[BUG]: #3547 does not handle missing tags for intermediate dimensions

[francesco-ballarin]

Summarize the issue

This line

dolfinx/python/dolfinx/io/gmshio.py

Line 306 in d407cbe

perm_sort = np.argsort(cell_dimensions)

and its usage as

• perm_sort[-2] to set up facet data in

  dolfinx/python/dolfinx/io/gmshio.py

  Line 321 in d407cbe

  num_facet_nodes = comm.bcast(cell_information[perm_sort[-2]]["num_nodes"], root=rank)

• perm_sort[-3] to set up ridge data in

  dolfinx/python/dolfinx/io/gmshio.py

  Line 328 in d407cbe

  num_edge_nodes = comm.bcast(cell_information[perm_sort[-3]]["num_nodes"], root=rank)

• perm_sort[-4] to set up vertex data in

  dolfinx/python/dolfinx/io/gmshio.py

  Line 335 in d407cbe

  num_vertex_nodes = comm.bcast(cell_information[perm_sort[-4]]["num_nodes"], root=rank)

  
  looks fishy after the update in Extract edge tags and vertex tags from gmsh and return a NamedTuple #3547.

For instance, in a 3D mesh with cell and edge data it would happen that perm_sort is an array with two entries, with perm_sort[-1] being associated to cell data and perm_sort[-2] (rather than perm_sort[-3]) associated to ridge data.
In other words, the current code assumes that in a D-dimensional mesh, if tags of dimension d are present, then tags of dimension d + 1, ..., D are also present.

cc @finsberg

How to reproduce the bug

Originally reported in https://fenicsproject.discourse.group/t/kernel-crashes-when-loading-1d-mesh-with-xdmffile/16710/5 , even though the example is far from minimal.

Minimal Example (Python)

No response

Output (Python)

No response

Version

main branch

DOLFINx git commit

No response

Installation

No response

Additional information

No response