[Haskell-cafe] Re: Can we come out of a monad?

[Anton van Straaten]

Kevin Jardine wrote:
> As a Haskell newbie, the first thing I learned about monads is that
> they have a type signature that creates a kind of mud you can't wash
> off.

There are places where you can't wash it off, and places where you can.

> eg.
> 
> f :: String -> MyMonad String
> 
> By mentioning the monad, you get to use its special functions but as a
> hard price, you must return a value with a type signature that locks
> it within the monad

That's perfectly correct: "you must return a value with a type signature 
that locks it within the monad."  That's because you're referring here 
to returning a value from a monadic function with a return type of 
MyMonad String.  But that's just one part of the picture.

Consider a caller of that function: after applying f to some string, it 
ends up with a value of type MyMonad String.  One of the things you can 
typically do with such values is "wash off the mud" using a runner 
function, specific to the monad.

They're called runners (informally) because what they do is run the 
delayed computation represented by the monad.  In the case of the State 
monad, the runner takes an initial state and supplies it to the monad in 
order to start the computation.  If these runners didn't exist, the 
monad would be rather useless, because it would never actually execute. 
  The result of running that computation typically eliminates the monad 
type - the mud is washed off.

You can even do this inside a monadic function, e.g.:

g m = do s <- get
          let x = evalState m s   -- wash the mud off m !
          ...

But the value of x above will be locked inside the function - you can't 
return such values to the caller without using e.g. "return x", to 
return a monadic value.

So you may be able to wash the mud off a monadic value, but if you want 
to pass that value outside a monadic function you have to put the mud 
back on first.

However, if you have a monadic value *outside* a monadic function, no 
such rule applies.

> The more I learn about monads, however, the less I understand them.
> I've seen plenty of comments suggesting that monads are easy to
> understand, but for me they are not.

Monads are very general, which means they're not easily learned by the 
common style of extrapolating from examples.  They're easy to understand 
in hindsight though!  :-}

Anton