Charlie, I believe you are misusing Big O notation in this case. For small values of n, the constants outweigh the multiplier. 10 * log(4) is greater than 1 * (4). What you are saying is absolutely true if your switch statements end up being 50 or 100 or 500 statements long, assuming that a binary or hash search can be used in such extreme cases. But for a set of simple comparisons, say, 10 long, your average case will be 5 (big assumption). Is that faster than 3 or 4 comparisons, plus associated overhead? The hash code is probably faster (maybe - that's an amortized analysis), but there is still overhead there.

I could be convinced otherwise. For practical purposes, I would have to see benchmarks in regular production apps that showed this was a significant part of execution time (which I have doubts about) and that adopting a different algorithm would improve things significantly. Ref: http://en.wikipedia.org/wiki/Amdahl's_law - Speedup in a Sequential Program