Re: define-struct

[James Henstridge <james daa com au>]

Andrei Zmievski wrote:

> At 09:57 AM 12/13/01 +0800, James Henstridge wrote:
>
> > Could you give some examples of the structs you want to use 
> > (define-struct ...) for?  I have found that most of the structs that 
> > are not registered with the gtype system often map nicely to native 
> > types in python (such as GdkPoint to a 2-tuple).
> >
> > One other definition we probably need to add is a (define-pointer 
> > ...) one.  Sonething like:
> >  (define-pointer CTreeNode
> >    (in-module "Gtk")
> >    (c-name "GtkCTreeNode")
> >    (gtype-id "GTK_TYPE_CTREE_NODE")
> >  )
> >
> > This is to cover types that are derived from G_TYPE_POINTER.  These 
> > types don't have any copy/free functions, so are not handled quite 
> > the same as G_TYPE_BOXED types.
>
>
> What about GdkPixbuf struct, for example? It's not derived from 
> GtkObject/GObject, but has _ref and _unref functions.

With the new defs format, I am mainly thinking about gtk 2.0.  With 2.0, 
GdkPixbuf is a GObject.

For gdk-pixbuf-0.x releases (the current versions for use with gtk 1.2), 
it should fall under the boxed type category (the ref/unref functions 
correspond to the copy/free of a boxed type).  Unfortuately, it isn't 
registered with the typesystem as such :(

This (define-pointer ...) would be for types registered as derivatives 
of G_TYPE_POINTER.


Before I forget, msw was hacking on support for marking definitions as 
deprecated (eg. gtk_widget_set_usize() being deprecated in favour of 
gtk_widget_set_size_request).  This is done through a (deprecated 
"warning message") clause.  For python, this uses the python warning 
framework, other languages might do the same, ignore the information in 
the defs file, or do a g_warning().

Looks like we need another draft of the defs spec (and look at 
finalising it -- glib/gtk are in API freeze now).

James.

--
Email: james daa com au
WWW:   http://www.daa.com.au/~james/