How do define a monad transformer that is the composition of two arbitrary monad transformers?-CodePudding

I want to write something similar to the following:

newtype FooT c d m a = FooT { unFooT :: (c (d m)) a }

instance (MonadTrans c, MonadTrans d) => MonadTrans (FooT c d) where
  lift = FooT . lift . lift

However, this snippet will not compile:

Could not deduce (Monad (d m)) arising from a use of ‘lift’

I understand why this won't compile; we don't know that the application of an arbitrary transformer d m is itself a monad. However, I'm not sure of the best way to proceed. Is there a clean way to make something like this work? Presumably it would go through if I could add a constraint along the lines of Monad (d m) to the left-hand-side of the instance declaration, but I don't know how to do so since m is not bound.

CodePudding user response：

With the QuantifiedConstraints GHC extension, this is

{-# LANGUAGE QuantifiedConstraints #-}

instance (MonadTrans c, MonadTrans d, forall m. Monad m => Monad (d m)) =>
         MonadTrans (FooT c d) where
  lift = FooT . lift . lift

m in the constraint is not the same m as in lift. The quantified constraint simply means what it says ("for any m :: Type -> Type, if Monad m require Monad (d m)"), and in lift that universal statement is being instantiated with the particular m being passed as argument to lift. Thus lift's m does not escape its scope.