Your $csv could use a ->new ({ binary => 1, auto_diag => 1 }) to catch errors better.

Uniqueifying arrays that keep the order can be done much easier. The perlfaq's give even more solutions:

tux$ cat test.pl
#!/pro/bin/perl

use strict;
use warnings;

my @data = (1, 3, 4, 5, 3, 6, 8, 4, 9);

$" = ", ";
print "(@data)\n";

{   my %seen;
    @data = grep { !$seen{$_}++ } @data;
    }
print "(@data)\n";

tux$ perl test.pl
(1, 3, 4, 5, 3, 6, 8, 4, 9)
(1, 3, 4, 5, 6, 8, 9)
[download]