[vtkusers] vtk and python - some questions.

[Prabhu Ramachandran]

hi,

>>>>> "Chris" == Chris Myers <myers at tc.cornell.edu> writes:

    Chris> This is somewhat tangential to the main thread regarding
    Chris> VTK, Python and SWIG, but it is related.

<snip>

    Chris> Has anybody thought about such things, or how they might be
    Chris> made to work?

        Well, if the python wrappers were perfect, this would never be
a problem. :) I think using py_cpp is a step towards the goals you are
seeking.  this is because, if all of VTK were wrapped with py_cpp then
vtkUnstructured grid would be a proper python object and not only
could you pass it along to any function that requires a
vtkUnstructuredGrid you could also derive a sub class and pass an
instance of the subclass!!  Imagine the number of cool things you
could do.

        OTOH if you had a contributed c++ class that you want to use
from python you could still wrap it using py_cpp and then expose just
that class and use it with the rest of the python classes.

        Hopefully, it should not be too hard to use py_cpp.  It
certainly looks interesting and is well worth the effort.  Let me see
if I can find some time soon. :)

    Chris> P.S.  Having looked at the VTK introspection code that
    Chris> Prabhu wrote to implement his VTK-CFD, I can understand his
    Chris> frustration with the lack of certain features in the
    Chris> Python/VTK interface.  SWIG-wrapped objects can be
    Chris> examined, for example, through the __class__ and __bases__
    Chris> attributes, to reconstruct available methods for those
    Chris> objects.  I dug around in the Python/C API a bit trying to

        There are positive things to what I had to do.  I learnt
regexp's while building vtkPipeline.py. ;) Besides the current
wrappings do allow one to do quite a bit (with a little imagination).
The only thing is that vtk objects behave strangely at times.  Most of
the functions work, except for wierd Segfaults which I have to trap
with dirty hacks.

    Chris> understand this better when Prabhu was asking a few months
    Chris> ago about introspection-related matters, but I don't recall
    Chris> reaching any real conclusions.  Are there attributes of
    Chris> python objects that could be incorporated into the existing
    Chris> VTK/Python wrappers without introducing too much extra
    Chris> code?

        Well, IMHO, the pipeline browser for python works well.  So I
really cannot blame the python wrappers.  But there is scope for some
fantastic improvements with the wrappers.  I sincerely hope I can be
part of those improvements. :)

        Actually, if we do improve the python wrappers, the pipeline
browser would really suffer and I'd have to rewrite the parser engine.
But that shouldnt be too hard.

prabhu