dolfin team mailing list archive

["Garth N. Wells" <gnw20@xxxxxxxxx>]

Anders Logg wrote:
> On Mon, Nov 03, 2008 at 02:44:41PM +0000, Garth N. Wells wrote:
> > Anders Logg wrote:
> > > On Mon, Nov 03, 2008 at 01:04:26PM +0000, Garth N. Wells wrote:
> > > > DOLFIN wrote:
> > > > > One or more new changesets pushed to the primary dolfin repository.
> > > > > A short summary of the last three changesets is included below.
> > > > >
> > > > > changeset:   5080:89eb83718857c7d9dbee0003a50983e28533243e
> > > > > tag:         tip
> > > > > parent:      5079:34987a77383c6b8a37a9c73d40ba4dc38115ca97
> > > > > parent:      5078:1182f3401713488f72e0bc52f8031530bf192bd9
> > > > > user:        Anders Logg <logg@xxxxxxxxx>
> > > > > date:        Mon Nov 03 13:57:42 2008 +0100
> > > > > files:       dolfin/fem/DirichletBC.cpp
> > > > > description:
> > > > > merge
> > > > Beat me to it.
> > > >
> > > > I'm running into an issue with DirichletBC and mixed elements because 
> > > > BoundaryCondition, from which DirichletBC inherits, creates a new 
> > > > FunctionSpace which messes up tests for equivalence and breaks the 
> > > > sharing of function spaces paradigm.
> > > >
> > > > Is there a problem if we remove SubDomain from the argument list for 
> > > > DirichletBC and just pass the corerct FunctionSpace? One can easily 
> > > > extract the necessary sub element FunctionSpace and send it to 
> > > > DirihcletBC. To me this is more intuitive.
> > > >
> > > > Garth
> > > We then need to find a nice way for users to extract the "sub space".
> > > Calling extract_sub_space, getting a pointer and then handling the
> > > deletion of that pointer won't look very nice.
> > Agree. So we remove the SubSystem argument to DirichletBC?
>
> Yes.

Done. Nice to delete almost half the constructors from three classes.

Now I actually understand how DirichletBC works. The new design has made 
what happens in a number of areas much clearer to me.

Garth

> > > How about replacing the SubSystem class with a new class SubSpace
> > > which works just like SubSystem but it also inherits from
> > > FunctionSpace:
> > >
> > >   SubSpace Vu(V, 0);
> > >   SubSpace Vp(V, 1);
> > Looks ok. I'm inclined to keep the name SubSystem because what we're 
> > extracting are not really subspaces.
> >
> > Garth
>
> I also thought about that, but came to the conclusion that in some
> sense it is in fact a subspace. For example, for 2D Stokes Vu above
> will be the subspace of all function which have a zero third component
> (although the third component is not written out).
>
> We could possibly name it SubSystemSpace.
>
>
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/dolfin-dev