Python variational solver interface fails when a form is emty

[Garth Wells]

The variational solver interface needs to be extended to support linear forms that are empty - this can occur when using lhs/rhs to separate terms in an equation F = 0. A solution would be to create a zero vector that is compatible with the matrix generated by a bilinear form.

[Anders Logg (logg)]

On Sat, Sep 10, 2011 at 11:25:46PM -0000, Garth Wells wrote:
> Public bug reported:
>
> The variational solver interface needs to be extended to support linear
> forms that are empty - this can occur when using lhs/rhs to separate
> terms in an equation F = 0. A solution would be to create a zero vector
> that is compatible with the matrix generated by a bilinear form.
>
> ** Affects: dolfin
> Importance: High
> Status: Confirmed
>
> ** Changed in: dolfin
> Status: New => Confirmed
>
> ** Changed in: dolfin
> Importance: Undecided => High
>
> ** Changed in: dolfin
> Milestone: None => 1.0-rc1

Can you provide an example?

--
Anders

[Garth Wells (garth-wells)]

Example code:

from dolfin import *

mesh = UnitSquare(32, 32)
V = FunctionSpace(mesh, "Lagrange", 1)

def boundary(x):
    return x[0] < DOLFIN_EPS or x[0] > 1.0 - DOLFIN_EPS
bc = DirichletBC(V, 1.0, boundary)

u, v = TrialFunction(V), TestFunction(V)
F = - inner(grad(u), grad(v))*dx
a, L = lhs(F), rhs(F)

# Compute solution
u = Function(V)
solve(a == L, u, bc)

[Garth Wells (garth-wells)]

Fix committed for non-symmetric assembly. WIll register another report for assemble_system.