David Feuer
David Feuer
Thus far, this proposal has gotten good reception on the libraries list.
* Thanks to GHC Trac ticket 16261, matching on `[]` in a rule LHS is pretty dicy. Generalize the write-back rules so we don't have to. * `NOINLINE` set-to-list conversions...
`foldrBits` was previously defined just like `foldr'Bits`, and `foldlBits` was defined just like `foldl'Bits`. That is, each of them built up an accumulator, lazily, across a whole leaf before producing...
`Arbitrary` instances are defined as orphans in the `tests`. This is reasonable, because they're tuned to test the implementations, rather than for users. However, it might be nice to export...
We want ```haskell alterF (const m) k s -- or even alterF (`seq` m) k s ===> seq s $ bool (delete k s) (insert_preserving k s) m ``` to...
The implementations of `foldl'Bits` and `foldr'Bits` look efficient. The implementations of `foldlBits` and `foldrBits`, on the other hand, don't look efficient for their purposes. I'd expect `foldlBits` to look more...
- As usual, the `IO` and `ST` code were actually the same. `PrimMonad` abstracts from that, and also adds support for monad transformers. - Removed utterly bogus `Show` instances for...
I believe we should offer ``` haskell freezeGraph :: (GraphM m mg, Graph g) => mg a b -> m (g a b) thawGraph :: (GraphM m mg, Graph g)...
It should be possible to unify the `IO` and `ST` support, and indeed expand support to other `PrimMonad` instances.
They are not even remotely legitimate and there should never be any reason to use them. IMO, they should be removed promptly in the next major version.