dolfin team mailing list archive

[kent-and@xxxxxxxxx]

Are there any thoughts on how to implement TensorFunctionSpace in Dolfin ?

Copying code for VectorFunctionSpace it could be something like:


 class TensorFunctionSpace(MixedFunctionSpace):
    "VectorFunctionSpace represents a vector-valued finite element
function space."

    def __init__(self, mesh, family, degree, dim=None):
        """Create tensor-valued finite element function space. The
function space
        may be created by

            V = TensorFunctionSpace(mesh, family, domain, degree, dim=None)

            mesh    : a Mesh
            family  : a string specifying the element family
            degree  : the degree of the element
            dim     : an optional argument specifying the number of
components

        If the dim argument is not provided, the dimension will be deduced
from
        the dimension of the mesh.
        """

        # Create subspaces
        dim = dim or mesh.geometry().dim()
        spaces = dim*[VectorFunctionSpace(mesh, family, degree, dim)]

        # Initialize base class
        MixedFunctionSpace.__init__(self, spaces)



But then I get the following when in use:

    condition or error(*message)
  File "/usr/local/lib/python2.5/site-packages/ufl/log.py", line 97, in error
    raise UFLException(self._format_raw(*message))
ufl.log.UFLException: Transposed is only defined for rank 2 tensors.

Any thoughts ?


Kent