dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

Bartosz Sawicki wrote:
> Hi there,
>
> I'm just starting to learn Dolfin/FFC... projects. I'm impressed by
> capacities it has, design and code quality. Thank you for such excellent
> work.

Thanks!

> At this moment I have one question, which probably shouldn't be sent
> into dolfin-dev, but there are no other list (for users):

It is perfectly fine to send it here. (If the traffic should increase 
much, we could add a list dolfin-users.)

> Is it possible to make simple post-processing using build-in features of
> Dolfin?

Yes.

> For example, I think about grad/rot/div operators. In the simpliest case
> /demo/pde/poisson, discrete function u is computed. Having u, I would
> like to compute also e=-grad(u). Of course, I can write own code
> iterating over cells and calculate gradient, but it won't be very
> elegant solution.
> How would you solve such problem?

Define a form that projects whatever quantity you need to a suitable 
space. Say you want to compute the gradient of a piecewise linear function:

P1 = FiniteElement("Lagrange", ..., 1)
P0 = VectorElement("Lagrange", ..., 0)

v = TestFunction(P0)
w = TrialFunction(P0)
u = Function(P1)

a = dot(v, w)*dx
L = dot(v, grad(u))*dx

Then solve the system and w will be the L2-projection of the gradient of 
u (which will be exactly equal to the gradient of u since the gradient 
is in P0).

You might want to project to P1 if you want to plot the gradient since 
most postprocessing tools can only handle P1.

Note that you can also define functionals like the integral of u along 
the boundary, the average etc. (Use M instead of a, L in FFC.)

/Anders


> Regards,
>
> Bartosz Sawicki
> Warsaw University of Technology
>
>
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/dolfin-dev