Actually, one problem I just realized is correlating the extra P2 vertices 
with a P2 vector function.  For the kind of stuff I do, I have a velocity 
field which is 2nd order everywhere, and I use it to move my mesh 
vertices.  So, if I have P2 mesh vertices I must ensure that the velocity 
DoFs correspond with the P2 mesh vertices.  I have had no problem with 
this before because I have my own code which is fairly straightforward. 
But there may be a problem in generalizing it.  argh....

So, it seems that if I wanted a P2 mesh, I must create a P2 function that 
lives on that P2 mesh where all the DoF's correspond.  I could then use 
the P2 function to interpolate any velocity field I have onto the P2 mesh 
vertices in order to update the mesh.   This P2 function would, of course, 
contain the vertices/control points for each triangle.

Would that `MeshFunction' be able to do this?

- Shawn

On Wed, 20 Aug 2008, Anders Logg wrote:

> On Wed, Aug 20, 2008 at 03:03:06PM -0400, Shawn Walker wrote:
>> I see that you want to keep the mesh topology separated from the
>> geometry, right?  But, even now, this is slightly violated in the current
>> .xml format, because with each vertex index, you give the coordinates.
>> If all you wanted was the topology, then the coordinates would not be
>> necessary. The coordinates are only need to make sense of the geometry of
>> the mesh (i.e. whether there is overlap of triangles or inverted
>> triangles, etc). So, technically, there should be a list of topological
>> vertices and a separate list of geometric vertices.
>>
>> Is this what you mean?
>>
>> - Shawn
>
> Yes. We might need to break the XML format by having a <topology> tag
> and a <geometry> tag which are separate.
>
> -- 
> Anders
>
_______________________________________________
DOLFIN-dev mailing list
[email protected]
http://www.fenics.org/mailman/listinfo/dolfin-dev

Reply via email to