Jørgen Schartum Dokken

Close this for now. Issue made: https://github.com/FEniCS/dolfinx/issues/2348

``` import dolfinx from dolfinx import cpp from dolfinx.mesh import refine from mpi4py import MPI import ufl gm = cpp.mesh.GhostMode.shared_facet mesh = dolfinx.UnitCubeMesh(MPI.COMM_WORLD, 5, 5, 5, ghost_mode=gm) mesh.topology.create_entities(1) mesh2 =...

> I'm not crazy about adding more syntax to a (simple) demo. Is this just for running pyvista from Docker? These are so that we can run the codes through...

I guess this question should be revisited now, as we currently have basix that can tabulate at runtime. @mscroggs, @garth-wells any ideas?

> Substitution `CellAvg(u) --> u / CellVolume` does not look correct to me, these things have also different units (u-units vs. u-units/volume). And it is also wrong for constant functions,...

The problem seems to be that ufl splits integrals by domain marker. I.e. ```python from ufl import * coord_element = VectorElement("Lagrange", interval, 1) mesh = Mesh(coord_element) element = FiniteElement("Lagrange", interval,...

This is mentioned in https://github.com/FEniCS/ufl/issues/70 The big question is, why do we split integrals by domain id?

@michalhabera The issue is already in the initialization of the form: https://github.com/FEniCS/ufl/blob/main/ufl/form.py#L36-L37 As the integrals are split there, I think it is really hard to reverse engineer them back together.

Similar issue with `ipygany` as jupyter backend

@drew-parsons I think this will help when bumping different parts of the FEniCS eco-system, making it easier to provide minimal bounds (If we introduce this in Basix as well).