Re: [Chicken-users] External procedures at runtime

[felix winkelmann]

On 7/24/07, John Cowan <address@hidden> wrote:
> felix winkelmann scripsit:
>
> > What exactly do you mean with "externally callable". Do you mean
> > that it should be possible to recognize procedures defined with
> > "define-external"?  This shouldn't be too hard: the lolevel unit has
> > some functions for decorating procedures with arbitrary data and a
> > custom variant macro of "define-external" could then expand into the
> > normal define-external + decorator code.
>
> That sounds good.  (BTW, there doesn't seem to be any way for a
> extended procedure to get hold of its *own* data, unless of course
> you pass it to itself as an argument.  This would be a good thing
> to fix.)

You could also use "letrec" to refer to the procedure itself from inside
the procedure.

> Okay, then the next issue is that there is no obvious way, given
> a procedure, to pass it through the C interface: there is no
> procedure type.

But a type of generic, unconverted Scheme data. The "scheme-object"
type passes the procedure directly (see also [1]). The actual extension
data is stored in the closure slots:

; loading library posix ...
; loading library srfi-1 ...
; loading library srfi-13 ...
#;1> (define (foo) 1)
#;2> ,d foo
procedure with code pointer b7c258d3 of length 3
0: #<procedure (f_4077 . v377)>
1: ()
2: #<lambda info (foo)>
#;2> (use lolevel)
; loading library lolevel ...
#;3> (extend-procedure foo "hello")
#<procedure (foo)>
#;4> ,d #
procedure with code pointer b7c258d3 of length 4
0: #<procedure (f_4077 . v377)>
1: ()
2: #<lambda info (foo)>
3: (#(extended) . "hello")
#;4>


cheers,
felix


[1] http://chicken.wiki.br/Data%20representation