Fair enough. Your original problem statement presented the situation as you had a list of words and were searching to determine which words had the search string as a prefix or suffix. That is what your original code did, and what your second code did not do.

I agree that storing a list of prefixes (in one hash) and suffixes (in another hash) is definitely preferable (if possible) and will result in a huge savings in time (and probably memory, too!). Restating the problemspace is a very good thing to do, when you're stuck like that.