Update to polysemy's new names

[Avi-D-coder]

This does not compile for me even with polysemy master. See This actually compiles.

3 of 4 makeSem_ produce errors like the following.

polysemy-zoo/src/Polysemy/Shift/Internal.hs:23:1: error:
    • Could not deduce: s ~ a
      from the context: Member (Shift ref a) r
        bound by the type signature for:
                   reset :: forall (ref :: * -> *) a (r :: [(* -> *) -> * -> *]).
                            Member (Shift ref a) r =>
                            Sem r a -> Sem r a
        at src/Polysemy/Shift/Internal.hs:(68,1)-(71,16)
      ‘s’ is a rigid type variable bound by
        an expression type signature:
          forall s. Shift ref s (Sem r) s
        at src/Polysemy/Shift/Internal.hs:23:1-16
      ‘a’ is a rigid type variable bound by
        the type signature for:
          reset :: forall (ref :: * -> *) a (r :: [(* -> *) -> * -> *]).
                   Member (Shift ref a) r =>
                   Sem r a -> Sem r a
        at src/Polysemy/Shift/Internal.hs:(68,1)-(71,16)
      Expected type: Shift ref s (Sem r) s
        Actual type: Shift ref a (Sem r) a
    • In the first argument of ‘Polysemy.Internal.send’, namely
        ‘(Reset x_asQf :: Shift ref s (Sem r) s)’
      In the expression:
        Polysemy.Internal.send (Reset x_asQf :: Shift ref s (Sem r) s)
      In an equation for ‘reset’:
          reset x_asQf
            = Polysemy.Internal.send (Reset x_asQf :: Shift ref s (Sem r) s)
    • Relevant bindings include
        x_asQf :: Sem r a (bound at src/Polysemy/Shift/Internal.hs:23:1)
        reset :: Sem r a -> Sem r a
          (bound at src/Polysemy/Shift/Internal.hs:23:1)
   |
23 | makeSem_ ''Shift

--  While building package polysemy-zoo-0.4.0.1 using:
      /home/user/.stack/setup-exe-cache/x86_64-linux-tinfo6/Cabal-simple_mPHDZzAJ_2.4.0.1_ghc-8.6.5 --builddir=.stack-work/dist/x86_64-linux-tinfo6/Cabal-2.4.0.1 build lib:polysemy-zoo --ghc-options " -fdiagnostics-color=always"

Is the plugin active? Running stack build --ghc-options='-fplugin=Polysemy.Plugin' leads to the same error.

[KingoftheHomeless]

Grr, yeah, I've had issues using makeSem_. It is incredibly fussy about how you name the type variables. I had massaged the type signatures to the point that it all compiled at the time, but perhaps something has changed?

I'll look into it, because I really, really want to keep the uses of makeSem_. Continuation effects are weird enough that the main methods need to be documented.

[Avi-D-coder]

Are names defined in zoo going to be changed to fit the new guidelines?

How do people feel about it?

[KingoftheHomeless]

I'm all for changing interpreter names to fit the current standard. I.e. instead of runAsyncFinal, you could do asyncToFinal/asyncToIOFinal. (Although the latter's perhaps not the best name, seeing as how the semantics of runAsyncFinal is like lowerAsync, and not asyncToIO?)

I could create a pull request for changing the interpreter names for Final and friends as well as the continuation effects.

[KingoftheHomeless]

I have (presumably) fixed the uses of makeSem_ and opened a pull request over at Avi-D-coder's with-polysemy-submodule branch https://github.com/Avi-D-coder/polysemy-zoo/pull/1.

It's really weird. I figure it must be a difference between 8.6.3 (or 8.6.4, which I use locally), and 8.6.5, but I don't see anything in the release notes that would hint at the change. On the plus side, this allows me to name the variables the way I originally wanted to, so that's nice.

[isovector]

@TheMatten mind syncing with people to figure out what's going on with makeSem_ here?

[isovector]

I'd like it if the names in this module rolled with the same convention as those in the main lib. I've been putting it off. Thanks for the PR!

[TheMatten]

@KingoftheHomeless @isovector it seems like reset uses name from type constructor (a) - use name from data constructor signature instead (s)

[KingoftheHomeless]

Previously, I had to use a when type annotating reset to get it to compile, but that's no longer the case. I have no idea why.

Have you tested with 8.6.3 (lts 13.0)? I can't do that since I'm on windows, and TH is broken on windows on 8.6.3.

[TheMatten]

Well, there was a change in logic that collects info about effect GADTs - maybe you were working with version of polysemy before this PR.

[KingoftheHomeless]

That may very well be it! In that case, we have nothing to worry about when it comes to compatibility.