Allow promotion of geometry dimension for built-in mesh creators.

[jhale]

There is currently no easy way to create e.g. a unit interval living in 2d or
3d geometric space.

After discussion of this implementation I will implement the others the same way.

[jhale]

I will now do the work on the C++ interface to get this parameter up to Geometry.

[jorgensd]

Shouldn’t xshape[1] and mesh.geometry().dim() be the same?

[jhale]

Yes - but in terms of the code being correct in meaning, the former is the shape of the geometry data and the latter is the geometric dimension.

[jorgensd]

Do we check if they match? We should probably add some sanity checks.

[jhale]

Will have to look at this post-release - it would still be very useful to have "affine domain" non-affine mesh so that we can push-forward an affine domain with an analytical map with higher-order accuracy.

[garth-wells]

Will have to look at this post-release - it would still be very useful to have "affine domain" non-affine mesh so that we can push-forward an affine domain with an analytical map with higher-order accuracy.

I would like to make CoordinateElement as independent of Basix as possible in order to support a range of possible maps. Maybe we should have something like CellMap that is more generic.

Dealing with this could also resolve #3815. We should gave for the mesh module/namespace to have no dependency on the fem module/namespace.

[jorgensd]

Will have to look at this post-release - it would still be very useful to have "affine domain" non-affine mesh so that we can push-forward an affine domain with an analytical map with higher-order accuracy.

I would like to make CoordinateElement as independent of Basix as possible in order to support a range of possible maps. Maybe we should have something like CellMap that is more generic.

Dealing with this could also resolve #3815. We should gave for the mesh module/namespace to have no dependency on the fem module/namespace.

It would be a good idea to decouple the CoordinateElement from basix. However, to do so, we need a rather large redesign from the C++ side and carefully think about what we need to expose to make the CoordinateElement complete and usable for external packages (dolfinx_mpc, asimov-contact etc), as by currently providing the escape hatch to the basix element, these libraries can access information we do not currently expose in DOLFINx.

As a side note; Many other finite element software do treat the CoordinateElement as any standard finite element.