Draft: support vtk files generated from Gmsh

[GenevieveBuckley]

Closes issue #1544
Possibly related to GenevieveBuckley/napari-meshio#4 ?

I had a go at addressing issue #1544. I am confident that commit 8c42f167ea2379570bce046548c1ddbcba2bcee0 is correct, but I am not confident about the rest, and whether the overall result is correct.

Details

I used the example file suggested in issue #1544:

A simple mesh example (~200kB) is provided here: binary_mesh_example.vtk.zip

This mesh can be displayed in Paraview:

When you checkout this PR branch:

git clone https://github.com/nschloe/meshio.git
cd meshio
gh pr checkout 1544

and read in the example Gmsh .vtk file:

import meshio
mesh = meshio.read(filename)

You get this summary output:

<meshio mesh object>
  Number of points: 2238
  Number of cells:
    tetra: 6383
  Cell data: CellEntityIds

So everything seems to work ok so far, although I am not sure if this is the expected output.

To confirm whether it's working correctly, I would really like to load the mesh from meshio, into a visualization program to look at it. Then we could see if it looks like the original input or not.

[GenevieveBuckley]

Why I'm not confident this is correct

A few years ago I made a napari plugin that wraps the meshio library: https://github.com/GenevieveBuckley/napari-meshio/
I had hoped that with the changes from this PR, I could install meshio as an editable dependency locally, and run my napari plugin on top of the changes. Then I should be able to load and view the example Gmsh .vtk file (linked in the description above).

However, when I do that and try to open the Gmsh .vtk file with the napari plugin:

• It doesn't produce an error, and the surface layer that gets added does show reasonable looking data if you inspect it in the terminal
• But the napari canvas still appears black
• When you switch napari to 3D mode, there is a vispy.gloo error:

RuntimeError: All attributes must have the same size, got:
<VertexBuffer size=25532 last_dim=1>: 25532
<VertexBuffer size=19149 last_dim=3>: 19149
<VertexBuffer size=0 last_dim=3>: 0
<VertexBuffer size=25532 last_dim=3>: 25532

I know this isn't a napari related project, but it does cast doubt on whether the proposed fix in this branch is actually correct.

To figure out if the problem is:

• with this PR branch, or
• with the napari-meshio plugin, or napari itself
  I'd need to test the output mesh from this PR branch, and visualize it successfully with another program.

Alternatively, a round trip test inspecting all the vertexes and surface information of the mesh before and after the round trip could also work (but I don't have the .geo code used to generate that example Gmsh .vtk file originally, and might have to create a new example).

This might be the limit of the time I have available to work on this, but I wanted to put what I had so far in writing. Perhaps someone else can push it further along.