> Just to note that one practical consequence of this change is that
> if you really need a PrimPatch constraint but omit it, instead
> of the type-checker telling you that, you get a stream of messages
> about the individual classes. You have to be paying a bit of
> attention to realise that the right fix is to add PrimPatch rather
> than writing out a huge constraint by hand.
I noticed that, too.
My conclusion from this and the problem with using class constraints on
type family applications is that we should think carefully before we
apply this idiom. As I see it, for RepoPatch the benefits outweigh these
problems, but for PrimPatch I'm not so sure.
|