RE: Blessables -- What Can You Make Into Objects?

[perlmonkey]

in reply to Blessables -- What Can You Make Into Objects?

This was a cool article. Dont know how useful, but very interesting.
I am sure I will find a use for it somewhere.

I tried to play with blessing RegEx's and ran into trouble.
You mention that you might be able to bless a s{}{}, but I
couldn't figure it out.

I wanted an object that would do "s/^\s+|\s+$/g" for me.
It would not compile if i did somethine like
$self = s{^\s+|\s+$}{}; so I ended up using qr{} instead,
putting the s/// in the member function:

package Cleaner;
sub new 
{
    my $class = shift;
    $class = ref($class) || $class;
    my $self = qr{^\s+|\s+$};
    return bless($self, $class);
}

sub clean 
{
    $_[1] =~ s/$_[0]//g;
}

package main;
$c = new Cleaner;
$foo = "\t\tFOO\t\t";
$c->clean($foo);
print $foo, "\n";

So this works, but I dont know it is the most optimized
routine. I did some benchmarks:

package main;
$c = new Cleaner;
$foo = "\t\tFOO\t\t";
use Benchmark;
timethese(1000000, {
    'blessed' => sub { my $bar = $foo; $c->clean($bar) },
    'normal'  => sub { my $bar = $foo; $bar =~ s/^\s+|\s$//g },
});

And got these results:

Benchmark: timing 1000000 iterations of blessed, normal...
   blessed: 43 wallclock secs (36.50 usr +  0.14 sys = 36.64 CPU)
    normal: 28 wallclock secs (24.19 usr +  0.08 sys = 24.27 CPU)

I guess the difference is just the function call, but thought maybe there is a faster way. Any Ideas?

[chromatic]

Looks like I'm wrong about being able to bless a s{}{} statement. That would be truly convenient. I'll change that in the tutorial.

In your benchmarking, lots of the time *is* taken up by the function call. Here's what I ran to prove that. (This has to be done in the Cleaner package, for obvious reasons.):

timethese(1000000, {
    'blessed' => sub { my $bar = $foo; $c->clean($foo); },
    'normal'  => sub { my $bar = $foo; $bar =~ s/^\s+|\s$//g; },
    'super' => sub { my $bar = $foo; $bar =~ s/$c//g; },
});

There's not nearly the performance penalty involved:

Benchmark: timing 1000000 iterations of blessed, normal, super...
   blessed: 10 wallclock secs ( 9.83 usr +  0.00 sys =  9.83 CPU)
    normal:  4 wallclock secs ( 5.42 usr +  0.00 sys =  5.42 CPU)
     super:  7 wallclock secs ( 6.71 usr +  0.00 sys =  6.71 CPU)

I suspect that if the regex were more complicated (and $foo were much longer), the difference wouldn't be quite as intense.