ffc team mailing list archive

[Shawn Walker <walker@xxxxxxxxxxxxxxx>]

On Sun, 23 Aug 2009, Kristian Oelgaard wrote:

> Quoting Shawn Walker <walker@xxxxxxxxxxxxxxx>:
>
> > Hello.  I would like to know how hard it would be to modify FFC (and UFL)
> > to do the following.
> >
> > Here is the sample .ufl file:
> >
> > Poisson.ufl
> > ------------------------------------------
> > element = FiniteElement("Lagrange", "triangle", 2)
> > vector  = VectorElement("Lagrange", "triangle", 2)
> >
> > v = TestFunction(element)
> > u = TrialFunction(element)
> > f = Function(element)
> > G = Function(vector)
> >
> > a = inner(grad(v), grad(u))*dx
> > L = v*f*dx
> > ------------------------------------------
> >
> > And I would like to do the following in C++:
> >
> > // in C++
> > a.G = Some_Function;
> >
> > Basically, I just want to attach to `a' some external function that does
> > NOT appear in the bilinear form.  How hard would that be?  I'm not sure
> > what the UFL syntax should be.  This is for the higher order mesh stuff.
>
> I'm sure it's possible, but does it make sense that a has a function attached on
> which it does not depend?
> The code that will be generated from this form, will not use G in any way, so
> how do you plan on using it? For higher order mesh stuff, should one not just
> use:
> triangle = Cell("triangle", 2) # Cell() is defined in geometry.py in UFL
>
> and then define the finite element as:
>
> element = FiniteElement("CG", triangle, 1)
>
> maybe we should/could attach the function G to the cell if you intend to use
> this to compute the geometry?
>
> Kristian

Yes, that is exactly what I want to do!  I'm not sure completely about the 
Syntax though.  Would this make sense:

--------------------------------------
# this is just a normal lagrange polynomial over a straight triangle
G  = VectorElement("Lagrange", "triangle", 2)

# now define what the higher order element is
my_triangle = Cell("triangle", G)

# then define the element
element = FiniteElement("CG", my_triangle, 1)
--------------------------------------

That way it is clear that G influences what the shape of the triangle is, 
and it splits up the implementation in a nice way.  How about this?

- Shawn