Re: Small hack to help interactively identify components of an object

Hammond, Mark ((@)
Tue, 20 Sep 94 09:57:00 PDT

> >>>> On Sun, 18 Sep 1994 23:51:11 GMT, MHammond@cmutual.com.au (Mark
Hammond)
> said:
>
> An alternate solution to the same problem is to automatically generate
> binding functions from C, C++ or some intermediate header file. This
> is the approach we have been taking. While doing this, I have found
> many uses for my type system. The approach is similar to what Guido
> has done for the gl module.

Yes - this is a good idea, but not as flexible as would be nice.
>
> > So apart from the nice features you describe that your system would
give,
> > I can see other benefits of such a tool.
>
> Are you just refering to the problem that you mentioned above, or do
> you have other ideas as well? I would be interested to know what
> people would find useful in such a system. Much of the coding is
> already done, but many issues still remain, like how to bind it to
> Python for example. So any suggestions would be welcomed.
There are many places I would like this mechanism.

For example, in the GUI we are building, the full Windows message structure
is passed as a param to
an even handler in Python. Currently, this is passed as a tuple. It would
be really nice to have
"named" structure elements. Dictionaries could be used, but not really nice
in this case.

I think that the uses are limited only by our imagination.

But in an earlier post, Guido mentioned the problems integrating this stuff
into the existing Python
code base, and his concerns are certainly justified. But it is heartning to
hear his support for the concept.

Maybe we could start with your stuff as functionality requirements, and
discuss a more "Python friendly" way of implementing it (involving Guido,
and making the appropriate trade-offs where required). This seems the best
way to get this stuff rolling in a productive manner.

Mark.