On Wed, Dec 31, 2003 at 11:54:27AM +0000, Graham Klyne wrote:
My *intuition* here is that the problem is with countLeaves2, in that it must build the computation for the given [sub]tree before it can start to evaluate it. Maybe this is why other responses talk about changing the state monad?
But why does this computation have be done in a state monad at all? countLeaves seems to me to be a pretty straightforward function from a Tree to an Int, with no need for intervening state other than to increment a counter: as such, I'd have expected a simple recursive function to serve the purpose. (Maybe there was something in the original application that was lost in the problem isolation?)
I think you might well be correct that I'm doing things the wrong way. The original program is a chess prog. and the function in question is the alphabeta search. I wanted to hold the transposition table (a cache of seen positions) among other things in the state monad. I thought this was the normal way to approach this, but am having doubts now. The recursive approach will indeed work, but I had hoped to avoid all the code associated with threading the state by hand. - Joe