Re: [Chicken-users] DBI

[Graham Fawcett]

On Fri, Feb 29, 2008 at 10:41 AM, felix winkelmann <address@hidden> wrote:
> On Thu, Feb 28, 2008 at 4:02 PM, Graham Fawcett
>
> <address@hidden> wrote:
>  >
>
> >  It would be a smart idea to change the implementation, then, so that
>  >  the unspecified value could not be tested with (eq?). That would
>  >  prevent it from being 'misused'.
>
>  The unspecified value is unspecified. It has both no identity and
>  no efforts are made to make it identity-less.

Right, but (eq? (void) (void)) => #t is a property that shouldn't
exist for two unspecified values: by definition it gives identity to
unspecified.

(eq? (void) (void)) => #f is better, raising an exception is better
still. To preserve the 'undefined' meaning, any functions that can
test #<unspecified> should really be internal, e.g. in the ##sys#
namespace.

But as you say, it's unspecified, so this is a divergent discussion. :-)

>  >  There does seem to be a good case for an immediate value that *can* be
>  >  tested this way, though. John et. al. wouldn't have used (void) in
>  >  eggs if there weren't. Record instances aren't really a great answer
>  >  (though I suggested them myself) since different records of the same
>  >  type will fail an identity test.
>
>  Unless you use a single unique instance.

That would be problematic in compiled code, would it not? Where is the
instance stored?

Graham