[vtkusers] serial to parallel XML vktUnstructuredGrid converter via filenames?

[Peter Schmitt]

Berk,
Thanks for the follow-up.  I know that the XML format is still
evolving, sorry for bugging the list about it :)

FYI, I am converting a dataset to pvtu, then visualizing in paraview. 
Once we have a better understanding of the data that we received from
collaborators, we are going to write our own custom parallel viz app,
at which point we will read data in parallel in our dataset's native
format (actually, we are going to generate data on n nodes and
visualize in real time on m nodes).

I avoided writting my own program/script to write the head pvtu file
as I think there is a greater chance of me implementing a buggy
exporter vs. letting VTK handle it.  My 4-step method (see previous
response) works pretty well for our purposes right now, so I'm happy.

For future reference, how does one officially put in a feature request for VTK?

Thanks!
Pete


On 7/11/05, Berk Geveci <berk.geveci at gmail.com> wrote:
> > I agree.  However, I sometimes get frustrated wtih VTK's XML support.
> > It almost seems like they added the xml support as an afterthought and
> > rushed to get it done.
> 
> We actually spent a lot of time designing the XML file format.
> However, it had a lot of requirements and not all of them were met to
> everyone's satisfaction. Furthermore, the readers and the format are
> still evolving (which makes me glad we chose XML as the base, it is
> easier to extend without breaking compatability). Keep in mind that it
> has a lot of new features not present in the previous file formats
> (support for distributed data being the most important) and it is not
> always easy to get things perfect the first time.
> 
> > anyway, here's my current workaround:
> > 1.  load a piece of data into memory
> > 2.  use vtkXMLPUnstructuredGridWriter.  This saves the "pvtu" file and
> > the first vtu file in a series (for some reason the first vtu file has
> > a truncated filename... can't figure out how to control that)
> > 3.  Delete & allocate new vtk objects
> > 4.  Step through and load every other piece of data, saving each one
> > with vtkXMLUnstructuredGridWriter as a vtu file with the appropriate
> > filename
> 
> The pvtu format was not designed to handle streaming (which is
> essentially what you are doing) but rather to write data in parallel.
> This is why the writer does not support this. However,
> 1. It is not hard to extend the writer to do this (actually, it is
> quite easy). If you put a feature request and assign it to me, maybe I
> can get to it in the future (not near)
> 2. The pvtu format does not contain data and is simple xml. Therefore,
> it is very very easy to write with a small program or script.
> 3. We actually have a writer that does this in ParaView. We use it to
> write animations and time dependent data. It has not been moved to VTK
> yet.
> 
> -Berk
> 


-- 
Pete