Re: Proposal fort GTK bindings

[Arun Sharma <asharma netscape com>]

On Tue, 04 Nov 1997 15:08:30 +0100, Michael Lausch wrote:

	m> I poropose the following archtiecture of the GTK pYthon
	m> bindings.

	m> provide bindings for all the GTK functions in a gtk
	m> module.

This is what Neil has done in his SWIG based bindings. I was trying to
save a function call in not doing it:

Gtk.Button("foo") {Python} -> gtk.gtk_button_new_with_label("foo")
{Python} -> gtk_button_new_with_label("foo") {C}

is slower than

gtk.button("foo") -> gtk_button_new_with_label("foo")

I don't know if the savings will be significant, but I didn't think
that not going with SWIG would introduce a huge implementation
penalty.

	m> Build a python class hierarchy using these functions in
	m> the Gtk module.

The class hierarchy would be relatively flat for now. I don't think
there is a need to expose abstract classes in gtk (gtk_box for
example) in Python.

However, object oriented principles have been used in that:

both gtk.vbox.pack() gtk.hbox.pack() eventually translate to 
gtk_box_pack_start().

Of course, any new classes which add extra functionality to these
classes would form a class hierarchy.

	m>  The user should only access to Gtk module.

Certainly.

	m> Why now implement the class hierarchy in C? Because it's
	m> not possible, without overhead, to inherit from C defines
	m> classes.

Could you please elaborate ?

	-Arun

[ CC'ed to gtk-list so that more people can take part in the discussion]

-- 
#include <stddisclaimer.h>