Actually, you can make a joinInner for the State monad. However, it does not allow the inner function (h) to change the state, because how state is threaded through a monad N is different for each N.

i :: (Monad n) => (a -> State s b) -> (b -> n c) -> (c -> State s d) -> (a -> State s (n d))
i f g h = joinInnerState . liftM (liftM h . g) . f

joinInnerState :: Monad n => State s (n (State s a)) -> State s (n a)
joinInnerState (State g) = State $ joinInnerAsReader . g
  where
    joinInnerAsReader (n, s) = (liftM (fst . ($ s) . runState) n, s)

joinInner is the only one of the 3 that works, because the outer M gives you an initial state to work with.

Sjoerd

On Jul 10, 2009, at 11:25 PM, Job Vranish wrote:

Yeah, I think the problem with my case is that while M is a specific monad (essentially StateT), N can be an arbitrary monad, which I think destroys my changes of making a valid joinInner/joinOuter/distribute.
Maybe someday Haskell will infer valid joinInner/joinOuter for simple cases :D
Thanks for you help. I'll definitely have to see if I can find that paper.

- Job Vranish

On Fri, Jul 10, 2009 at 3:09 PM, Edward Kmett <ekmett@gmail.com> wrote:
The problem you have is that monad composition isn't defined in general. You would need some form of distributive law either for your monads in general, or for your particular monads wrapped around this particular kind of value.
 
What I would look for is a function of the form of one of:
 
distribute :: N (M a) -> M (N a)
joinInner :: M (N (M a)) -> M (N a)
joinOuter :: N (M (N a)) -> M (N a)

that holds for your partiular monads M and N.
 
IIRC Mark P. Jones wrote a paper or a lib back around '93 that used these forms of distributive laws to derive monads from the composition of a monad and a pointed endofunctor.
 
-Edward Kmett
 
On Fri, Jul 10, 2009 at 11:34 AM, Job Vranish <jvranish@gmail.com> wrote:
I'm trying to make a function that uses another monadic function inside a preexisting monad, and I'm having trouble.
Basically my problem boils down to this. I have three monadic functions with the following types:
f :: A -> M B
g :: B -> N C
h :: C -> M D
(M and N are in the monad class)
I want a function i where
i :: A -> M (N D)

the best I can come up with is:
i :: A -> M (N (M D))
i a = liftM (liftM h) =<< (return . g) (f a)

I'm starting to feel pretty sure that what I'm going for is impossible. Is this the case?
_______________________________________________
Haskell-Cafe mailing list
Haskell-Cafe@haskell.org
http://www.haskell.org/mailman/listinfo/haskell-cafe



_______________________________________________
Haskell-Cafe mailing list
Haskell-Cafe@haskell.org
http://www.haskell.org/mailman/listinfo/haskell-cafe

--
Sjoerd Visscher
sjoerd@w3future.com