On Thu, 28 Jul 2005, skaller wrote:

> I'm not arguing "I don't like this approach". I already know
> of several ways to do variable length arrays, you haven't
> shown me anything new: it seems you're simply not accepting
> my assertion: it isn't possible to do it efficiently
> in Ocaml without magic: either Obj.magic or C code with
> an Ocaml interface: both solutions are fragile and
> require secret knowledge of ocaml implementation.

Plus keeping fingers crossed that major compiler changes will not break 
those implementations. (What if, at one point, GC would start making extra 
homogeneity assumptions over arrays - if it does not do so already?)

> Thus, to safely use variable length arrays they
> have to be provided by INRIA. That doesn't imply
> they *should* be provided by INRIA of course.
> I can always use a different data structure,
> ignore safety, or use another programming language.

First of all, it certainly would be nice if OCaml provided more 
flexible arrays. There are specialized applications where it really is a 
pain having to emulate them manually, and especially if it's about 
numerics, extra consing / indirection often have to be reduced to a 
minimum.

I originally was somewhat worried whether extensible arrays really would 
be such a dramatic improvement - what if the next person needs displaced 
arrays, or any other functionality from CL's arrays with their many bells, 
gongs, and whistles? But thinking a bit more about it, it seems as if
virtually anything of interest could be accomplished by other means, with 
the exception of just extensible arrays.


-- 
regards,               tf@cip.physik.uni-muenchen.de              (o_
 Thomas Fischbacher -  http://www.cip.physik.uni-muenchen.de/~tf  //\
(lambda (n) ((lambda (p q r) (p p q r)) (lambda (g x y)           V_/_
(if (= x 0) y (g g (- x 1) (* x y)))) n 1))                  (Debian GNU)