[SciPy-user] ANN: first release of IVuPy-0.1

Prabhu Ramachandran prabhu_r at users.sf.net
Sun Feb 12 22:43:00 CST 2006

>>>>> "Gerard" == Gerard Vermeulen <gerard.vermeulen at grenoble.cnrs.fr> writes:

    Gerard> On Sun, 12 Feb 2006 11:22:23 -0700 Fernando Perez
    Gerard> <Fernando.Perez at colorado.edu> wrote:

    >> Gerard Vermeulen wrote: > I am proud to announce IVuPy-0.1
    >> (I-View-Py).

    Gerard> My knowledge of VTK is practically zero, but I will try.
    Gerard> The motivation for developing IVuPy originated in the
    Gerard> request of a PyQwt-user (the owner of a company
    Gerard> specializing in high-end technical software) complaining
    Gerard> that VTK was: (1) slow, (2) memory hungry, and (3) hard to
    Gerard> program.

(1) and (2) depends on what is done and how it is done.  (3) is matter
of personal opinion.  :-)

    Gerard> Besides that, VTK does not integrate very well with PyQt
    Gerard> (I have tried to help to fix it, see the QVTKRender*.py
    Gerard> files in VTK).  You have to do the right thing to get it
    Gerard> working.

I thought you sent in some patches to VTK that fixed these issues?  Or
were there more problems?

    Gerard> I would be happy to take the fastest VTK version of the
    Gerard> first example of
    Gerard> http://ivupy.sourceforge.net/examples.html into account in
    Gerard> the comparison.

The VTK folks would also be interested in real benchmarks.  There are
a lot of developers working on VTK so if it is possible to improve
performance, they will.  The VTK sphere benchmark might be of use


It should not be very hard to create a simple Python script to do this
with IVuPy and VTK and compare the results.  It would be great if
someone did the benchmarking and sent the results to the VTK folks
also.  I'm not volunteering any more of my time on this, so count me

    Gerard> The concepts needed to program VTK and Coin3D seem to be
    Gerard> quite different.

    Gerard> Coin3D is a scene graph library.  Nodes in the scene graph
    Gerard> can be for instance: (1) simple shapes such as cones,
    Gerard> Coin3D is more low-level than VTK.  For instance, a 3D
    Gerard> axis frame is missing from Coin3D (it is on my todo list
    Gerard> for IVuPy) and I believe that VTK comes with more
    Gerard> algorithms such as polygonizing an implicit 3D surface or
    Gerard> contouring levels on a 3D surface.

Right, VTK provides a ton of visualization and imaging algorithms for
a wide variety of data sets.  Coin3D is as you say a lower level
toolkit that targets a different problem domain.

One more problem with Coin3D is that it is distributed under the GPL,
and unless I am mistaken, building commercial apps requires a license
from SIM.  The same applies to Qt.  Also, Coin3D is Qt/Xt only.  No
wxPython etc.  I don't think that should be a big problem though since
the number of Qt specific classes are relatively small.  However, VTK
does try to work with other GUI toolkits and already supports GTK,
wxPython, Tkinter and Qt (apparently badly).

That said, IVuPy does sound very interesting indeed.  Thanks.


More information about the SciPy-user mailing list