From: Yi DAI <plm....@gmail.com> Subject: About the primitive macros `and' and `or' Date: Fri, 25 Dec 2009 13:18:01 +0800
> Hi, > > > I don't see the point why Scheme provides the general `and' and `or' as > primitive macros (which does stand in our way when we wanna (apply and > things) instead of primitive procedures. For efficiency? I don't think there > would be much compared to the following definitions: > > (define (gand . l) ; 'g'eneral 'and' > (define (iand l) > (if (null? l) > #t > (let ((b (car l)) > (bs (cdr l))) > (cond ((null? bs) b) > (b (iand bs)) > (else #f))))) > (iand l)) > > (define (gor . l) ; 'g'eneral 'or' > (define (ior l) > (if (null? l) > #f > (let ((b (car l)) > (bs (cdr l))) > (cond ((null? bs) b) > (b b) > (else (ior bs)))))) > (ior l)) As a function, gor needs to have its arguments evaluated before the actual application takes place, moreover, the order of evaluation is not specified. Hence (define x 0) (gor (= x 0) (/ x)) gives a numerical overflow error, whereas (define x 0) (or (= x 0) (/ x)) returns true. If the terms to be evaluated contain non-commutable side effects, things get out of control, whereas in the macro `or', you may rely on the order of the side effects. Klaus Schilling
