On Fri, Oct 28, 2011 at 2:04 PM, Joe Marshall <jmarsh...@alum.mit.edu> wrote: > On Fri, Oct 28, 2011 at 10:58 AM, Stephen Bloch <sbl...@adelphi.edu> wrote: >> >> OK, here's a variant of "binary->nary" that produces the results we want >> when "op" has contract "X X -> boolean" rather than "X X -> X" >> >> (define (binary->nary relop) >> (letrec ((f (lambda args >> (or (empty? args) >> (empty? (rest args)) >> (and (relop (first args) (second args)) >> (apply f (rest args))))))) >> f)) >> >> This applies relop to each successive (overlapping) pair of elements and >> "and"s the results. > > Yes! Thank you! > > Of course this fails for n-ary add and multiply, but presumably a > person who preferred > this generalization would have to answer the dual question of the > original poster: > Why do the relational operators generalize to 0 or 1 arguments but addition > and > multiplication do not?
You seem to be assuming that we have to pick one binary->nary for all binary operators. I would choose this one for relations and the other one for associative operators with identities. Other kinds of operators may have other kinds of natural generalizations. --Carl _________________________________________________ For list-related administrative tasks: http://lists.racket-lang.org/listinfo/users
