How can I assemble the (dense, rectangular) Jacobian of a side constraint?

[pefarrell]

(Possibly not a bug/user error)

I'm trying to assemble the Jacobian of a side constraint in a PDE-constrained optimisation problem. The constraint is that u*dx == V. I test this equation with a TestFunction from FunctionSpace(mesh, "R", 0). However, when I take its Jacobian and try to assemble it, I can't seem to get an explicit PETSc matrix out. (I intend to pass the assembled Jacobian to IPOPT.)

Consider the following code:

from firedrake import *

mesh = UnitSquareMesh(2, 2)
V = FunctionSpace(mesh, "CG", 1)
R = FunctionSpace(mesh, "R", 0)
v = TestFunction(R)
u = TrialFunction(V)

form = inner(u, v)*dx
J = assemble(form, mat_type="aij")
J.force_evaluation()
print(J.petscmat.type)

I was expecting it to print "aij", but instead it prints "python". How can I get the matrix entries out? (mat_type="dense" isn't recognised.)

[dham]

The rectangular blocks produced by the "R" space are represented as PyOP2 dats wrapped in PETSc Python matrices. I think J.petscmat.getPythonContext().dat will give you the entries which live on this process. This possibly should be done with a PETSc dense matrix. At the time, I didn't know such things existed.