Dolfinx discontinous expression

Hi all, I want to define a discontinous expression on two subdomain. The expression equals to 0 and 1 in the two subdomain respectively. How to realize it using dolfinx? Thank you in advance.

Hi, is this a expression that can be expressed geometrically, like Domain a 0.5<=x[1], Domain 2 x[1]=<0.5, or is it defined through a mesh function?

You could for instance do the following for geometrical search:

import dolfinx

mesh = dolfinx.UnitSquareMesh(dolfinx.MPI.comm_world, 9, 9)
V = dolfinx.FunctionSpace(mesh, ("DG", 0))
v = dolfinx.Function(V)
x = V.tabulate_dof_coordinates()
for i in range(x.shape[0]):
    midpoint = x[i,:]
    if midpoint[0]> 0.5 and midpoint[1]>0.25:
        v.vector.setValueLocal(i, 2)
        v.vector.setValueLocal(i, 1), "DG.xdmf").write_checkpoint(v, "v", 0)

and this if the data comes from a MeshFunction

v2 = dolfinx.Function(V)
mf = dolfinx.MeshFunction("size_t", mesh, mesh.topology.dim, 0)
mf.mark(lambda x: numpy.logical_and(x[0] < 0.5, x[1]>0.2), 2)
for i in range(x.shape[0]):
    if mf.values[i] == 2:
        v.vector.setValueLocal(i, 2)
        v.vector.setValueLocal(i, 1), "DG_mf.xdmf").write_checkpoint(v, "v", 0)

dokken, it’s the answer I want. Many thanks.

Dolfinx is more pythonic and convenient than dolfin. I like it.

That is one of the goals of dolfinx. For instance, you can use numba to create custom assemblers, use normal Python functions (lambda functions) to locate dofs, integrate expressions etc.

Numba could be used in this case to accelerate the loops.


Hi everyone, sorry if I re-open this but I was wondering how to avoid the for loop indicated by @dokken (just to save some lines).

I would like something like:

foo = dolfinx.fem.Function(V)
foo.interpolate(lambda x: 1. if x[0] < 0.5 else 0)

But this does not work. The only way I found to avoid the for loop is to do the following:

x = ufl.SpatialCoordinate(mesh)
foo_ufl_exp = ufl.conditional([0], 0.5), 1., 0)  # stands for 'x[0] < 0.5 ? 1 : 0'
foo_exp = dolfinx.fem.Expression(foo_ufl_exp,
# interpolate

But this look even less clear than the for loop. Anything I am missing?

See numpy.where.

1 Like

foo.interpolate(lambda x: np.where(x[0] < 0.5, 1, 0)) does the job indeed.

Thank you very much @nate