change `topology::unstructured::to_polygonal` to `to_polytopal`?

[xjrc]

The new name would more accurately describe what the function does, but it would require updating some downstream libraries. This decision will be important as it will also impact #642 and the final names used for the functions in this PR.

Tagging: @cyrush, @nselliott

[cyrush]

as a bridge, should we have both?

[xjrc]

I think my preference would be skipping the bridge because the code isn't used very many places (I think just my code and potentially @terryhaut 's code), but I'm fairly indifferent. Is there any way we can at least mark to_polygonal as deprecated and have it complain when called?

[cyrush]

@nselliott Q:

For the non mpi lib blueprint, there was only one method to_polygonal that supported both polygons and polyhedra.

For the mpi lib blueprint, there are both to_polygonal and to_polyhedral methods.

What are your thoughts about to_polytopal to unify both cases?

[nselliott]

@cyrush I have no problem with a name change, but perhaps there should be some differentiation between the naming of the non-MPI and MPI versions? They aren't really serial/parallel versions of the same thing.

[cyrush]

to_polygonal case for non-mpi lib is under mesh::topology::unstructured. With this, it is implied to operate on a mesh topology.

The mpi methods operate on the entire mesh.

mpi::mesh::to_polygonal for assumes the input mesh is a 2D structured mesh with adjsets (and amr level info ?). mpi::mesh::to_polyhedral for assumes the input mesh is a 3D structured mesh with adjsets (and amr level info ?).

I am a little confused that these methods don't actually use MPI to communicate?

I think that's a good thing, it signals that the adjsets construct provides enough info, but I also wonder that's just part of the current assumptions.

If we truly don't need to use MPI, we should think about moving these methods into blueprint::mesh

[nselliott]

They do use MPI, for example lines 448-459 in conduit_blueprint_mpi_mesh.cpp

[cyrush]

I see. I was confused by the interface. We need to change these to take a communicator and not assume MPI_COMM_WORLD