dolfin team mailing list archive

["Garth N. Wells" <g.n.wells@xxxxxxxxxx>]

Anders Logg wrote:
> On Wed, Dec 06, 2006 at 11:39:44PM +0100, Garth N. Wells wrote:
> > Anders Logg wrote:
> > > Do you mean "DofMap associated with mesh and finite element"?
> > Yep.
> >
> > > I'm ok with that. We could have a DiscreteFunction with two members:
> > >
> > >     Vector
> > >     DofMap
> > That would work.
>
> On second thought, I'm not sure this would work. The current class
> FiniteElement has a few other functions than what is accomplished by
> the DofMap. We have the value type for the function space (scalar,
> vector, tensor) and we have evaluation at vertices with vertexeval()
> which is used for output at vertices. So just the DofMap is not
> enough.
>
> I know the DofMap has a FiniteElement that can be accessed, but the
> FiniteElement is more central. The DofMap is just a way to reorder the
> dofs for efficiency.
>
> How about letting
>
>     Function = (vector, mesh, element, dofmap = 0)
>
> so that the dofmap is optional (the pointer migth be 0). If it is
> nonzero, the implementation must look at the dofmap when it does
> interpolation.

It is nicer having just

  Function = (vector, mesh, element)

The danger is that it offers more possibilities for mistakes -  dofs 
might be remapped without a Function knowing. It's still sometime away, 
so we don't have to settle on anything just yet.

Garth

> ?
>
> /Anders