I am trying to load a mesh written by gmesh (see attached) but it fails because the first cell in the cells list is type 1, causing vtk_dims[cell_types[0,0]] raise a KeyError in meshio.py as vtk_dims = {3 : 2, 9 : 2, 5 : 2, 3 : 3, 10 : 3, 12 : 3} does not handle single point cells. I don't know is these single point cells are a bug in gmesh or a feature that sfepy does not care about.

SfePy currently supports only triangles, quads, tetrahedrons and hexahedrons. What are those single point cells used for? Would it be enough to modify the mesh reader so that it ignores the unknown cell types?

One more thing - your mesh is a 3D surface (nice!), but in 3D, sfepy needs a volume mesh (e.g. of tetrahedrons).

I don't know why Gmesh adds those singular cells in vtk file and not in the other formats it supports (msh and medit for example). Would it be difficult not to load them ? Just an idea : in VTKMeshIO.read, when the 'while 1' loop ends, pop the lines of the array that has undesirable value in the first column, count the dropped cells and adjust n_el and n_val

You are right about the 3D surface, but I have the same problem with a 3D volume mesh (see cylindre.vtk). I attached in the original bugreport a lightweight vtk file showing singular cells that the sfepy does not handle.

I can now read the mesh, but there are also some line segment cells - the computation will not work with that (no bar elements in sfepy), so we should probably ignore any cells besides those with the highest dimension.

Try the attached patch, please.

With the patch, calling Mesh.from_file (in problemDef) does not raise the KeyError anymore. Thanks!

Good, here is a better patch. If all works, you can close the issue :)

It seems to work too. At least it loads the mesh file without raising any error.

Good, let's close this.

Migrated to http://github.com/sfepy/sfepy/issues/97