sdowney:
ok, i'll bite. why should i prefer join rather than concat in the list monad. and, moreover, why is this a lambdabot trick?
Oh, but you shouldn't prefer it! It's obfuscating! It's a trick simply because lambdabot knows the rewriting: 12:16 dons> ?pl reverseWords = concat . reverse . groupBy ((==) `on` isSpace) 12:16 lambdabot> reverseWords = join . reverse . groupBy ((==) `on` isSpace) Refactoring autobots to the resuce!
i suspect that the answer actually has a deep connection to the 'dummies' thread next door. while any program that produces the right result is correct, there are some that are more correct than others. a good introductory guide to haskell should lead you in that direction.
reasoning by analogy, for c++, andy koenig's accelerated c++ had a huge impact on teaching c++ because it didn't start with C or pidgen c++, but instead required the student to accept that many mechanisms would be explained later, but that they should be used -now-.
i suspect that being biased towards higher order functions, and writing new functions such that monads (although the more i learn i think that should be typeclasses) can take advantage of those functions, is the skill that needs to be learned / taught. the equivalent of the central dogma of OO, where it's all about objects having identity, state , and behavior.
Yes, and I think a proper understanding of higher order functions, parametric and bounded polymorphism and monadic computations is sorely lacking in a lot of undergraduate courses. With monads in C#, and parametric polymorphism in Java, I wonder how long it will be before you're expected to understand these concepts to program real world Java/C#/... -- Don