[Axiom-developer] Re: concatenate and bit-vector

[Stephen Wilson]

Gabriel Dos Reis <address@hidden> writes:

> On Thu, 9 Aug 2007, address@hidden wrote:
> 
> | Steve, Gaby,
> | 
> | The semantics of concat are those of vmlisp.
> | 
> | It might be possible to prove that concat is never called with
> | bit-vector semantics. The subtle part is to prove that it can't
> | occur in generated code (e.g. algebra).
> 
> I'm sorry but that is nonsense to me.  
> 
> You clearly have a buggy codes, but you're unwilling to fix, invoking twisted
> logic.  
> 
> That concat has two version.  One for AKCL, one for non-AKCL.
> The only way its call can always yield meaningful results is that the
> arguments satisfy the common semantics restrictions, which is that
> the arguments be of type string.  That is all you need to know to fix your
> code. 

I agree that 'concat is a bit wierd, an I would like it to be a little
more general, but I still dont understand what is broken about it.

One gets the same results for each version of concat, under GCl or
SBCL as appropriate.

  (concat "foo" #\-  'bar) ==> "FOO-BAR"

  (concat #*1010 (make-array 2 :element-type 'bit :initial-contents '(0 1)))
     ==> #*101001

I do not understand what is buggy about the code per say, other than
the semantics just dont jive well IMHO in CL terms w.r.t 'concatenate
in the case the arguments are string designators thus possibly leading
to confusion.

Am I missing something?

Steve