Question

As a learning exercise, I’m trying to implement a heapsort in Haskell. I figured the State monad would be the right choice to do this, since heaps rely pretty heavily on moving data around inside a single structure (and do notation would be useful). Besides, I’m looking to cement my understanding of monads in general.

The examples on the State monad in Learn You A Haskell (and a number of other tutorials), say that State is defined as:

newtype State s a = State { runState :: s -> (a,s) }

I should be passing a function of type s -> (a,s) (which may or may not be curried in other arguments) to the State value constructor. So my functions look like this:

pop :: Ord a => State (Heap a) a
pop = State pop'
pop' :: Ord a => Heap a -> (a, Heap a)
-- implementation of pop' goes here

push :: Ord a => a -> State (Heap a) ()
push item = State $ push' item
push' :: Ord a => a -> Heap a -> ((), Heap a)
-- implementation of push' goes here

This doesn’t compile, with the following error:

Not in scope: data constructor `State'
Perhaps you meant `StateT' (imported from Control.Monad.State)

From reading the API docs for Control.Monad.State, it looks as though the State value constructor has been removed from the module since these tutorials were written. As a beginner, I find the documentation to be far from self-explanatory. So my question is:

1. Am I right in believing that the State value constructor is gone?
2. What should I be using instead?

Answer 1

1. Yes, it’s gone, replaced by StateT. (The State monad is now defined in terms of the StateT monad transformer.)
2. You should be using the state function instead.

I would question whether your approach is correct, however. Instead of worrying about how State is implemented, consider using do-notation and the get and put functions.