`Norm` should accept `VectorFunction` objects

[alex-m-h]

When one tries to compute the L2-norm of a H(curl)-conforming FemField, it has to be done by converting the symbolic field into a SymPy ImmutableDenseMatrix object, e.g.

domain = Cube(name='domain')
V = VectorFunctionSpace(name='V', domain=domain, kind='hcurl')

v = element_of(V, name='v')
l2_norm_sym = Norm(ImmutableDenseMatrix([v[0], v[1], v[2]]), domain=domain, kind='l2')

domain_h = discretize(domain, ncells=[5,5,5], periodic=[False, False, False])
V_h = discretize(V, domain_h, degree=[3,3,3])

l2_norm_discrete = discretize(l2_norm_sym, domain_h, V_h)

If instead the symbolic vector field is used directly to define the norm, i.e.

l2_norm_sym = Norm(v, domain=domain, kind='l2')

SymPy throws the exception
sympy.matrices.common.NonSquareMatrixError.

It would be more intuitive, if the symbolic vector field could be used directly to define the norm.

[yguclu]

This appears to be a problem in SymPDE, not Psydac.