Add some bifunctors-style newtypes

[garyb]

The name Cowrap is probably a bit odd, but I haven't come up with anything better yet.

[garyb]

/cc @safareli

[paf31]

I wonder if it's worth providing the adjoint, which constructs a profunctor from an invariant.

[paf31]

Personally, I'd rather just remove Invariant, but that's a separate, larger conversation 😄

[garyb]

Wait! I just realised something, this is purescript-deprecated/purescript-monoid#34 if we provide

instance semigroupJoin :: Semigroupoid p => Semigroup (Join p a) where
  append (Join a) (Join b) = Join (a <<< b)

instance monoidJoin :: Category p => Monoid (Join p a) where
  mempty = Join id

[paf31]

👍

[garyb]

Any idea what the adjoint would look like? I just took a stab at it, but not really sure what I'm doing there and couldn't figure out how to implement dimap.

[paf31]

I found the previous discussion of this here.

[garyb]

Ah yeah! I remember now. Will add that too then.

[paf31]

Feel free to give it a better name 😄

[garyb]

Cojoin was all I had, I'm not sure it's any better 😆

[paf31]

Split possibly? Since we're trying to split the type argument into two.

[paf31]

👍 LGTM. Curious though - is there a use case for these?

[garyb]

Curious though - is there a use case for these?

Uhh... most of them are pretty speculative 😉

It came from a discussion with @safareli, he was asking if we could impose Functor and Contravariant superclasses, or provide instances for Functor and Contravariant for every Profunctor - since we have endofunctors, no, but I figured providing newtypes for them would "prove" the relationship is there, and perhaps be occasionally useful.

Then when I started on that, I figured that profunctors are still a bifunctors, just not in the purescript sense, so I went back to BF to see which of the various newtypes may translate to PF. So Clown and Joker are definitely a bit 🤷‍♂️, and Join was right up until two minutes ago when I realised it's our abstracted Endo 😆

[paf31]

Do you think the names Clown and Joker might be confusing though?

[safareli]

there is already Joker and Clown defined in https://pursuit.purescript.org/packages/purescript-bifunctors

[garyb]

I'm happy to change them, just figured since they work the same way for bifunctors, reusing the names here was reasonable.

[garyb]

There we go, Split sounded good to me.

[paf31]

Can we hide the constructors here?

[garyb]

Oops, good call.

[paf31]

👍 LGTM, thanks!

[paf31]

I just thought - should this have reused the existing newtypes? Bit late now, but probably worth considering.

[garyb]

I just explored that a bit...

• We could reuse Clown and Joker, defining the instances in Data.Profunctor
• Wrap doesn't work as it has an overlapping Functor instance (one with Bifunctor and one with Profunctor constraints on p)
• Join could work, but we'd have to define those "endo" instances with the newtype, which would mean some dependency reordering - it probably works out, but bifunctors and monoid are both near the top of the hierarchy, so I'm a little hesitant to mess with them right now.

I think next time we do breaking releases for all the core libraries we can take a look at it again though 👍