Revisit tuples - Githubissues

treeowl commented 3 years ago

Previously, <<*>> and biliftA2 were strict in their tuple arguments. Unfortunately, the Bifunctor instances (defined in base), are lazy in their tuple arguments. This inconsistency made tuples utterly useless for Biapplicative traversals. Make <<*>> and biliftA2 lazy for tuples.
Add Stricter and Lazier newtype wrappers. Stricter produces strict Functor, Applicative, Bifunctor, and Biapplicative instances from (possibly) lazy ones. Lazier produces lazy instances from (possibly) strict ones for product types, using GHC.Generics.

treeowl commented 3 years ago

Question: should Stricter and Lazier have Bifoldable, Foldable, Bitraversable, and/or Traversable instances? The Foldable and Bifoldable ones are trivial, albeit verbose. What about traversals? It's much less obvious to me what those are supposed to look like.

treeowl commented 3 years ago

Another question: If base adds a Functor constraint to Bifunctor, will it harmonize the Functor and Bifunctor instances, or are those doomed to be forever mismatched?

treeowl commented 3 years ago

Oh, I guess the traversals are dead simple too, but I'm not sure we really want them. Any thoughts?

treeowl commented 3 years ago

Hmm... Not sure why this is struggling to build. But I'm thinking maybe most of this PR doesn't belong in this package.

ekmett / bifunctors

Revisit tuples #94