Here is another way to do...

I don't think it is as fast as could be, but I believe that Corion's initial posting assumes that you are comparing with only the previous line, and also that the last line is always 'good' to keep (but what do I know, he is a Saint, and I am a chump, no wait, an Acolyte, about the same thing though).

#!perl -w
use strict;

open (DATAFILE,"D:/JUNK/data.txt");
my $file_content = do {local $/; <DATAFILE> };
my @fa1 = split(/\n/,$file_content);

my $cnt=0;  my %hv=();
#Set to hash with a count to retain order of list
foreach (@fa1) {
  $hv{$_}=$cnt;
  $cnt++;
}

my %already_deleted=();
foreach my $v1 (@fa1) {
  if (exists($already_deleted{$v1})) { next; }
  my $len1 = length($v1);
  foreach my $v2 (keys %hv) {
    if ((exists($already_deleted{$v1})) || (exists($already_deleted{$v
+1})))  { next; }
    my $len2 = length($v2);
    if (($len2 > $len1) && (substr($v2,0,$len1) eq $v1)) {
      #Delete the hash key value
      delete $hv{$v1};
      #Store deleted hash values to help move through the
      #foreach statements quicker with next statements
      $already_deleted{$v1}=1;
    } else {
      #Nothing
    }
  }
}

#Retain the original order, then print 'em out.
my @keys = sort{$hv{$a} <=> $hv{$b}} keys %hv;
foreach (@keys) {print "$_ \n"};



___DATAFILE___
Topology/IPClassA
Topology/IPClassA/Device
Topology/IPClassA/Device/log_ratio
Topology/IPClassA/Device/poll_interval
Topology/IPClassA/Device = 2
Topology/IPClassA/Device = 2/poll_interval
Topology/IPClassA/Device = 2/is_managed
Topology/IPClassA/Device = 2/Port
Topology/IPClassA/Device = 2/Port/ifPhysAddress
Topology/IPClassA/Device = 2/Port/poll_interval
Topology/IPClassC
Topology/IPClassC/Device
Topology/IPClassC/Device/poll_interval
Location
[download]