Re: signals versus vfuncs

[muppet <scott asofyet org>]

On Friday, January 9, 2004, at 07:53 AM, Murray Cumming Comneon com 
wrote:

> An important part of this is that we can always get the C++ instance 
> from the C instance. We use quarks with g_object_set_data() to store a 
> pointer to the 1 and only C++ wrapper instance in the C instance. 
> Without that, the overriden C vfunc in gtkmm would not be able to call 
> the C++ instance's method.

we do something very similar in Glib::Object, for the same reasons --- 
we store a pointer to the wrapper in instance qdata.  in fact, we also 
store the pointer to the C object in "magic" attached to the Perl 
object, so to Perl developers it looks like a normal blessed hash, in 
which you can store whatever you want.  there's a fancy black-magic 
hack in which we share reference counts between the Perl and C objects 
in order to ensure that they go away at the right time, completely 
removing the burden of reference counting from the Perl developer.  
it's very nice.  ;-)


> I see that Christof understands gtkmm internals very well. I'll be 
> sending
> relevant patch requests to him in future.

:-D

--
elysse (pregnant): are your hands cold?
me: uh, i suppose so.
elysse: will you put them on me?