dolfin team mailing list archive

[Johan Jansson <johanjan@xxxxxxxxxxxxxxxx>]

Hi!

Erik Svensson, Christoffer Cromvik and I had a discussion about how to
compute boundary IDs for edges and faces. We came to the conclusion
that if every node has a list of the boundaries it belongs to, then
that is sufficient to compute the boundary ID for a face.

Let's say we have a corner in a cube where each face of the cube has a
different boundary ID.

D
 \   3
  \
   A------C
 2 |
   |  1
   |
   B

A belongs to 1, 2, 3
B belongs to 1, 2
C belongs to 1, 3
D belongs to 2, 3

The face defined by ACD then has the boundary IDs:
(1, 2, 3), (1, 3), (2, 3)

The boundary ID which occurs in every node of the face is then the
boundary ID of the face. So the boundary ID for face ACD is 3. For
face ABC it's 1 and for face ABD it's 2.

What do you think, does this seem like a workable algorithm?

The XML format would change to this:

<dolfin xmlns:dolfin="http://www.phi.chalmers.se/dolfin/";>
  <mesh>
    <nodes size="8">
      <node name="0" x="0.0" y="0.0" z="0.0">
        <boundaryid name="1"/>
        <boundaryid name="2"/>
      </node>
      <node name="1" x="1.0" y="0.0" z="0.0"/>
...

It's backwards compatible (a node without boundary IDs is defined just
as before), so no changes are necessary for existing meshes.

What is left is the generation of the boundary IDs for faces and
edges, but I think Erik is planning to do that.

  Johan