On 5/7/12, H. S. Teoh <hst...@quickfur.ath.cx> wrote: > you may be better off adapting D's AA implementation > to have an AA that consists of just keys.
Yeah I do that in my code right now. It's pretty easy, just write an some opApply/opBinary functions, add to!string(innerhash.keys) for good measure and you're set (heh, set). > Basically, you still have the > hash table and buckets, just without the value field in each slot. (I > wonder if it's possible for the new AA implementation to support this by > specifying void as the value type.) I think that can be done now with void[0], e.g.: void[0][int] set; I guess the way to add keys is to use a stub variable: void[0][int] set; void[0] val; set[1] = val; But I'm not sure how well serialization libraries can handle this type, for my code I just use int[Key] in a wrapper struct.
