Derive instance {Enum, Bounded, Num, Real, Integral} Compose f g a

[Bodigrim]

Let's add

deriving instance Enum (f (g a)) => Enum (Compose f g a)
deriving instance Bounded (f (g a)) => Bounded (Compose f g a)
deriving instance Num (f (g a)) => Num (Compose f g a)
deriving instance Real (f (g a)) => Real (Compose f g a)
deriving instance Integral (f (g a)) => Integral (Compose f g a)

I'd argue that the absense of such instances is just an omission, but if anyone is curious, my particular use case deals with things like Compose Odd Positive Integer, where newtype Odd a and newtype Positive a are wrappers with desired Arbitrary instances.

[mixphix]

Are there other Data.Functor.* types that are missing instances?

[chshersh]

I don't mind this addition if it helps solve a real problem. I believe, all those typeclasses are composable with transparent semantics?

[mixphix]

This permits the literal 1 :: Compose Odd Positive Int. I'm not sure this is desired behaviour.

[Bodigrim]

This permits the literal 1 :: Compose Odd Positive Int. I'm not sure this is desired behaviour.

As long as one is fine with 1 :: Odd Int and 1 :: Positive Int (which might be somewhat questionable, I agree), I don't see an issue with 1 :: Compose Odd Positive Int.

Basically the proposal is just to use {-# LANGUAGE GeneralizedNewtypeDeriving #-} for newtype Compose f g a = Compose (f (g a)). If for some reason there is already instance Foo (f (g a)), there is very little to win from not providing instance Foo (Compose f g a), it's just coerce-away anyways. We already have this kind of instances for Eq, Ord, Show, and I'd like to extend to other standard type classes as well.

[Bodigrim]

Are there other Data.Functor.* types that are missing instances?

Other newtypes under Data.Functor (which are Identity and Const) already have full set of instances.

---

I've raised a draft MR with the proposed changes: https://gitlab.haskell.org/ghc/ghc/-/merge_requests/10419

[Bodigrim]

Dear CLC members, let's vote on the proposal to add instance {Enum, Bounded, Num, Real, Integral} (Compose f g a) as detailed in https://gitlab.haskell.org/ghc/ghc/-/merge_requests/10419/diffs.

@tomjaguarpaw @chshersh @hasufell @angerman @parsonsmatt @mixphix

---

+1 from me, obviously.

[tomjaguarpaw]

+1

[hasufell]

+1

[angerman]

We do not have an impact assessment for this, do we? Primarily for completeness. Every change should have an impact assessment.

As I believe this is purely additive and a fail to see how this can break existing code (someone enlighten me please, if I'm wrong). +1.

[mixphix]

+1

[chshersh]

+1

[Bodigrim]

@angerman yes, this is a purely additive change.

Thanks all, approved.