[DiaCanvas] a few random remarks about demo.py

[<aur...@wa...>]

Le jeu 06/02/2003 =E0 10:49, arj...@he... a =E9crit :

> The only special thing is the metaclass. I would like to hear (read) yo=
ur opinion.

First I have to state a few things : I'm quite new to
- python,
- C (as in not C++)
- the whole Gnome Mess^H^H^H^H Framework

So I may lack some of the required basics to understand what's really goi=
ng on,
and have to do some wild guesses... Also the fact that I work on MOPs the=
se days
will help a little wrt the metaclass stuff...

About the metaclass : seems like a neat trick to automagically set up som=
e=20
relations between any hand-made class definition of a compound graphical =
thing
and the underlying machinery.

So :

- gobject.type_register (new_class)
  helps declare a new hand-made class to the real C cruft under the hood
  I would like to know better, but the sources are vast... where to look =
exactly ?
  DiaCanvas ? Glib ? I'll see that later.

- dia.set_callbacks (new_class)
- dia.set_groupable (new_class)

  depending on what superclass our class extends, so to speak

In the case of MyBox, it's both.
The exact semantics of 'set_callbacks' escapes me. Does it mean that the =
instance
is to handle callbacks ? (if so 'set_has_callbacks' would be better...)

So I can't say it's real good or just plain ugly. _Seems_ working.

Note about the garbage collection :

Bad things happen (segfault in fact) when invoking the gc via the menu.
Note that if you remove all the "del box, del my_box, etc..." from demo.p=
y,=20
gcing won't do any harm anymore.

This may be due to the=20
canvas.root.add(my_box2) not correctly reference-incrementing the instanc=
e=20
first associated with my_box2... some C code not doing the Right Thing ?


About selections :

When I select object, I would like to rotate/transform the selection the =
same=20
way I move it.
Selecting objects really is a strange thing here.
Maybe my anticipations are misplaced :-)

ahhhhh, I just understood : being selected is not the same as being focus=
ed.
Seems quite unnatural to me.=20

So when I select 'my_box', it can happen that the text blurb has the focu=
s.
Then, "deleting the focused" sends the whole program to segfaultland - si=
nce
it shouldn't be possible to delete it but through the whole MyBox stuff.


Synchronizing MyBox elements :

I just added to the MyBox class :
    def on_rotate (self, x, y):
        self.text.request_update()
        self.ellipse.request_update ()
        return dia.CanvasBox.on_rotate (self, x, y);

it doesn't crash, it does nothing, not even update the small circle
grouped with the box. Except for moves, the circle winds up being=20
updated AFTER the transformation in all cases.
What I don't grok is why the text blurb rotates with the box, and
not the circle.=20

Various iterator methods :

    def on_shape_next (self, iter):
	if iter is self.ellipse:
	    return None
	iter =3D dia.CanvasBox.on_shape_next(self, iter)
	if iter is None:
	    iter =3D self.ellipse
	return iter

This boggles me down... really :-)

It would be nice if we could write :

self.add_construction(self.ellipse)
then add=20
        self.update_child(self.ellipse, affine)
to
    def on_update (self,  affine):
	dia.CanvasBox.on_update (self, affine)
	self.update_child(self.text, affine)

But an ellipse appears not to be a CanvasItem. Is this really a good thin=
g ?


More broadly, I fail to see the connection between methods and their C
couterparts (staring at dia-canvas-element.c doesn't help much).

That will be all for today.

Regards,
Aur=E9lien.