2009/9/25 Casey Hawthorne :

Well that makes sense, but for a learner, how is he/she supposed to know that 'i' could be '(i,i)' or for that matter a tuple of n of those i's?

"STUArray s i e"

Could you also have a tuple of states?

Obviosly, 'e' could be a tuple, for instance (Int,Char)

Well, 'i' is just a type variable, just like 'a' and 'b' are type variables in the function type map :: (a -> b) -> [a] -> [b] I guess you know that you can "map" a suitable function on a list of pairs, right ? Type variable mean you can use whatever type you want, including pairs, lists, or whatever. So you can have list of whaterver you want, and use whatever you want as indices in arrays.... BUT! But the whatever you want can be constrained a bit: here, the constraint is that the type of the indices must be in Ix, this is what the "Ix i =>" means. For instance, we said we can put whatever you want in a list. But ask GHCi what is the type of inc = map (+1) : Prelude> :t map (+1) map (+1) :: (Num a) => [a] -> [a] You see that you can use "inc" on list of whatever you want (the "a") *provided* the "a" is in Num, the "Num a =>" part of the type signature. Now, you have to look if the type you want to use for your indices is in Ix. Look at [1] and you see that (Ix a, Ix b) => Ix ((,) a b) is an Instance of Ix. (The right part can be read as (a,b) instead of (,) a b). So a pair is in Ix provided its elements are in Ix too. [1] http://hackage.haskell.org/packages/archive/base/4.0.0.0/doc/html/GHC-Arr.ht...

Am Freitag 25 September 2009 09:22:25 schrieb Casey Hawthorne:

Well that makes sense, but for a learner, how is he/she supposed to know that 'i' could be '(i,i)' or for that matter a tuple of n of those i's?

minh thu already explained this very well.

"STUArray s i e"

Could you also have a tuple of states?

You can't choose the state 's', the documentation says "The strict state-transformer monad. A computation of type ST s a transforms an internal state indexed by s, and returns a value of type a. The s parameter is either * an uninstantiated type variable (inside invocations of runST), or * RealWorld (inside invocations of Control.Monad.ST.stToIO). " Without evil hackery (or stToIO), you can only use ST actions/ST(U)Arrays via runST :: (forall s. ST s a) -> a or runST(U)Array :: Ix i => (forall s. ST s (ST(U)Array s i e)) -> (U)Array i e which have rank 2 types (universally qualified type as type of argument [result]), the 'forall s' within the parentheses says it has to work whatever type the rts chooses (actually none), so if you write myFancyArray :: forall s1, s2. ST (s1,s2) (STUArray (s1,s2) Int Int) you can't use it.

Obviosly, 'e' could be a tuple, for instance (Int,Char)

Not for STUArrays, but for STArrays, there's no problem.

-- Regards, Casey