in reply to How to remove duplicates from a large set of keys
In the end, you will still need to have all keys in memory, or at least accessible, that means, you will need some kind of hash, either a plain (in memory) hash, or a tied hash, that you tie to a dbm file for example.
You can possibly save on the side of the keys, by generating the checksums for the keys yourself, by using Digest::MD5 or something comparable, but that will only help you as long as your keys are on average longer than their MD5 length. You can also consider building a trie of your keys by building a linked list of keys with a common start, either a letter or a string. This increases the number of lookups you need to make, but can reduce the amount of memory you need, of your keys are long enough and have enough common prefixes. Still, a million keys shouldn't eat too much memory - about 1 million*32 bytes for the hash entries, plus the length of the keys in bytes.