Second, if you really want to code it yourself, and depending on how large the menu of possible selections is, you could generate lookup hashes ahead of time. If, for example, you have 200 possible surnames and 50 possible ages, then you could generate a hash with keys 'surname-age' (e.g. 'smith-25') whose values are something like a reference to a list of indices into @Persons (e.g. [ 10, 102, 145, ... ] ). Seems like a classic "space-vs-speed" trade off.

An intermediate approach might be to somehow cache the results of each filter loop, so that the slow part only happens once.