Consider removing all the redefined typeclasses

[fumieval]

This package has its own definitions of typeclasses (GShow, GMonoid, etc), but they are often not viable because they are incompatible with the existing instances. For example you can't derive GShow of a datatype if it contains a field that's an instance of Show but not GShow.

Same goes for everything else. Default method definitions are no longer a reason to redefine classes, now that we have DerivingVia.

[RyanGlScott]

The fact that GShow et al. aren't compatible with existing instances is by design, as these classes are primarily intended for educational purposes. That is, they exist to demonstrate how one can define GHC.Generics-based defaults for common type classes, without actually having to use the original type classes themselves. This has a number of advantages:

• Having "teaching-only" classes allows us to simplify some of the internal details. For example, having a GHC.Generics-based default for Show that accurately captures all of the intricacies of deriving Show is quite involved, but since we're using GShow, not Show, we are permitted to make some simplifications, at the expense of being slightly less faithful to the "reference implementation" of deriving Show.
• Some of the type classes that generic-deriving includes (e.g., GSemigroup) correspond to base classes (e.g., Semigroup) that were only introduced to base relatively recently. Having to support the original Semigroup in generic-deriving would complicate generic-deriving's library dependencies and GHC support window. With GSemigroup, however, we can bypass that altogether: generic-deriving can continue to have minimal dependencies and an extremely wide GHC support window. (The latter is important because people often use generic-deriving as a a way to backport GHC.Generics features to old versions of GHC.)

For these reasons, I usually recommend that people use generic-data rather than generic-deriving if they want GHC.Generics-based defaults for Semigroup, Show, etc. that use type classes from base.

Admittedly, the documentation for generic-deriving doesn't communicate this very well at the moment.

[RyanGlScott]

I've attempted to improve the documentation about this in #73.

[fumieval]

I see, thanks for explanation. I'm content as long as the purpose is documented