How to get average facet normals for each node?

[conpierce8]

Hi @steve,

Sorry this is a little late, since it seems you’ve already found a solution. But for your awareness, I wanted to point you to the following thread, which seems like it might accomplish what you’re looking for in a way that uses more FEniCS and less numpy: Obtain velocity normal to boundary - #2 by dokken

Regarding the first of your questions below, it does make sense to have multiple DOFs sharing the same coordinate. Since you mentioned elasticity, I assume your dependent variable is displacement (a vector). Thus for a given vertex (i), you will have three displacement components u_x^{(i)}, u_y^{(i)}, u_z^{(i)} at each mesh vertex, and thus three DOFs which all share the same coordinate.

Regarding your second and third questions, unfortunately, I don’t have an answer.

@steve

• When writing get_dofcoords_from_dof_index() , I did not expect that different dof indices would have identical coordinates. Why is that? This is especially strange to me, since I fed the function of unique dofs. A dof can be on multiple facets, I get that, but unique dofs should have unique coordinates. What am I missing here?
• About entity_closure_dofs() :
  • What does this method do, exactly? Why does entity_dofs() not work, for example? The manual did not tell me much.
  • Did I correctly interpret the 2nd argument ( dimension ) as the dimension of the boundary? That is, for a 2D mesh the boundary would be 1D and hence dimension=1 ?
• In the above link to a thread, where dokken posted how to get node coordinates, he goes from cells to facets in order to build up the list of dofs: He uses mesh.topology and dofmap, but
  dofs = V.dofmap().entity_closure_dofs(mesh, dimension, surfacefacetindices)
  does the same thing without all the loops. Did I miss something?