Re: [Sbcl-help] string type uncertainty?

[Todd S. <ts...@op...>]

Pascal Bourguignon <pj...@in...> writes:

> Christophe Rhodes writes:
> > [ off-list to give you a chance to reflect. ]
> 
> Thank you :-) 
> 
> > Pascal Bourguignon <pj...@in...> writes:
> > 
> > > Christophe Rhodes writes:
> > >> Pascal Bourguignon <pj...@in...> writes:
> > >> 
> > >> > The point  is that  while a (ARRAY  (ARRAY BASE-CHAR  *) *) has  to be
> > >> > implemented as  an (ARRAY T  *), it must  not forget that  it contains
> > >> > actually (ARRAY BASE-CHAR *).
> > >> 
> > >> Yes it must.  It is required to by the ANSI standard for the
> > >> language.  See the CLHS page for System Class ARRAY:
> > >> 
> > >>   If element-type is the symbol *, arrays are not excluded on the
> > >>   basis of their element type. Otherwise, only those arrays are
> > >>   included whose actual array element type is the result of upgrading
> > >>   element-type; see Section 15.1.2.1 (Array Upgrading).
> > >
> > > Beware, I've always specified the type of the elements. I've only used
> > > the * wildcard for the dimension.
> > 
> > Perhaps you failed to read the second sentence of the quote?
> 
> Yes. You're right.   That's unfortunate.  They link such  a high level
> expression  as: (ARRAY  (ARRAY BASE-CHAR  *) *)  to  an implementation
> detail  such  as UPGRADED-ARRAY-ELEMENT-TYPE.   This  is without  much
> consequence until you treat type declarations as assertions.

I'm not so sure.  The CLHS page for Declaration TYPE has this to say:

   Within the lexical scope of an array type declaration, all
   references to array elements are assumed to satisfy the expressed
   array element type (as opposed to the upgraded array element
   type). A compiler can treat the code within the scope of the array
   type declaration as if each access of an array element were
   surrounded by an appropriate the form.

The Examples section there reinforces this, except that it seems to
invoke undefined behavior by storing 127 into an (array (signed-byte
5) 1) array.  I think that's probably an errata, but maybe I'm missing
something?

-- 
Todd Sabin                                          <ts...@op...>