makeSem doesn't work well with effects with "pure type" arguments

[googleson78]

With a declaration like this:

-- intended to always be used with a type application
data Eff m a where
  Stuff :: forall ty. Int -> Eff m (F ty)

type family F ty where
  ...

makeSem generates non-compiling declarations.
I'm guessing it generates something like this:

stuff :: forall ty r. Member Eff r => Int -> Sem r (F ty)
stuff x = send $ Stuff x

but ghc can't figure out that the ty from the forall is the one I want to use for ty in the Stuff invocation.

Instead we need something like this:

stuff :: forall ty r. Member Eff r => Int -> Sem r (F ty)
stuff x = send $ Stuff @ty x

but who knows if this breaks something else :/

[bolt12]

You can always default to manual writing the boilerplate, right?

…

On Fri, 1 May 2020, 12:26 Georgi Lyubenov, ***@***.***> wrote: With a declaration like this: -- intended to always be used with a type applicationdata Eff m a where Stuff :: forall ty. Int -> Eff m (F ty) type family F ty where ... makeSem generates non-compiling declarations. I'm guessing it generates something like this: stuff :: forall ty r. Member Eff r => Int -> Sem r (F ty) stuff x = send $ Stuff x but ghc can't figure out that the ty from Stuff invocation is the same as the ty in the forall. Instead we need something like this: stuff :: forall ty r. Member Eff r => Int -> Sem r (F ty) stuff x = send $ Stuff @ty x but who knows if this breaks something else :/ — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#339>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AE2SULI3X3OQCQ2WXHKFUMDRPKWW5ANCNFSM4MXBFRCQ> .

[googleson78]

@bolt12 Yes, that's what I'm doing for the time being