Re: Type BOOLEAN

On Tue, 8 Jun 2004 19:13:56 -0600, Kevin Rosenberg <kevin-HJRc7zDS/[email protected]> wrote:

> To the degree, that "declaration are assertations", it's not okay to
> lie to Lisp. This is implementation specific, of course. OpenMCL,
> for example, computes a slot named ccl:type-predicate for each
> effective-slot-definition. It uses this predicate to check values
> stored with (setf slot-value-using-class).

Yes, sorry, that was of course just plain dumb of me. I think CMUCL
19a with Gerd's PCL might do something similar.

> I have arranged a hook in sql/metaclasses.lisp named
> compute-lisp-type-from-slot-specification where it's possible to
> change the actual lisp type for a :type attribute in a slot. So, it
> is possible to have ":type boolean" actually translate to lisp as
> ":type t". While possible, I've tried to stick as much to Lisp
> semantics as possible such that: (typep slot-value slot-type) => t
>
> As stated, I've thought about this issue and I'll be glad to discuss
> it further.

How about having a user-specified slot-type GENERALIZED-BOOLEAN which
translates to the CL type T but is stored as

  (if val "t" "f")

in the database?

Cheers,
Edi.