Incorrect gshow for infix constructor with zero or one argument

[Lysxia]

> data U = (:+) Int deriving Generic
> instance GShow U
> gshow ((:+) 0)
:+ 0

Expected output:

(:+) 0

[RyanGlScott]

OK... and why do you expect that output? :)

I think perhaps you're trying to silently appeal to how a derived Show instance would behave in the same scenario, but keep in mind that GShow is not the same class as Show. Indeed, we make no guarantees about output of GShow at all—it, like every other G-prefixed class in this library, is simply a toy example to demonstrate how one can use GHC.Generics. GShow is not intended to be a drop-in replacement for Show (and indeed, it can't be by virtue of being a different class).

To put it another way—we could change this behavior, but I don't see what it would buy us here. I think what you want is a GHC.Generics–based implementation of show, not the GShow class.

[Lysxia]

Since generic-deriving is a toy example, I understand that this isn't really a problem, so feel free to just close this.

In fact, I was considering forking generic-deriving to work with the actual classes in base. Such a fork would have more reason to match GHC's derived instances. Eventually I changed my mind, but I thought this discrepancy would at least be worth pointing out, as I expected gshow to mirror show and produce a valid Haskell expression, at least for some simple class of types.

[RyanGlScott]

Indeed, I think having a library which provides generic implementations of standard type classes would be a good idea. For various reasons, generic-deriving probably won't ever be that library, one reason being that generic-deriving has an extremely minimal dependency footprint, so backporting generic implementations of things like Semigroup would be tricky, since Semigroup has only been in base since 4.9.

In any case, I'll opt to close this. Thank you for the bug report!