Jørgen Schartum Dokken

It is in general broken. even a simpler example, such as ```python from mpi4py import MPI import numpy as np import ufl import basix.ufl import dolfinx points = np.array([[0.0, 0.0,...

This might also be because the error handling with the "new" value size is wrong.

> > This might also be because the error handling with the "new" value size is wrong. > > My gut feeling is that the difference between `reference_value_shape` and `value_shape`...

> The difference between `reference_value_shape` and `value_shape` is not documented or explained anywhere in the code. This means bugs because it's not clear which one should be used. > >...

> Thanks for the examples. Can you give a math explanation too? My initial expectation would be that the value shape remains (2,), i.e. remain in the plane of the...

Thinking even more about this, it’s clear that then name value_shape links to the shape of the basis values when pushed to the physical space. Thus it is natural that...

Do we have any resolution to this? It is quite frustrating that manifolds are kind of broken.

I guess we need to be very careful with the verification, as we would need to ensure that the coordinate element is a first order lagrange element (so the only...

I think it should be possible to handle. I've thought about this for a while, and insertion of hanging nodes would be something one could do during the refinement process,...

A prototype have been made at: https://gist.github.com/jorgensd/89acdc5534f3e01741aa6a3db279ff2d Needs some further fixes for parallel problems.