Hashes take up a lot of space, and sorting requires making a copy of all the keys. If memory is a big issue, you can sort and process just the input, then do a line-by-line merge with the dictionary file, placing your output in a secondary file. Then when you finish, delete the original and rename the secondary. This way you never have to store more than the input x2 in memory, which is far better than (dictionary + input) x2. You're also working with arrays, not hashes, and you don't have to sort the dictionary, just the input.