Max et. al, I'm impressed by your solutions to the problem. I still object to them in practice, though, because they depend on either extra up-front work (uncurry-n) or extra understanding on the part of the caller (:end, '() and calling without parameters).
Haskell and the other ML languages are able to support partial application for two reasons: they have a simpler notion of function, and they have a type-checking compiler to catch human errors. Hopefully kinghajj can forgive me for not knowing about Printf; in the two years I spent studying Haskell I never saw a user-defined variadic function nor in fact a call to the Printf library. I am under the impression the OCaml compiler does extra tricks just to support printf but even if it doesn't you won't find many variadic functions in OCaml either. One thing I like about Clojure's #() syntax is that I can tell by looking at the code that what's being passed back is a function. In Haskell, the only way one knows whether you've got a function where you meant a value is through the compiler. Clojure's compiler isn't that sophisticated and hopefully never will be, it being a Lisp family language. This means some of the knowledge has to be either in the head of the user, the editor, or the code itself. Even knowing Haskell well, you will occasionally write "sum = foldr (+)" instead of "sum = foldr1 (+)" or "sum = foldr (+) 0". You survive because the compiler tells you you've screwed up. Haskell's nature is to prefer a simple, mathematically complete facility over two more human-friendly facilities, so it has a simplified notion of function. Clojure's notion of function is more amenable to (some) humans and it would require extensive overhauling to support syntactic partial application cleanly. It wouldn't surprise me if it would need a Hindley-Milner type checker just to produce useful errors. To see the downside, notice that Haskell has map, zipWith, zipWith2, zipWith3, zipWith4, ... zipWith8 in the standard library. If you want to apply a function with 9 arguments to each value successively in 9 lists, you have to write your own function to do it. In Clojure, all of these are just map. Upsides and downsides. With respect to eyeris, while there may not be a difference to the compiler between #(* % 2) and (fn [x] (* x 2)), there is a large difference to humans and not knowing when (* 2) is a function or a value would be sufficiently detrimental (IMO) to outweigh the benefits. Again with respect, because I think Haskell get great utility from this facility, but only because it exists in a Haskell universe with Haskell functions, Haskell error messages and Haskell typechecking. I just don't see how one could import this particular feature without importing a lot more of Haskell, destroying a lot of the value of Clojure in the process. And the utility of (* 2) over #(* 2 %) seems pretty limited, and (partial * 2) doesn't seem that bad. Perhaps there could be another reader macro, #p(* 2) or something, being the same as (partial * 2)? I would have to be more familiar with other people's code to know whether that would be generally useful. Incidentally, have any of you taken a look at Liskell? <http://liskell.org/ > — Daniel Lyons http://www.storytotell.org -- Tell It! --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en -~----------~----~----~----~------~----~------~--~---
