On Tue, Jun 2, 2009 at 3:50 AM, Dan
You hit the nail on the head. "Why I am doing this" is because of boilerplate. Boilerplate gives me rashes and bulbous spots on the nose.
Consider the following Ruby code:
def check(zeClass, zeValue) zeValue.is_a? zeClass end
This does not require a new function for every class defined in Ruby. (To be fair, though, the class of a Ruby object tells you precious little, compared to a Haskell type constructor).
I figured there would be a clever Haskell idiom that would give me a similarly concise route. Does it really require Template Haskell? I can barely parse regular Haskell as it is..
So the question is, why do you need to know if x is an Atom or a Bool? The Haskell idiom is to pattern match and just do what you want with the data: f (Atom s) = ... f (Bool b) = ... instead of f x = if isAtom x then ... atomData x ... else ... boolData x ... Alternatively, you can define a fold[1] once: myval :: MyVal -> (Bool -> a) -> (String -> a) -> a myval (Bool b) bool atom = bool b myval (Atom s) bool atom = atom s f x = myval bool atom where bool b = ... atom s = ... This is a small amount of boilerplate that you write once for each type; it's possible to automate it with TH, but usually it's not worth it, in my opinion. Coming from Ruby (the same route I took to get to Haskell!), you should be aware that Haskell does have somewhat more "boilerplate" than Ruby, but it has its benefits as well. I am a convert to the Church of Purity and Type-Safety :) And you can use type classes for many metaprogramming tasks. -- ryan [1] "fold" here is the general term for this type of function. Examples are foldr: http://haskell.org/ghc/docs/latest/html/libraries/base/src/GHC-Base.html#fol... maybe: http://haskell.org/ghc/docs/latest/html/libraries/base/src/Data-Maybe.html#m... either: http://haskell.org/ghc/docs/latest/html/libraries/base/src/Data-Either.html#...