Separate geometry and dofs/interpolation

[pmueller2]

On the last NuTo meeting we decided to introduce a geometry mesh. The idea is to pull it out of the mesh we have so far and in a next step to transform the remnants (a kind of a dof mesh) to a function space like data structure.

We plan on doing this in several steps to always have working intermediate states.

1) Continuing @vhirtham s work on the Coordinate/Dof Node separation we introduced a GeometryMeshFem which is referenced by the dof mesh. The gmsh import and unit meshes now generate geometry meshes. The dof mesh is still fully functional and does not depend on the new GeometryMesh (apart from the reference) 2) Coordinate related functionality is removed from the dof mesh. 3) and probably also 4., 5. and 6.: Removal of the dof mesh and replacement with an DofInterpolation/FunctionSpace/or similar.

to be continued ... (details, challenges, dof selection by coordinates and such)

[TTitscher]

Will there still be coordinate elements? And the layered ElementCollection?

For non-isoparametric elements, the function space for geometry and dof differs, right? Could this be a problem? Maybe in the application of bcs?

[pmueller2]

• Coordinate Elements: Yes. The GeometryMesh will be very similar to the MeshFem replacing ElementCollections with Coordinate Elements (the mesh may have a method that returns a corresponding function space).

• at least for the moment the FunctionSpace thingy will be a MeshFem with only a single dof and no coordinate elements (but references to them)

• the layered Element Collection will be gone. If you want to have two differing interpolations on a given geometry you have to create two different FunctionSpaces/DofMeshes/YouKnowWhoIam. If they are defined on the same geometry they also have the same number of elements with the same ordering. So selecting boundary elements on the coordinate mesh gives you the corresponding element numbers in the FunctionSpace/DofMesh.

• The transition from the crippled DofMesh to the Function space will be specified later.

[pmueller2]

We think that at this point coordinates and dofs are separated as good as possible without introducing major changes.

• Coordinate Element and DofElement have still the same base class because of the ElementCollection returning coordinate elements via ElementInterface refs which is used in multiple places. Additionally there is no CoordinateElementIGA so far
• The introduction of the function space will be moved to another issue