dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Fri, Jun 25, 2010 at 08:00:58AM -0000, Lars wrote:
> Question #115551 on DOLFIN changed:
> https://answers.launchpad.net/dolfin/+question/115551
>
>     Status: Answered => Open
>
> Lars is still having a problem:
> Thank you, Anders. I am interested to work on the implementation.

Good luck! It will be interesting to see whether you can make it work.

Note that even if you get the mesh to work, other parts of FEniCS
(like FIAT and FFC) need some work to get everything running on hex
meshes.

> I read your paper "Efficient representation of computational meshes", which is a good explanation of the data structures.
>
> Can you recommend more documentation on the MeshEditor concepts and
> implementation, or do I have to study the sources in detail? I think I
> have read somewhere that without the MeshEditor you can construct meshes
> of arbitrary cells - can I find some code examples (c++ or python) how
> to achieve this?

You would probably need to read the source in detail.

What MeshEditor does is that it interacts with the low-level data
structures of the Mesh class (MeshTopology and MeshGeometry).

You probably need to rename some of the overloaded functions in
MeshEditor. In particular add_cell which is overloaded on the number
of arguments (2 arguments is an interval, 3 is a triangle, 4 is a
tet) since a tetrahedron has the same number of vertices as a
quadrilateral. Better names would be add_interval, add_triangle etc.

> And finally, is there any reason why the MeshEditor handles only meshes
> of one single CellType? Why not allow a combination of triangles and
> quadrilaterals, for example, or tets and hexes and wedges in 3D?

It's because the Mesh only has one cell type which is specified by the
CellType* _cell_type member in the Mesh class. This is a limitation
that we might also consider changing, but that would also require a
bit of work.

--
Anders

Attachment: signature.asc
Description: Digital signature