dolfin team mailing list archive

[Bartosz Sawicki <sawickib@xxxxxxxxxxxxx>]

> Would it be simple to implement this so that it works also for uniform
> refinement?

Uniform refinement is done by refine_cell method implemented separately 
for different types of elements. So mappings between old and new mesh 
should be done in the same level. It should be quite strait forward to do.


> It would also be practical to be able to transfer mesh functions of
> all kinds to the refined mesh (at least for cells and vertices), not
> just the one named "material indicators".

That's exactly what I was planning at the beginning - whole MeshData 
structure should follow refinement. But it is impossible to define 
general transformations for arrays, maps, if we don't their meaning.
Update for vertices MeshFunction vertices is hard to define too (new 
vertex should inherit value from closest neighbor? what if it is exactly 
in the middle between them?)

We can easily change the code to manipulate every cell based 
MeshFunction, not only the one named "material indicators". But IMHO it 
would be inconsistent with boundary conditions, where only arrays named 
"bondary indicators", "facet numbers", and "cell numbers" can be 
transformed. That is why I decided to update only specified components 
of MeshData, the ones which meaning is known.

cheers,