Re: instance?

[Per Bothner <per at bothner dot com>]

Chris Dean wrote:
> I like all of these ideas, even if I disagree with some of the syntax.
> How about this syntax:
>
>   (define-namespace date <java.util.Date>)  ; create the namespace
>   (date:new)                                ; (make <java.util.Date>)
>   (date:some-method x)                      ; (invoke x 'some-method)
>   (date:instance? x)                        ; (instance? x <java.util.Date>)

I have no objections to this.
I'm not saying I'd veto (date:? x) either, but it is a little strange
looking, and I think (date:instance? x) would be more readable.

>   (date:class)                          ; <java.util.Date> equivalent

This should perhaps be (date:type) or date:type (a variable).
<java.util.Date> evaluates to gnu.bytecode.ClassType instance,
not a java.lang.Class instance.

> We could even change the name of define-namespace to define-alias in
> Kawa Scheme if we thought it was useful to separate the usage above with
> the semantics of the namespace concept in XML.

Er... define-alias already exists, and can be used for "typedefs".
See kawa/lib/arrays.scm.

Dominique Boucher wrote:

> Also, it would be really nice to be able to use the namespace in the
> type declarations. Something along the line of
>
> (define-namespace foo "class:some.java.class")
>
> (define (a-function (arg :: foo))
> ...)

That would pollute name lookup.  define-namespace defines a name
*prefix*, not name in the normal scope.  It's implemented similar to:
  (define-constant $Namespace$ "class:some.java.class")
--
	--Per Bothner
per@bothner.com   http://per.bothner.com/