Re: GObject docs improvements

[David Nečas (Yeti) <yeti physics muni cz>]

On Fri, Feb 09, 2007 at 01:42:19PM +0100, Stefan Kost wrote:
> > Deriving from classes that implement interfaces and changing
> > the implementation would worth an explanation and example.
>
> Huh, sounds scary.
> ...
> Do you have pointers to an example?

I must admit while I do this in some real world code and it
works, I have never verified I do it right by analysing the
gtype.c code (too scary).  That's one of the reasons why
I suggested it...

A testing (non-real world) example is attached, of which
only MyPrintable, MyParent and MyChild2 are interesting
(MyChild does not change the implementation).  If I attempt
to summarize it:

1. Implement the interface *again* in the subclass (this is
the controversial point):

G_DEFINE_TYPE_EXTENDED
    (MyChild2, my_child2, MY_TYPE_PARENT, 0,
     G_IMPLEMENT_INTERFACE(MY_TYPE_PRINTABLE, my_child2_printable_init))

2. Get the parent interface in the interface init function
with g_type_interface_peek_parent() and override methods:

static void
my_child2_printable_init(MyPrintableIface *iface)
{
    my_child2_parent_printable = g_type_interface_peek_parent(iface);
    iface->print = my_child2_print;
}

3. Use thus obtained parent implementation to call it in
the overriden method (if suitable):

    my_child2_parent_printable->print(printable);

> While checking for an example, just noticed that the "Implemented  
> Interfaces" stuff in Gtkdoc seems quite borked. Need to fix that.

In what sense?  It seems to work for me -- though I should
note I use svn trunk gtk-doc.

Yeti


--
Whatever.