Am Donnerstag 24 Dezember 2009 22:21:50 schrieb michael rice:
Hmm... here are the functions I was looking to trace, the second one being an example from Scheme text "Concrete Abstractions" that I rewrote after seeing the first. Compared to the CL/Scheme memoization code, the Haskell seems like, how shall I put this, a sleight of hand, so much so that I'm driven to look behind the scenes to try to understand what is occurring. I remember that someone said, pattern matching is strict and LET is lazy, so I know the trick depends on laziness, but knowing that and understanding it are still a world apart.
Does tracing a function *always* require memoizing it?
No, of course not. That was just a use of tracing I knew where it was. You can put calls to trace more or less everywhere, like mySine x = sin x `debug` "Calling sine " ++ show x (note: using debug = flip trace instead of trace needs fewer parentheses and makes commenting out the tracing code easier). Of course, tracing can be particularly enlightening for recursive functions.