David Feuer
David Feuer
Not all the reconstructed pieces end up getting incorporated. For `intersection`, none of them do. (I wish that were true of `difference` as well, but we follow the exact algorithm...
Why not be more general? ```haskell class Monad m => Disposable m a where dispose :: a %1 -> m () ```
Which order should the class parameters go in? I can't decide which partial application seems more likely.
`unions` should be fine. However, some of the conversions from lists rely pretty heavily on pattern matching the list. I don't know how to make a fromFoldable for Data.Sequence that's...
Hmm.... For compatibility with potential non-GHC implementations, this needs a few changes: 1. Don't use `MagicHash`. Even though it's confusing, we'll need to use something else to distinguish these unboxed...
Making progress... still don't know if performance will work. Do we have any map/coerce tests? I forget.
@meooow25 Do you happen to have the bandwidth to see what effect this PR has on the benchmarks?
Thanks a lot! My computer is old and wimpy and I have no time right now. I expect some things to slow down slightly, which is probably okay if not...
Thanks a lot! I'm glad to see that performance does not seem to decrease. Unfortunately, I believe the two apparent improvements are almost certainly benchmarking artifacts. I'm disappointed that there...
Hmm.... It's not at all clear to me that having a custom `mapMaybeWithKey` implementation is actually better than just doing ```haskell mapMaybeWithKey f = fromDistinctAscList . Data.Maybe.mapMaybe (\(k, v) ->...