Re: [Chicken-users] tinyclos and variable argument functions

[John Lenz]

On 04/06/05 01:47:57, felix winkelmann wrote:
> >
> > I also think there might be a problem with define-macro... With the
> above patch to tinyclos,
> > the following code gives somewhat strange results:
> >
> > #;2> (define-method (foo (a <top>) b) (print "two " a " " b))
> > #;3> (define-method (foo (a <top>)) (print "one " a))
> > #;4> (foo 3 2)
> > Error: bad argument count - received 3 but expected 2
> > #;4> (foo 3)
> > one 3
> > "one "
> > #;5> (define-method (foo (a <top>) (b <top>)) (print "two other " a " "
> b))
> > #;6> (foo 3 2)
> > two other 3 2
> > "two other "
> >
> > What is going on here is that when expanding the first define-method,
> define-method is
> > passing the list of specializers to (add-global-method) as '(<top>).
> This is because
> > of lines 482 in chicken-highlevel-macros.scm:
> >     (if (or (not (pair? args))
> >             (and (not (pair? (car args)))
> >                  (not (scan (cdr args))) ) )
> > Since define-method stops generating the list of specializers when it
> finds no more
> > pairs left in the list of arguments, b's type (which is <top>) is never
> getting added
> > to the list. It makes for a wierd case where wrapping b inside a (b
> <top>) works
> > whereas it doesn't if b is just left open.
> >
> > Thus, during add-method, since the list of specializers for the first
> method is equal
> > to the list for the second, the second define-method is replacing the
> first.
> >
>
> But is this so weird? A non-specialized argument ("b" in the first
> define-method) must be an instance of something in the end...

No, it doesn't.  The way define-method is written, in the first define  
method the list of specializers is (<top>).  That is, define method passes  
a list of specializers that is one shorter than the actual argument lists.   
Try something like the following:

(define (printspecs var sym specs function)
  (print specs))

(define <top> 'top)

(set! ##tinyclos#add-global-method printspecs)

(define-method (foo (a <top>) b) (print "hey"))
(define-method (foo (a <top>) (b <top>)) (print "yo"))
(define-method (foo (a <top>) b (c <top>)) (print "2"))

You can see that the first one prints
(top)
and the second prints
(top top)

The reason is because define-method stops processing the list of arguments  
when it checks and sees there are no more pairs left in the list.  (See the  
scan function).

John