[Paraview] Convert Large CGNS Dataset to numpy

[Cory Quammen]

On Mon, Dec 5, 2016 at 11:04 AM, Thomas PEDOT <thomas.pedot at gmail.com> wrote:
> Hi,
> I have some CGNS files I want to "convert" to numpy array. To write my
> script I first use a more direct method :
>
> Converting CGNS to VTK
> Use vtk_to_numpy method
> process numpy array
>
> This task is time consuming and I need a more direct option (if possible).
>
> For this, I tried to use paraview CGNSReader and simple module.
>
> I'm able to convert CGNS data to numpy with small data set (Some MB) with
> this approach :
>
> Case = CGNSSeriesReader(filename)
> Case.UpdatePipeline()
> rawData = servermanager.Fetch(Case)
> import vtk.numpy_interface.dataset_adapter as dsa
> # Wrap the raw data object to access NumPy friendly API
> data = dsa.WrapDataObject(rawData)
>
> But as it said here : Do not do this if data is large otherwise the client
> will run out of memory.
>
> So, I can I do this without running out of memory ? I don't need to load all
> Point Arrays at once.

That warning is more for if you've got some very large data set loaded
into a ParaView server running on a large computing resource. If you
call this, it will load all that data onto the client, which may not
have enough RAM to hold the very large data set.

If you are working locally with a data set that the ParaView client
can load, then you shouldn't have any problems with the script above.

> Is it possible to select just 'Temperature' for example and only one block ?
> I guess it might be less memory consuming.

To select just 'Temperature', you should be able to do something like:

Case.PointVariables = ['Temperature']

If you need to, you can add an ExtractBlock to your pipeline:

extractBlock1 = ExtractBlock(Input=Case)
extractBlock1.BlockIndices = [1]

I probably wouldn't do the above unless you try your original script
and run into out-of-memory errors.

HTH,
Cory


-- 
Cory Quammen
Staff R&D Engineer
Kitware, Inc.