Re: Continuations and scheme level wrappers

[Mikael Djurfeldt]

Gary Houston <address@hidden> writes:

> > From: Mikael Djurfeldt <address@hidden>
> > Date: Thu, 30 Nov 2000 18:01:46 +0100
> > 
> > I saw your recent change maing continuations applicable smobs.
> > 
> > Question: Will the new continuations be properly tail-recursive?
> > 
> > They need to be...
> > 
> > They should, for example, handle a situation where a recursion
> > repeatedly creates a continuation and calls it tail-recursively
> > without leaving anything on the stack.
> 
> Hi Mikael,
> 
> Do you mean something like this?
> 
> (define cont #f)
> 
> (define (a)
>   (call-with-current-continuation (lambda (c)
>                                   (set! cont c)))
>   (b))
> 
> (define (b)
>   (if (< (random 10) 5)
>       (cont #f)
>       (a)))

Yes,

  (define (a)
    (call-with-current-continuation (lambda (c) (c #f)))
    (a))

should not grow the process, but nor should

  (define (b)
    (call-with-current-continuation (lambda (c) (b))))

> Judging by process size, this seems to be tail recursive when using
> the normal evaluator

Yes, I now had a quick look at your changes, and they look good to me.
I reacted to the "avoid Scheme wrapper" comment in the ChangeLog and
thought you meant the wrapper for call/cc in r4rs.scm.

Actually, now you've partially factored out continuations from the
evaluator so that Guile becomes slightly more modular.  That is very
good.

> but not when using the debugging evaluator.

Hmm... this is a bug which should be fixed.

Best regards,
Mikael