Cannot read VTK files with METADATA

[Karol Lewandowski]

When we try to load a vtk file (moab.load_file("plane.vtk", 0, "");) created for example with ParaView 5.7.0 we get the following error:

MOAB ERROR: Parsing error at line 230: expected one of { VERTICES LINES POLYGONS TRIANGLE_STRIPS } got "METADATA"!
MOAB ERROR: match_token() line 378 in src/io/FileTokenizer.cpp
MOAB ERROR: --------------------- Error Message ------------------------------------
MOAB ERROR: Failed to load file after trying all possible readers!
MOAB ERROR: serial_load_file() line 623 in src/Core.cpp
MOAB ERROR: load_file() line 514 in src/Core.cpp

However, when we delete the metadata lines from vtk e.g.:

INFORMATION 2
NAME L2_NORM_RANGE LOCATION vtkDataArray
DATA 2 0.516745 1.17908 
NAME L2_NORM_FINITE_RANGE LOCATION vtkDataArray
DATA 2 0.516745 1.17908

Everything works fine.

[Vijay M]

Thanks for raising the issue. We will add some switches to accomodate these section changes. I will let you know when a PR is ready for review.

[Vijay M]

@Iulian Grindeanu can we update VTK reader to ignore the metadata sections for now ?

[Iulian Grindeanu]

is there somewhere a description of the format ? it is a newer vtk format, and we say we support basic, legacy formats; I could not find a reference to metadata fields even in
https://vtk.org/wp-content/uploads/2015/04/file-formats.pdf

so it must be something newer than that ; suggestions ? @Karol Lewandowski

‌

[Antoine Mazuyer]

Yes, it seems to have issues with MOAB and the file format support. Paraview and vtk writes only the version 5.1 now. And MOAB reads only v3.0 or below.

[Iulian Grindeanu]

is there a manual / description of the version 5.1 format ?

[Antoine Mazuyer]

Hi @Iulian Grindeanu . According to this post, there is no official documentation for the new legacy format, it’s still work in progress

[Iulian Grindeanu]

Thanks Antoine; so I assume the only solution is to look at the source code, and try to reverse engineer the format; maybe the source code is documented. Does anybody know ? Is Paraview using directly vtk reader/writers for the new legacy code? Or do they process it ? Probably I have to see vtk source code, not Paraview

[Antoine Mazuyer]

You are right : the vtk code is included within Paraview, and this is this code which is called by Paraview.

My conclusions were that Paraview/VTK has the backward compatibility to read the old file format (<= 3.0), but can’t save them anymore (just the 5.1).

As far as I know, the main difference are in the way the cells are written. Indeed, there is “three” subsections after the “CELL” sections (CONNECTIVITY, CELL_TYPES and OFFSETS) that are handled by this :

https://gitlab.kitware.com/vtk/vtk/-/blob/master/IO/Legacy/vtkUnstructuredGridWriter.cxx#L38

and this

https://gitlab.kitware.com/vtk/vtk/-/blob/master/IO/Legacy/vtkDataWriter.cxx#L2292

[Antoine Mazuyer]

I gathered simple data set in the “old” and “new” format : one sphere meshed with tetrahedra (I also put some simple field data for the new version), and one cube meshed with arbitrary polyhedron.

I hope it can help.

‌

https://gofile.io/d/RB2jTf

‌

[Herve Gross]

The most explicit documentation I have found is here : https://vtk.org/Wiki/VTK/Polyhedron_Support

Thanks for your help with this issue!

[Vijay M]

@Iulian Grindeanu Any progress on this ?

[Iulian Grindeanu]

no, sorry; it seems that the solution is to reverse engineer / understand what paraview does; we have a list of tokens that we process; and should adapt what we do to the new format; even vtk examples site is outdated:
https://kitware.github.io/vtk-examples/site/VTKFileFormats/

we support polyhedra in our moab vtk reader / writer since 4.9.1; there is no doc though on how to interpret the metadata, and what can be skipped safely;

‌

[Iulian Grindeanu]

@Antoine Mazuyer can you please update the link with your example? it expired and I missed it :)

[Iulian Grindeanu]

so there are several issues; legacy format is vtk 3.0, which we support, including polyhedra and polygons; paraview 5.7 exports in format 4.2, which seems to be very similar to 3.0, except for metadata info; paraview 5.8 exports in vtk format 5.0, which is different, because it introduces, as Antoine pointed out, new subsections, OFFSETS, CONNECTIVITY and then old CELL_TYPES ; The only way to even obtain example files is to run paraview 5.7, 5.8 and export old files in new format; I just deleted 5. 7 from my laptop, I have to install again

[Vijay M]

@Iulian Grindeanu Any progress ?

[Iulian Grindeanu]

no, sorry