Are you working on a Unix system? If so you might want to use some of the available Unix tools. You could use "cut" to splice out the URL, feed the results to a file which could then "sort". Once that sort is completed, you could then easily count of the occurances of each URL without having to store a large number of lines or create many temp files. After the sort you could do something like:

# Untested
my $current = '';
my $count = 0;
while (<>) {
   if ( ($current ne $_) && ($current ne '') ) {
      print "$current   ::   $count  \n";
      $count = 0;
      $current = $_;
   } else {
      $count++;
   }
}
[download]

You would invoke at the command line as ./foo.pl < sorted.file > file.count

Since the file is already sorted for you and contains only the URL, all of each URL will be grouped together. Therefor, once a URL changes you will know that you are done counting a particular URL. No need to store in memory any more than the current URL and the current count; Once the URL changes you dump out the count and move on to the next one.