I'd argue that most programmers do, or are supposed to. Either because they are not the designers of the system they build and what they are expected to do is to "fill in the blanks and make it work", or because they implement something that operates with another system and should strictly adhere to the interface/protocol.

The ones who have anything to say about the design are usually the system designers (duh), small project contributors (because there is no spec) and UI people (because they're designers). So yes - if your spec is "1. make it work on this input; 2. make it fast", then making it work in O(1) for this and only this input is basically the best thing you can do (unless you can make it faster for every input in less time of course). Spending time on generalising it is based on second-guessing the intent.

I'd even go as far as saying that specs have absolutely no intent. You have an intent and describe it while writing the spec and you'd better formalise your intent precisely or you will end up with something you didn't request. If you want something on every input, you'd better write it down (when you give a task to someone else).

Actually no. Here's a BSD one: http://hackage.haskell.org/package/judy

(Yeah, I do have some data to back that up - I wrote & benchmarked an autocomplete implementation for a financial software firm. String tries were roughly 10x slower than binary searching an array for a 20k word corpus, which is probably around the size you're dealing with here. They have terrible cache locality - each character makes you follow a pointer, which is probably a cache miss, and the total size of the trie is on the order of 26 * 8 * 20k = 4M. They become a bit better if the number of distinct words in the corpus is small (so that everything fits into cache) or if the number of words is huge (so that your hashtable or array blows the cache anyway).