difference in behaviour of split

Question

Suppose U belongs to a mixed function space.
Following
solver.solve()
I have U which holds the solution.

Now consider

#  Works
(v, p) = U.split(True)
v_nodal_values = v.vector().array()


# Doesnt work
v, p = split(U)
v_nodal_values = v.vector().array()
AttributeError: 'ListTensor' object has no attribute 'vector'

What is the difference between them ? Where in the documentation could I have found the difference ?

Answer 1

Not sure how well that's documented. I remember this issue has been brought up a few times in this forum. In fact searching for "split" issues on bitbucket returned this

https://bitbucket.org/fenics-project/dolfin/issue/194/split-u-versus-usplit

So that's still work in progress, I suppose.