[Haskell-cafe] (Why) must this function remain nameless?

26 May 2015

      The following code (somewhat contrived) code compiles on GHC without issue,
and towards the end repeatedly uses the lambda function "(\fn (D x) -> f fn
x)". But when I simply try to give it a name, like so:

h = (\fn (D x) -> f fn x)

I then get a compile error.

How can I name this function, or must it remain forever nameless?

( Ideone link: http://ideone.com/mtuYnK )

---

{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE FlexibleInstances #-}
{-# LANGUAGE FlexibleContexts #-}
{-# LANGUAGE UndecidableInstances #-}
{-# LANGUAGE OverlappingInstances #-}
{-# LANGUAGE ScopedTypeVariables #-}
{-# LANGUAGE ExistentialQuantification #-}
{-# LANGUAGE MultiParamTypeClasses #-}
{-# LANGUAGE ConstraintKinds #-}
{-# LANGUAGE NoMonomorphismRestriction #-}

import GHC.Exts (Constraint)

type family Parent a

class C a b where
  f :: b -> a -> String

instance (C (Parent a) b) => C a b where
  f _ _ = f (undefined :: b) (undefined :: Parent a)

data A1 = A1
data A2 = A2
data A3 = A3
data A4 = A4

type instance Parent A2 = A1
type instance Parent A3 = A2
type instance Parent A4 = A3

data F1 = F1
data F3 = F3

instance C A1 F1 where
  f _ _ = "F1"

instance C A3 F3 where
  f _ _ = "F3"

type family Constraints t a :: Constraint

data D t = forall a. (Constraints t a) => D a

type instance Constraints A1 a = (C a F1)
type instance Constraints A2 a = (C a F1, C a F3)
type instance Constraints A3 a = (C a F1, C a F3)
type instance Constraints A4 a = (C a F1, C a F3)

main =
  do
    putStrLn (f F1 A1)
    putStrLn (f F1 A2)
    putStrLn (f F1 A3)
    putStrLn (f F1 A4)
    putStrLn (f F3 A3)
    putStrLn (f F3 A4)
    putStrLn $ (\fn (D x) -> f fn x) F1 ((D A1) :: D A1)
    putStrLn $ (\fn (D x) -> f fn x) F1 ((D A2) :: D A1)
    putStrLn $ (\fn (D x) -> f fn x) F3 ((D A3) :: D A2)

-- h = (\fn (D x) -> f fn x)

[Haskell-cafe] (Why) must this function remain nameless?

Clinton Mead