Looks like a reasonable change to me. It seems odd to pull in a tuple library just for that one usage. In the memoize code I maintain, I just use the sequence of args as-is -- I wouldn't even bother unrolling them into a vector -- as I recall. But maybe that's the key to it being a "fast" memoize instead of the regular memoize?

My guess (without doing the benchmarks) is that the most "fast" comes from using a mutable map instead of PHM.

Merged

@U04V5VAUN, do you want to cut the new release, since you merged? It looks like you and I are the only ones with potemkin deploy perms on clojars. I'll ask Zach if he can designate others as admins.

Yah, I can do that. Will be tonight (CET) though.

0.4.7 is out