PSA: Be more careful about type classes included in cats(-core)

[adelbertc]

Filing this as an issue just to have a central place to talk about this. It seems Cats is getting larger and larger with a lot of perhaps niche type classes. One of the goals of Cats was to be more modular/not grow into this gigantic project with everything included.

Examples that come to mind:

• FunctorFlatten Add AlternativeFlatten #1337
• Cocartesian Add Cocartesian #1362
• Reducer Reducer typeclass #1360
• LeftFoldable Add a trait LeftFoldable[F[_]] such that Foldable[F[_]] <: LeftFoldable[F[_]]  #1339
• Copair Created Copair typeclass #976

I'm not saying we shouldn't include these, or that the current set of type classes we already have are blessed - I just want to bring up the fact I think we should really think about if a type class is necessary/useful/general enough to be in a library like cats-core before including it.

cc-ing some folks who look to be active on those tickets @non @ceedubs @johnynek

[eed3si9n]

+1 on keeping things lean esp if Cats 1.0 is going to have multi-year bincompat story.

[johnynek]

I think this is a good thing to consider for sure. I don't think we want to add every type-class imaginable.

For what it's worth, I submitted mine only because it is very common in work I do (more so, than the less general): https://github.com/typelevel/cats/pull/1225/files (that said, I'm starting to wonder if it is not just a method on Alternative).

[ceedubs]

I'm going to go ahead and close this out. Cats is past the 1.0 release and has been pretty stable. There are good points brought up here, but this issue is titled in such a way that will forever resist being closed out, so I'm going to go ahead and do so now :P