Jason Catena wrote:
By this example State doesn't seem to give you anything more than a closure would, since it doesn't act like much of an accumulator (by, for example, storing 6 as its new internal value).
Reader r = (r->) Writer m = Monoid m => (m,) State s = (s->) . (s,) So, yes, the state monad can "change the internal value".
Could you use State for something like storing the latest two values of a Fibonacci series?
Sure. let s = (Int,Int) and fibs = start where start 0 = 0 start 1 = 1 start n = evalState (recurse $! n-2) (0,1) recurse n = do (x,y) <- get let z = x+y z `seq` if n==0 then return z else do put (y,z) recurse $! n-1 will return the nth Fibonacci number, with memoization, while only holding onto the previous two memos.
Would Haskell memoize already-generated values in either case?
No. Doing so would lead to enormous memory leaks.
Could we write a general memoizer across both the recursive and State implementations, or must we write a specific one to each case?
There are a number of generic memoization libraries on Hackage, just take a look. -- Live well, ~wren