and there may be multiple rows per key (although each row may well have different information for that key)

You cannot have "multiple rows" per key. That means your "key" is not a "key" since it does not uniquely identifies one record and one record only. It also is a strong indication that your data file is not normalized. That is not bad, it just makes life a lot more difficult.

But your problem has in essence nothing to do with keys and database design. It is simply: upon the first occurrence of a certain value in a certain field, do something. Easy as pie!

use Modern::Perl;

my %keys;

while (<DATA>) {
    my ($key, $data1, $data2) = split /,/;
    unless (exists($keys{$key})){
        # do something
        say $key; # just as an example
        $keys{$key} = 1;
    }
}

__DATA__
key1, abd, 123
key2, dhde, 9+6+
key3, dxdc, edazedaz
key1, dea, 564dz
key1, deksizi, 4833
key4, eoz, 852662
key2, dzdadazd, 9566
[download]

Output:

key1
key2
key3
key4
[download]