Re: Scheme is too complicated

[Wolfgang Hukriede <whukriede@ifm.uni-kiel.de>]

Maciej Stachowiak <mstachow@mit.edu> writes:

>  jglascoe@jay.giss.nasa.gov writes:
>  > On Fri, 30 Oct 1998, Maciej Stachowiak wrote:
>  > 
>  > > (define (hash->keys mytab)
>  > >   (let ((real-tab (cdr mytab))
>  > >   (entry->key cadr)
>  > >   (end (vector-length real-tab)))
>  > >     (let loop-over-tab ((index 0)
>  > >           (accum ()))
>  > >       (if (= index end)
>  > >     accum
>  > >     (loop-over-tab (+ index 1)
>  > >          (let loop-over-bucket ((l (vector-ref real-tab index))
>  > >                  (accum accum))
>  > >            (if (null? l)
>  > >                accum
>  > >                (loop-over-bucket (cdr l)
>  > >                   (cons (car l) accum)))))))))
>  > 
>  > hrmm.  functional, pretty, and sleek.  But I bet this guy is faster:
>  > 
>  > > (define (hash->keys mytab)
>  > >   (let ((real-tab (cdr mytab))
>  > >   (entry->key cadr)
>  > >   (end (vector-length real-tab)))
>  > >     (do ((index 0 (+ 1 index))
>  > >      (accum () (do ((l (vector-ref real-tab index) (cdr l))
>  > >           (accum accum (cons (car l) accum)))
>  > >               ((null? l))
>  > >             accum)))
>  > >     ((= index end))
>  > >       accum)))

...

Except that I didn't follow the thread, and that the first let
probably should be a let*, and that I observe that "entry->key" is
defined but never used in both versions ..

.. and while you claim

>  Incidentally, you mentioned order of magnitude improvements of length
>  of Perl/Python source code vs. C, I think Scheme often gives
>  improvements on the same scale.

isn't that whole bunch not just:

(define (hash->keys mytab)
   (map cadr (apply append (vector->list (cdr mytab)))))  

Huh?  [Example:

(define eghash (cons 'eghash (list->vector
  '(()
    ()
    ((v1 k1) (v2 k2))
    ()
    ((v3 k3))
    ()))))

> (hash->keys eghash) ==> (k1 k2 k3) ]

-Wolfgang.