Re: bonobo, gtk widgets, DOM

["Gustavo J. A. M. Carneiro" <gjc inescporto pt>]

A Seg, 2003-07-14 às 21:23, Luca Padovani escreveu:
> On Mon, 2003-07-14 at 15:06, Gustavo J. A. M. Carneiro wrote:
> 
> >    Definitely alternative c) is the best.  Maybe XPath identifiers are
> > appropriate for this purpose?...
> 
> this is interesting, the problem is that not all the communication is
> from the user to the widget. Had it been this way, it would only be the
> user to provide Xpath expressions, and our DOM implementation already
> supports xpath interpretation. But when the widget has to notify the
> user with a particular node, it would have to generate an xpath
> expression for that node, and I don't see a sensible way to do this in
> general.
> 
> >   Alternative a) also has some potencial, but you should create DOM node
> > CORBA wrapper objects only on-demand.  Don't create a BonoboObject for
> > each DOM node in a tree!
> 
> oh that's for sure. BTW, the bonobo wrapper for DOM might be generated
> automatically almost completely, but it looks like an overkill solution
> anyway.
> 
> Is it possible that nobody else had this problem before?

  In fact, I did have a similar problem.  In NumExp, a math program, a
math expression is represented internally a tree of "elements".  The
bonobo server communicates elements with the outer world with integer
identifiers.
  In the beginning, to obtain an identifier for the address of an
element I would cast the address of the element to a long int.  
  Currently, a different implementation is used, where a random number
is generated to represent each element.  See [1] for the implementation
of this, though I'm not 100% sure this is bug free.  Anyway, if you find
it useful, use it in whatever license you want.
  Regards, and good luck.

PS: Several times I found GtkMathView potentially interesting, but I
never could compile it. Too many dependencies, dependencies on unstable
an difficult to build dom library (which brings more dependencies),
etc.  I become some frustrated that I started my own (python) mathml
project[2].  Still, I feel a bit bad about all this duplication of code.

[1]
http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/numexp/numexp-core/bindings/bonobo/nxp-ticket-source.c?rev=1.2&content-type=text/vnd.viewcvs-markup
[2] http://pymathml.sf.net


-- 
Gustavo João Alves Marques Carneiro
<gjc inescporto pt> <gustavo users sourceforge net>