
If M is a monad transformer (e.g. StateT) and able to be parameterised over an arbitrary monad, you can do something related, which may actually be more useful to you. Given the definition of StateT from mtl: newtype StateT s m a = StateT { runStateT :: s -> m (a, s) } you could define this join-like function: join2 :: (Monad n) => n (StateT s n a) -> StateT s n a join2 m = StateT (\state -> m >>= flip runStateT state) (I don't know a good name for this function; it's called "joinIM" in iteratee) Perhaps this would apply to your situation? Cheers, John Lato
Message: 18 Date: Fri, 10 Jul 2009 17:25:55 -0400 From: Job Vranish
Subject: Re: [Haskell-cafe] Problems with nested Monads To: Edward Kmett Cc: Haskell Cafe mailing list Message-ID: Content-Type: text/plain; charset="iso-8859-1" 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
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
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