What's the proper way to open parallel VTK files (*.pvtu)?

First thanks for this wonderful package! It is a much more sane way to manipulate VTK files in Python…

I wonder what’s the proper way to open parallel VTK files? It consists of a header file (*.pvtu) that points to normal VTK files (*.vtu). In ParaView, you can simply open the top-level *.pvtu file, and ParaView will automatically assemble the small VTK files into a single object.

Directly calling vtki.read('*.pvtu') leads to OSError: This file was not able to be automatically read by vtki.. So I open each small VTU file and use vtki.pointset.UnstructuredGrid.merge() to put them together. The full reproducible code is at https://github.com/JiaweiZhuang/vtk_experiments/blob/master/vtki_read_pvtu.ipynb, with small sample data files.

However, merge() seems to have problems at the boundaries between different subdomains:

The blue/green lines are artifacts that shouldn’t exist in the original data. They correspond to the boundaries of domain-decomposition:

Here each color represents a VTK file.

Thus I wonder what’s the proper way to open and merge parallel VTK files? Would it be useful to add an API similar to Xarray’s open_mfdataset() to deal with multiple files?

Disclaimer: I have zero experience with VTK, but need to process some VTK files for a class project. Please forgive me if I miss anything obvious 😃

[banesullivan]

It’s now on the master branch and an example is live in docs: http://docs.vtki.org/examples/00-load/read-parallel.html

[JiaweiZhuang]

Yes I think that’s okay!