Your code confuses me, but that could be lack of coffee on my part. The way I understand it, for each new name/surname combination, you want to check whether it already exists in a file, and if not, add it to that file. If your file is not that big, your best bet is probably doing this with a hash of hashes:

#!/usr/local/bin/perl -w

use strict;

# Read in file and build HoH
open INPUT, "<names.csv" or die "Blerch: $!\n";
my %existing_names;
foreach $line (<INPUT>) {
  my ($name,$surname)=split /,/, $line;
  $existing_names{$surname}{$name}++;
}  
close INPUT;
...
# Check and add new record
sub duplicate_record {
  
  my $name=shift;
  my $surname=shift;

  if (exists ($existing_records{$surname}{$name})) {
    do_complain_about_duplicates();
    return;
  }
 
  $existing_records{$surname}{$name} 
  
}
[download]

At the end of the program, you would write back the file from the hash, like:

open OUTPUT, ">names.csv" or die "Hcrelb: $!\n";
foreach my $surname (keys %existing_names) {
   print OUTPUT "$_,$surname\n" for keys %{$existing_names{$surname}};
}
close OUTPUT;
[download]

This code basically builds a hash of hashes out of your existing records, with surnames on the first level and names on the second level, which will make it easier to search for existing records later on. Note that none of this code is tested, so it may fail in unpredictable ways, not excluding actually functioning as expected.

CU
Robartes-