[plt-scheme] why isn't DrScheme odd enough?

On Sep 19, Shriram Krishnamurthi wrote:
> According to the Wadler/Taha/MacQueen paper on even and odd lazy
> lists, the following program should produce an error:
> 
> (define (countdown n)
>   (cons n (delay (countdown (- n 1)))))
> 
> (define (dmap f l)
>   (if (null? l)
>       l
>       (cons (f (car l))
>             (delay (dmap f (force (cdr l)))))))
> 
> (define (cutoff n l)
>   (cond
>     ((= n 0) '())
>     ((null? l) '())
>     (else (cons (car l)
>                 (cutoff (- n 1) (force (cdr l)))))))
> 
> (cutoff 5 (dmap sqrt (countdown 4)))
> 
> In fact, it does no such thing.  I wondered whether DrScheme had
> adopted "even" laziness somewhere along the way, so I switched to R5RS
> Scheme (hence the lack of brackets, etc., above) and ditto.
> 
> Anyone spot what I'm missing?

Yes, (sqrt -1) is not an error in mzscheme...  This is the surprise
they're describing:

  > (cutoff 4 (dmap values (countdown 4)))
  (4 3 2 1)
  > (cutoff 4 (dmap / (countdown 4)))
  /: division by zero

-- 
          ((lambda (x) (x x)) (lambda (x) (x x)))          Eli Barzilay:
                  http://www.barzilay.org/                 Maze is Life!