Helper Utilities

Exodus.jl includes a few higher-level helpers for building connectivity graphs out of raw block data — useful for assembling sparsity patterns, mesh coloring, or other finite-element bookkeeping without re-implementing the traversal yourself.

Element connectivities

Collects the connectivity of every element across every block in the mesh into a flat Vector{Vector{B}} (one entry per element, in block order):

conns = collect_element_connectivities(exo)

Node-to-element connectivity

Inverts element connectivity into, for each node, the list of elements that contain it:

node_to_elem = collect_node_to_element_connectivities(exo)

Element-to-element connectivity

Builds, for each element, the sorted, de-duplicated list of all elements that share at least one node with it (i.e. its node-adjacency neighborhood):

elem_to_elem = collect_element_to_element_connectivities(exo)

Each of these has an in-place ! variant (collect_element_connectivities!, collect_node_to_element_connectivities!, collect_element_to_element_connectivities!) for reuse of pre-allocated buffers in performance-sensitive code.

Parallel numbering helpers

A few additional helpers support stitching together global numbering information from a decomposed mesh (see Parallel / Decomposed Databases for the underlying communication-map types they build on):

Exodus.collect_global_element_and_node_numberings(file_name, n_procs)
Exodus.read_element_cmaps(rank, exo)
Exodus.read_node_cmaps(rank, exo)
Exodus.read_ghost_nodes_and_procs(rank, exo)
Exodus.read_internal_nodes_and_procs(rank, exo)

collect_global_element_and_node_numberings assumes decomp has already been run, producing per-rank files named file_name.n_procs.RRRR. It determines elements' owning processor, and for shared nodes (owned by multiple processors), applies a reduction function (maximum by default) to pick a single owning processor per node.