Prelude: I don't know if it's worth trying to do this, or if the more precise arity analysis is better than theoretical allocation improvements.
We can legally define
instance MonadTrans Codensity where
lift = frob
frob :: forall m a . Monad m => m a -> Codensity m a
frob = coerce (blah (Mag (>>=)) :: m a -> Mag2 m a)
newtype Mag m a = Mag (forall b . m a -> (a -> m b) -> m b)
newtype Mag2 m a = Mag2 (forall b . (a -> m b) -> m b)
blah :: Mag m a -> m a -> Mag2 m a
blah (Mag p) q = Mag2 (p q)
blah is obviously the identity (up to arity), because all it does is remove a newtype constructor and put a different one on. The effect is to shift the forall b to the right of an arrow; this does not change the run-time representation because type lambdas compile away. Unfortunately, I don't see a way to convince GHC that blah is just a cast. I believe, however, that it should be safe to define blah = unsafeCoerce id.
Prelude: I don't know if it's worth trying to do this, or if the more precise arity analysis is better than theoretical allocation improvements.
We can legally define
blahis obviously the identity (up to arity), because all it does is remove a newtype constructor and put a different one on. The effect is to shift theforall bto the right of an arrow; this does not change the run-time representation because type lambdas compile away. Unfortunately, I don't see a way to convince GHC thatblahis just a cast. I believe, however, that it should be safe to defineblah = unsafeCoerce id.