InteriorBasis raises IndexError on MeshLine loaded from MSH

[gdmcbain]

Following #175, another problem is that InteriorBasis(mesh, ElementLineP1()) raises

IndexError: index 2 is out of bounds for axis 0 with size 2

if mesh is loaded from a Gmsh file:

mesh = MeshLine.load(argv[1])

Example: example.tar.gz, where:

• line.geo:

Point(1) = {0, 0, 0};
Point(2) = {1, 0, 0};
Line(1) = {1, 2};
Physical Point("left") = {1};
Physical Point("right") = {2};
Physical Curve("domain") = {1};

(The same error is raised without the Physical entities too.)

• line.py

mesh = MeshLine.load(argv[1])
element = ElementLineP1()
basis = InteriorBasis(mesh, element)

[gdmcbain]

The same error is raised with just

mesh.mapping()

[gdmcbain]

Or MappingAffine(mesh).

[gdmcbain]

It looks like the problem is that mesh.p.shape[0] == 3 so that we've gotten redundant zeros for mesh.p[1:].

Is the error upstream in meshio or here in skfem.Mesh in load or from_meshio?

Does the same problem affect two-dimensional meshes? Gmsh always uses three-dimensional coordinates for nodes and I think these are passed on by meshio.read.

[gdmcbain]

I had thought to use Mesh.dim to strip the trailing coordinates but I see that it's defined a posteriori in terms of Mesh.p.shape.

scikit-fem/skfem/mesh/mesh.py

Lines 99 to 101 in 9ff1fd5

	def dim(self):
	"""Return the spatial dimension of the mesh."""
	return int(self.p.shape[0])

Would it be reasonable to define it as an a priori class variable, dim = 1 for MeshLine, 2 for Mesh2D and 3 for Mesh3D?

[gdmcbain]

Neither can something like

isinstance(cls, MeshLine)

be used in Mesh.from_meshio since MeshLine isn't defined at that point (similarly for Mesh2D).

[gdmcbain]

Ah: it was just that MeshLine doesn't override Mesh.strip_extra_coordinates as Mesh2D does.

[kinnala]

I suppose this can be closed now?

[gdmcbain]

Yes, thanks; this was fixed by 9036ffd which was included in #176.