wrong result of LagrangeInterpolator for vector enriched element spaces

[Jaroslav Hron]

The LagrangeInterpolator seems to give wrong results for vector enriched element spaces.

from dolfin import *

mesh = RectangleMesh(0.0, 0.0, 1.0, 1.0, 4, 4, "crossed")

V = VectorFunctionSpace(mesh, "CG", 2)
B = VectorFunctionSpace(mesh, "Bubble", 3)
U=V+B

e=Expression(("1.0","0.0"))
u=interpolate(e,U)
plot(u)

u=Function(U)
LI=LagrangeInterpolator()
LI.interpolate(u,e)
plot(u)
interactive()

[Jan Blechta]

I'm not sure if it has a chance to work (doc says This class interpolates efficiently from a GenericFunction to a Lagrange Function) but it should at least check a validity of the arguments.

[Mikael Mortensen]

I'm looking into it. Not sure whether or not it should work for this enriched element. I have not tested it previously.

[Jaroslav Hron]

There seems to be similar problem in dolfin interpolate - see #489 - it might be connected?

[Jan Blechta]

After resolving #489, the original example now raises

RuntimeError: evaluate_dof(s) for enriched element not implemented.

on u=interpolate(e,U). Switching it to u=project(e,U) raises

RuntimeError: tabulate_coordinates is not defined for this element

on LI.interpolate(u,e).

Blueprint for correct reimplementation: https://bitbucket.org/fenics-project/ffc/issue/69.

[Jan Blechta]

BTW, https://bitbucket.org/fenics-project/ffc/issue/69 suggest that dofs of enriched element will be linear combinations of Dirac deltas. I'm not sure whether tabulate_coordinates (used by LagrangeInterpolator) will have a sense in this situation.

[Martin Sandve Alnæs]

Removing milestone: 1.6 (automated comment)