dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Tue, Sep 20, 2011 at 01:25:31PM +0100, Garth N. Wells wrote:
> On 20 September 2011 13:12, Anders Logg <logg@xxxxxxxxx> wrote:
> > On Tue, Sep 20, 2011 at 09:17:25AM +0100, Garth N. Wells wrote:
> >> On 20 September 2011 07:10, Anders Logg <logg@xxxxxxxxx> wrote:
> >> > MeshFunctions? I think MeshValueCollection works but not yet
> >> > MeshFunctions.
> >> >
> >> > My suggestion is to either store MeshFunctions as MeshValueCollections:
> >> >
> >> >  <mesh_function>
> >> >    <mesh_value_collection>
> >> >    </mesh_value_collection>
> >> >  </mesh_function>
> >> >
> >> > (when dim is not 0 or D) or to say that parallel I/O for MeshFunction
> >> > is not supported and make the conversion between MeshFunction and
> >> > MeshValueCollection easy (constructors).
> >> >
> >>
> >> I think that the suggestion to store MeshFunctions as
> >> MeshValueCollections is the best solution. To keep it simple, I would
> >> use MeshValueCollection for entities of any dimension with a tweak of
> >> the tags for vertices.
> >>
> >> The tools are in place to read MeshValueCollections from a separate
> >> file. Just a few small additions are required to make it
> >> work (right now the domain data needs to be in the mesh file). Once
> >> this is in place reading MeshFunctions should be easy. I'll add the
> >> code today.
> >
> > How is the <domain> tag related to reading of MeshValueCollection? It
> > is implemented separately so one should only need to call
> >
> >   XMLMeshValueCollection::read(dim, type, node)
> >
>
> Do you need another reminder that parallel is not the same as serial? ;)

Yes, I need many of those. :-)

--
Anders