ffc team mailing list archive

[Michael Brandl <michibrandl@xxxxxxx>]

Hi,

thanks to everybody for the help, it works fine with the 
"range"-statement. This will be a little less manageable when treating 
larger expressions with several terms that are indexed in different 
ways, though, so I am looking forward to the mentioned fix in UFL.

However, since I am new to using Dolfin: Is FFC already using UFL? Is it 
reasonable for me to expect a fix to this problem within the next weeks, 
or should I focus on getting my fcc files to work with "range"-statements?

Thank you, and thanks for the corrections in the manual,

/Michael Brandl



Anders Logg wrote:
> On Wed, Oct 08, 2008 at 11:58:23PM +0200, Michael Brandl wrote:
> > Hi,
> >
> > I am new to this list and using ffc for my bachelor thesis.
> >
> > Hopefully I am not asking something stupid which has already been asked; 
> > I tried to look through FFC-dev's archieves, but didn't find an answer.
> >
> > What I want to do is using a mixed element (with a vector element and a 
> > finite element), but then to access the single components of a test or 
> > trial function defined on the vector element.
> >
> > A concrete example would be this modification of the "Mixed formulation 
> > of Stokes" from the documentation:
> >
> > P2 = VectorElement("Lagrange", "triangle", 2)
> > P1 = FiniteElement("Lagrange", "triangle", 1)
> > TH = MixedElement([P2, P1])
> > (v, q) = TestFunctions(TH)
> > (u, p) = TrialFunctions(TH)
> > f = Function(P2)
> > a = (dot(grad(v), grad(u)) - div(v)*p + q*div(u))*dx
> > L = (v[i] * f[i])*dx
> >
> > where I only changed the last line (from being "L = dot(v, f) * dx" into 
> > its componentwise computation).
> >
> > I get the following error:
> > --------
> > Preprocessing form file: Test.form --> Test.py
> >
> > Traceback (most recent call last):
> >    File "/usr/bin/ffc", line 180, in ?
> >      sys.exit(main(sys.argv[1:]))
> >    File "/usr/bin/ffc", line 107, in main
> >      execfile(outname, ns)
> >    File "Test.py", line 27, in ?
> >      L = (v[i] * f[i])*dx
> > TypeError: list indices must be integers
> > -----
> >
> > I am afraid that I am not good enough in Python to find a solution to 
> > that, therefore I would appreciate any kind of help very much.
> >
> > Thank you!
>
> This is expected. FFC handles mixed elements like the one above by
> first creating a vector-valued element that holds all components of
> the system. In this case a vector-valued element with three components
> is created.
>
> Then TrialFunctions() etc creates a basis function in this space (with
> three components) and returns components of this basis function split
> into the parts that make up the element:
>
>   w = TrialFunction(TH)
>   u = [w[0], w[1]]
>   p = w[2]
>
> So u is just a Python list and it does not know how to handle i which
> is an FFC object of class Index.
>
> This will be fixed in UFL as already noted by Martin.
>
> > By the way, while going through this example (Mixed formulation of 
> > Stokes)from the documentation I found some minor errors there:
> > It should be
> > "P2 = VectorElement("Lagrange", "triangle", 2)"
> > instead of
> > "P2 = FiniteElement("Vector Lagrange", "triangle", 2)",
> > "a = (dot(grad(v), grad(u)) - div(v)*p + q*div(u))*dx"
> > instead of
> > "a = (dot(grad(v), grad(u)) - div(v)*P + q*div(u))*dx";
> >
> > and finally the sentence about the example file with Heat.form seems to 
> > be misplaced; I didn't manage to find this example there.
> >
> > Thank you very much,
> >
> > yours sincerely,
> >
> > Michael Brandl
>
> Thanks for reporting this. I see that Kristian has already updated the
> manual.
>
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> FFC-dev mailing list
> FFC-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/ffc-dev