Question

I have a working perl script that scans a directory and uses imgsize
http://dktools.sourceforge.net/imgsize.html to get the width, etc of png files. Does anyone have any tips for speeding up this process (right now, it averages 5 minutes for every 1000 files)? I was just wondering if the code could be optimized some how. Thanks.

use strict;
use warnings;

use File::Find;

my @files;
my $directory = '/Graphics/';
my $output_file = '/output_file';
my $max_height = 555;
my $count = 0;

open ( OUTPUT, '>>', $output_file );

find( \&wanted, $directory );

foreach my $file ( @files ) {
        if ( $file =~ /\.png$/ ) {
                my $height = `imgsize $file | cut -d\'\"\' -f4`;
                if ( $height > $max_height ) {
                        print OUTPUT "$file\n";
                }

                $count++;

                my $int_check = $count/1000;
                if ( $int_check !~ /\D/ ) {
                        print "processed: $count\n";
                }
        }
}

print "total: $count\n";
close ( OUTPUT );
exit;

sub wanted {
  push @files, $File::Find::name;
  return;
}

---

Solution: Turns out that I was able to use the Image::Info module. I went from processing 1000 imgs every 5 minutes to every 12 seconds. Here’s the relevant snippet of code, if anyone is interested:

 use Image::Info qw(image_info);

    foreach my $file ( @files ) {
            if ( $file =~ /\.png$/ ) {
                    my $output = image_info($file);
                    my $height = ${$output}{height};

                    if ($height > $max_height) {
                            print OUTPUT "$file\n";
                    }

                    $count++;

                    my $int_check = $count/1000;
                    if ( $int_check !~ /\D/ ) {
                            print "processed: $count\n";
                    }
            }
    }

Answer 1

The Perl code you’ve shown is likely not the culprit. You can profile it with Devel::NYTProf, just like @choroba has said. But I’d bet money that most of the time comes from forking two external processes per image (imgsize and cut). You should look into Perl modules that can retrieve the image’s height without running any external process. Modules like Image::Info come to mind.