => q{
    test_if_then();
    },
    bench_max => q{
    test_max();
    }} 
      );
print "</PRE></TD></TR></TABLE>\n";

sub test_grep {
    $now = (sort grep {$_ <= $now} keys %{$url{'monday'}})[-1];
}

sub test_if_then {
    $now = ($now < $_ && $_ < 8 ? $_ : $now) for keys %{$url{'monday'}
+};
}

sub test_max {
    foreach ( keys %{$url{'monday'}} ) { $now = $_ if $_ > $now }
}

sub my_profile {
    my $title = shift;
    my $test_sub = shift || return;
    zero_stats;
    &$test_sub;
    stats_td( $title );
#    print_stats;
}

sub stats_td {
    my $title = shift;
    my %stats = %{stats()};
    print "<TABLE><TR><TH COLSPAN=2>&nbsp;$title<BR><HR></TH></TR>\n";
    while ( my ( $stat, $value ) = each %stats ) {
    $stat =~ s/</\&lt\;/;
    $stat =~ s/>/\&gt\;/;
    if ( $value ) {
        $value =~ s/>/\&gt\;/;
        $value =~ s/</\&lt\;/;
        print "<TR><TH>$stat</TH><TD>$value</TD></TR>\n";
    }
    }
    print "</TABLE>\n";
}
[download]

I just whipped this up without much planning, so there are some problems with it. One is that the call to stat() should be in my_profile() not stats_td(). But, because of it's location in stats_td(), it is a O(1) error, and so it shouldn't skew the results much.

Paris Sinclair    |    4a75737420416e6f74686572
pariss@efn.org    |    205065726c204861636b6572
I wear my Geek Code on my finger.