I've been toying with something similar. The approach I took was to define multi-methods for the various operators:
(defmulti add class) (defmulti sub class) (defmulti mul class) (defmulti div class) ... Then I of course write the implementations for various different types. Then, because the multi-method names don't look nice, I wrote a macro which transformed the regular operators into calls on the multi-methods, i.e.: (macroexpand-1 '(extended-math (* 3 A))) --> (mul 3 A) Its still not as clean as a having the operators defined as multi-methods from the get-go, but you don't have to muck with overriding the core names. One reason to leave the core operators alone is performance. Multi-method dispatch is far more expensive than a regular dispatch. I don't have this in front of me to get some real numbers, but when I tested, the same operations with the multi-methods were about 10x slower than the direct dispatch. /mike. On Mon, Dec 8, 2008 at 1:50 PM, ppierre <[EMAIL PROTECTED]> wrote: > > On 8 déc, 17:14, Mark Fredrickson <[EMAIL PROTECTED]> > wrote: >> >> Alternatively, could I provide a "multi- >> >> math" lib to redefine the core math functions? >> >> > Type classes would be king. >> >> Do you mean this in the Haskell sense? > Yes : http://www.haskell.org/tutorial/classes.html > Haskell make a dispatch on type at runtime. > >> > (ns test.test >> > (:refer-clojure :exclude [+ -])) >> >> > (defn + >> > [a b] >> > 33) >> >> > (+ 1 2) ;= 33 >> >> Thanks for the suggestion, > > You can "overload" core function : > (binding [clojure.core/compare (constantly 1)] > (sort [1 5 2 6])) ;(6 2 5 1) > > But it wouldn't work if the function is inline. > (binding [clojure.core/= (constantly false)] > [(= 1) > (= 1 1) > (not= 1 1) > ]) ; [false true false] > >> this will give me a chance to play around >> with multi-method and see what I think. > > I'm not sure, but : > defmethod macro simply call addMethod on clojure.lang.MultiFn object. > So you can add method from an other namespace. > > Basically, you would build something like type class. > But it wouldn't play nice with core operator. > > pierre > > > --~--~---------~--~----~------------~-------~--~----~ 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 To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/clojure?hl=en -~----------~----~----~----~------~----~------~--~---
