New user needs startup documentation

[RedKing7-GA]

I'm a new user interested in using libmesh to read meshes and manipulate them. Where can I find documentation that will walk me through using the various "functions" to do this?

[roystgnr]

Sorry for letting this notification go unnoticed.

There's no simple walkthrough, but the first introduction example creates a Mesh object, reads its data from file, and writes to another file. The meshtool app has a few modification options in between. The MeshModification namespace has a few other modification options encapsulated. For any more custom manipulation you'll want to iterate over nodes and/or elements yourself; see the other introduction example codes for a few iterator ranges. The most up-to-date build of the libMesh Doxygen documentation is probably the one hosted at the MOOSE Framework site.

[RedKing7-GA]

I've taken a brief look at this, but still can't figure out what I need. I just want to read an existing mesh, carve it up into a number of smaller subdomains and read the subdomain descriptions. I didn't see where this could be done with DistributedMesh; I see nothing in the header file that gives me a clue as to how to do this.

[roystgnr]

DistributedMesh vs ReplicatedMesh isn't about how the mesh is carved up on a conceptual level (e.g. here's the subdomain where the solid mechanics get solved, there's the one where the fluid mechanics get solved), but rather on a software level (e.g. here's the partition subset where MPI rank 0 does calculations, there's the subset where MPI rank 1 does). "Distributed" just means that neither processor can see all the mesh (saving memory and usually CPU time at the cost of more complicated code and inter-processor communication).

I'm not sure what you mean by "subdomain descriptions". We support "subdomain names" internally and with a number of file formats, and I guess those can be arbitrary-length strings, but in practice everybody uses names like "fluid" and "solid", not "In this essay I will demonstrate that..." But if those are what you need, each Elem has a subdomain_id() that lets you read and set which integer id associated with a subdomain applies to that element, and a Mesh (look in the MeshBase APIs) lets you read and set a subdomain_name() for each id.

[RedKing7-GA]

Let me clarify.

First, I don't intend to do any type of "mechanics" with the mesh.

I simply want to take an unstructured mesh with lots of elements and divided it into smaller, contiguous meshes. I want to retrieve the node locations, connectivity information, and nearest neighbor information for each of the smaller meshes.

[roystgnr]

Node locations are MeshBase::point(); connectivity is Elem::neighbor_ptr() or Elem::node_ref() depending on which kind you mean. To create a new smaller mesh from a subset of a larger one your best bet is to iterate over the larger one (e.g. MeshBase::element_ptr_range()) and copy (MeshBase::add_point(), MeshBase::add_elem()) into the appropriate smaller one as desired, then MeshBase::prepare_for_use() on each smaller one to reconstruct things like neighbor pointers that can be inferred without copying.