On Mon, Jul 27, 2009 at 7:58 PM, Dan Weston wrote:

The following inferred type has a constraint that can be trivially satisfied, but isn't:

Control.Monad> :t \ (m,f,x) -> (m >>= f) x \ (m,f,x) -> (m >>= f) x  :: forall t a b. (Monad ((->) t)) => (t -> a, a -> t -> b, t) -> b

-- In Control.Monad there is forall t. instance Monad ((->) t), -- so why is the vacuous Monad constraint still there? -- Nor can I remove it with a type annotation:

That instance is in the annoying module Control.Monad.Instances.

Control.Monad> :t \ (m,f,x) -> (m >>= f) x :: forall t a b. (t -> a, a -> t -> b, t) -> b

<interactive>:1:13:    Inferred type is less polymorphic than expected    [snip]

-- Then if I just import a module: Control.Monad> :m + Control.Arrow

Presumably it is also here.

-- Now, the Monad ((-> t) constraint disappears:

Control.Monad Control.Arrow> :t \ (m,f,x) -> (m >>= f) x \ (m,f,x) -> (m >>= f) x  :: forall t a b. (t -> a, a -> t -> b, t) -> b

-- Although it still will not accept an explicit type annotation: Control.Monad Control.Arrow> :t \ (m,f,x) -> (m >>= f) x :: forall t a b. (t -> a, a -> t -> b, t) -> b

This is just because :: binds tighter than lambda, I think:

...

...

:t (\(m,f,x) -> (m >>= f) x) :: forall t a b. (t -> a, a -> t -> b, t) -> b

typechecks just fine. Luke