David Feuer
David Feuer
I do think your test is a bit unusual. First off, why sum the sets at all? Why not just force them to WHNF? They're totally strict structures. Secondly, why...
@jwaldmann Correct.
> Correct, but it would be wasteful. The `Tip`s contain all the elements, it would be sufficient to just hash the contents of those (ignoring `Bin`s). Oh yeah, I think...
We could offer ```haskell serialize :: (Word -> r -> r) -> r -> IntSet -> r ``` This could be used to hash efficiently, with `serialize` written so it...
I doubt that's quite what we'd want for `Ord`, but maybe close. The use case is that I'd be more comfortable with "publicly" exporting serialize/deserialize than a function that's only...
It should be a matter of a few minutes to strengthen the constraint from `Eq` to `Ord` (a potentially breaking change; others' code may have to be modified to adjust,...
> I think that's about as far as reasonably possible for improving performance. > > For many `PrimMonads` we would like to simply create buffer and just write to it....
I'm going to need to see a proof that the stated time bounds still hold. Or if they don't, some reasonably tight new bound. We need to be sure that...
Did you see my question? The bounds we have for these operations lean on the highly nontrivial published proofs for bounds on intersection and difference. My concern is that by...
One option to consider is to switch to an unbalanced split (or the "hedge" algorithms we used to use) when the sets/maps get small enough (below some fixed size). That...