Semigroup constraint for the class Action

[AliceRixte]

I started to use monoid-extras, but one detail that stopped me : the Action class doesn't inherit of the semigroup structure of it's first argument. Is there a reason why the declaration of Action isn't

class Semigroup m => Action m s where

instead of the current

class Action m s where

I propose here a quick solution, which might break some code depending on Action. It seems to me that the necessary changes to the Product and Sum instances make much more sense like this , since it should be elements of the monoid Sum (or Product) that should act on their underlying set, and not the other way around.

Cheers !

[byorgey]

Thanks for the PR! I will try to take a look at this soon.

[byorgey]

This does make a lot of sense to me. I think probably the reason the Semigroup constraint was not there is that this library was initially created long ago, at a time when Semigroup was not a superclass of Monoid, so having a Semigroup constraint would have excluded types which were an instance of Monoid but not Semigroup. However, Semigroup has been a superclass of Monoid since GHC 8.4 (https://gitlab.haskell.org/ghc/ghc/-/wikis/proposal/semigroup-monoid) and we currently don't support any versions older than that, so we should be fine!

This does break the dual-tree package though it looks like for a simple and easily solved reason. I want to first make sure I can fix that and there are no other issues in the diagrams ecosystem; assuming that all works (as seems likely) we can then merge and release this.

[AliceRixte]

That's great ! Don't hesitate if you need anything else to make this work :-)