dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Tue, Dec 04, 2007 at 07:55:31PM +0100, Kristian Oelgaard wrote:

> Hi,
> 
> I'm having a few difficulties with 1D elements in DOLFIN, more specifically with
> the assembly over exterior integrals and when applying boundary conditions.
> 
> In Assembler::assembleExteriorFacets(), a boundary mesh is being created. This
> fails when the mesh consists of intervals. It does so in CellType where only
> Interval, Triangle and Tetraheron are available.
> 
> I successfully implemented a special case in Assembler::assembleExteriorFacets()
> if (mesh.topology().dim() == 1){//handle 1D}, but is this what we want and could
> it be done more naturally?

I think we can add a new cell type for 0-D cells. The natural thing
would be to call it Point, but it's already used in DOLFIN for
something else. Maybe we should rename Triangle to TriangleCell etc
since a Triangle is not really a Triangle. It's just a collection of
functions on Triangle cells. Then we can have PointCell, IntervalCell,
TriangleCell, TetrahedronCell.

> The second problem (BCs) is due to how VertexIterator is defined.
> Using the 'topological' (default) or 'geometrical' method the following simple
> SubDomain will fail:
> 
> class DirichletBoundary : public SubDomain
> {
>   bool inside(const real* x, bool on_boundary) const
>   {
>     return std::abs(x[0]) < DOLFIN_EPS && on_boundary;
>   }
> };
> 
> When a facet with index 0 (this is the facet I want) is used to generate a
> VertexIterator it will look at the connectivity and return the vertex with index
> 1 on a standard mesh. A VertexIterator generated from the facet with index 1
> will iterate over vertices 0 and 2. This destroys the logic of the geometrical
> check in the inside() function.

I don't understand this. Which mesh is this on?

-- 
Anders


> However, using the 'pointwise' method the above SubDomain works as expected.
> 
> Any thoughts on where these problems can be tackled most efficiently?
> 
> Kristian
> 
> 
> 
> 
> 
> 
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/dolfin-dev