dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Sat, Jan 02, 2010 at 10:39:33PM +0100, Garth N. Wells wrote:
> I'm not too keen that Mesh::refine() calls AdaptiveObjects::refine(). It
> breaks the nice separation of the Mesh library from other parts of
> DOLFIN. Can we do it differently? Should the user call something like
>
>   AdaptiveObjects.refine(mesh, markers)l
>
> to refine the mesh?
>
> Garth

Ideally, the mesh library should not depend on other parts of DOLFIN.
But I couldn't find any other way to handle this. The problem is that


  mesh.refine()

would otherwise break all other objects depending on it, including all
function spaces and all functions. So mesh.refine() makes little sense
unless one only works with the mesh which I assume is not that
common.

In my opinion, making sure refinement works without producing invalid
dofmaps and function vectors is more important than for the mesh
library not to depend on dolfin/adaptivity. It's also the case that
mesh.refine() is just a one-line convenience function which happens to
be an alias for AdaptiveObjects.refine(mesh, markers).

--
Anders

Attachment: signature.asc
Description: Digital signature