[vtkusers] Re: [vtk-developers] RFC: patch for VTK errors as python exceptions and trapping of bad_alloc

[Prabhu Ramachandran]

Charl Botha wrote:
 > I have just updated the Python VTK errors as exception + bad_alloc
 > trapping patch to the latest ParaView-3-2 branch checkout of VTK (it
 > should be applicable to head) and made it available at [1].  This
 > patch:
 > 1. turns all VTK errors into Python exceptions.  It even maintains
 > exception information if a user observer is invoked before the end of
 > the VTK method that triggered the error.
 > 2. surrounds all VTK method calls in the Python wrappers with
 > try...catch, so that we can catch things like bad_alloc as well.

Very cool!  I haven't tried it yet but have a few comments:

  1. You should really do: "import os.path; os.path.normpath...".

  2. Normpath will also convert "http://foo" to "http:/foo", do you
  really want that?

  3. Would it be possible to separate out the two new features into two
  separate checkins just in case?

  4. I think it would be better to use the warnings module to indicate
  warnings rather than raise exceptions.  That way a user could filter
  out warnings if they wanted to.

 > I've been running with this for quite a while, it seems quite solid
 > and I would like to commit it.  However, I would appreciate some
 > general comments, and especially on the try...catch, as it's causing
 > my VS 2005 to generate hundreds of "unreachable code" warnings in the
 > wrapper code.  Does anyone know how to get around these?  I assume

Not sure.

 > it's because the surrounded VTK method doesn't have a throws
 > classifier.  Any other thoughts?

Thanks for all the effort on this!

cheers,
prabhu