[petsc-dev] PetscSection

[Matthew Knepley]

On Sat, Nov 10, 2012 at 5:01 PM, Jed Brown <jedbrown at mcs.anl.gov> wrote:
> On Sat, Nov 10, 2012 at 3:49 PM, Matthew Knepley <knepley at gmail.com> wrote:
>>
>> I am not sure I understand. Labels just define point sets. If you want
>> to segregate point
>> sets by dimension, that is fine. Nothing in the interface is sensitive
>> to this, nor should it be.
>
>
> I can think of only one way in which numbering points by stratum versus
> dimension makes a tangible difference: whether contiguous ranges can contain
> unsorted mixed-dimensional points. I want to understand whether this is
> actually important and whether there is a different use case that is
> important. If not, then we could entirely discard the concept of stratum,
> sort points by co/dimension, and query based on labels.
>
> Where is stratum as a concept irreplaceable?

I cannot say off the top of my head. It is very convenient to have the
breadth-first
levels of the graph done this way, and it is also nice for loops, but
I know you are
arguing that people more often write loops over dimension.

>>
>> >>
>> >> You don't need coordinates for interpolation?
>> >
>> >
>> > Are we talking about the same interpolation? If I have cell-to-vertex
>> > connectivity, I can create the faces without coordinates, yes.
>>
>> Ah, we are not. Mesh interpolation.
>>
>> The way I have done interpolation, you have to know the mesh dimension,
>> DMComplexGetDimension(), and then it assumes that height 0 stuff is cells
>> of that dimension.
>
>
> Then we're basically identifying stratum with dimension, suggesting that we
> should be able to remove stratum from the API in favor of co/dimension.

The is one operation that I do not yet know how to do outside of this. This may
be ignorance.

   Matt

--
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which
their experiments lead.
-- Norbert Wiener