I attempted to create a datatype representing a tuple of infinitely many types:
{-# LANGUAGE DataKinds #-}
{-# LANGUAGE FlexibleContexts #-}
{-# LANGUAGE QuantifiedConstraints #-}
{-# LANGUAGE TypeFamilies #-}
{-# LANGUAGE TypeOperators #-}
import GHC.TypeNats
infixr 5 :!!
data OmegaTuple (t :: Nat -> *) (n :: Nat) = t n :!! OmegaTuple t (n+1)
This is fine.
I also attempted to declare the direct product of infinitely many semigroups:
instance (Semigroup (t n), Semigroup (OmegaTuple t (n+1))) => Semigroup (OmegaTuple t n) where
(x :!! xs) <> (y :!! ys) = x <> y :!! xs <> ys
Yet GHC complains like this:
• Illegal nested constraint ‘Semigroup (OmegaTuple t (n + 1))’
(Use UndecidableInstances to permit this)
If I understood correctly, using UndecidableInstances
will make GHC fall in an infinite loop.
Another attempt:
instance (Semigroup (t n), forall k. Semigroup (t k) => Semigroup (t (k+1))) => Semigroup (OmegaTuple t n) where
(x :!! xs) <> (y :!! ys) = x <> y :!! xs <> ys
Then GHC complains like this:
• Illegal type synonym family application ‘k + 1’ in instance:
Semigroup (t (k + 1))
Is it really impossible to teach GHC mathematical induction?
You can manually wrap the dictionaries, that allows lazily building up an “infinite class dictionary”:
{-# LANGUAGE GADTs #-}
{-# LANGUAGE DataKinds #-}
import Data.Kind (Type)
import GHC.TypeLits
data SemigroupSequence (t :: Nat -> Type) (n :: Nat) where
SemigroupSequence :: Semigroup (t n)
=> SemigroupSequence t (n+1) -> SemigroupSequence t n
class SemigroupFamily t where
semigroupSequence :: SemigroupSequence t 0
and then
mappendOmega :: SemigroupSequence t n
-> OmegaTuple t n -> OmegaTuple t n -> OmegaTuple t n
mappendOmega (SemigroupSequence sd') (x :!! xs) (y :!! ys)
= x <> y :!! mappendOmega sd' xs ys
instance (SemigroupFamily t) => Semigroup (OmegaTuple t 0) where
(<>) = mappendOmega semigroupSequence