Re: Gtk::Container child properties

[Juan Rafael García Blanco <juanrgar gmail com>]

Hi,

I’ve just started working on this. I think we would need to provide a similar API for child properties as for 
other properties, because I don’t think all GtkContainer (maybe not even one) expose child properties through 
regular methods.

I see a major problem in doing this and I can’t find a valid approach that does not imply reimplementing a 
lot of things; maybe you could help me. The point is child properties are only available in Gtk+/gtkmm, while 
most of the infrastructure for generating code belongs to glibmm, and we don’t want to make glibmm depend on 
Gtk+ I think.

First, we would need to make child properties appear in gtk_signals.defs or other .defs file. That could be 
done more or less easily creating an additional get_defs in gtkmm.

Second, we would need to extend gmmproc to handle child properties. I see here a major problem. I’m not sure 
we could use the same _WRAP_PROPERTY / _PROPERTY_PROXY macros for child properties, i.e I don’t know if we 
could handle differences between gobject properties and child properties at pm level or at m4 level. What is 
your feeling about this?

Third, we need to create a new PropertyProxy. I think this is the easiest; I’ve written some code for this on 
a local branch.

Could you please comment on this?

Thank you very much.

Best regards,
Juan.


On 13 Apr 2014, at 10:17, Kjell Ahlstedt <kjell ahlstedt bredband net> wrote:

> 2014-04-12 20:35, Murray Cumming skrev:
> > On Sat, 2014-04-12 at 19:16 +0200, Juan Rafael García Blanco wrote:
> >
> > > Hi,
> > >
> > > I’ve observed that subclasses of Gtk::Container tend to declare properties on their children. In the 
> > > gtkmm source code I’ve read that they are used for theming so that there is no point in wrapping them. 
> > > Does this currently hold?
> > Probably not.
> >
> >
> > >  IMO there is no such difference between properties and child properties.
> > A child property is just a property of a container's child, such as it's
> > alignment or whether it should expand, I think. I'm sure they are
> > useful, if we can find a nice API for it. But it might not be necessary
> > to use the generic property API. Maybe all the functions do what we need
> > already. I don't know.
> See also this 10 year old bug report: https://bugzilla.gnome.org/show_bug.cgi?id=140515