Cannot convert .stp file to .h5m

[ctrossei]

Hello,

Thank you for creating this library!

I have been trying to use it to convert a .stp file to .h5m as described in the examples ( surface_mesh / single_stp_file.py). However it seems the process fails for me because original_ids and scrambled_ids (line 565-566 of core.py) end up having a different size and the comments in the code tell me that this should not be the case.

Unfortunately I cannot share the .stp file I am using, however if there are some things I can try to investigate this issue, please let me know.

[shimwell]

Thanks for letting me know, I shall keep an eye out for similar behaviour

[ctrossei]

So, I think I made it work for my case, although I will admit I am making a lot of guesses of what might have been wrong and how the code works.

I disabled all checks regarding the size of the original_ids, scrambled_ids and material_tags / material_tags_in_brep_order. The meshing phase (gmsh, volumes = _mesh_brep(...) went well, and generated 4266 faces and 770955 nodes (vertices?).
The code went on to the _vertices_to_h5m part and spent a lot of time in there, the memory usage increased to 30 GB until the process eventually segfaulted. Looking at the code in the _vertices_to_h5m function, it looked like lines 145 and 146

moab_verts = moab_core.create_vertices(vertices)
moab_core.add_entity(face_set, moab_verts)

were adding the same vertices repeatedly to the moab_core object (because it is inside two for loops), and so I modified the function for this not to happen, ie I commented out lines 145 and 146 and added moab_verts = moab_core.create_vertices(vertices) line 105 right after the moab_core object definition. I did not add back the moab_core.add_entity(face_set, moab_verts) line, however.

After this change the _vertices_to_h5m very quickly generated an .h5m file that I mbconverted to a.vtk that looks like my original step file at first glance.

[shimwell]

Nice work. The other thing I occasionally try is to manually mesh with gmsh

manually meshing the step file gives good control over the mesh parameters

1 load up gmsh GUI
2 load in the step files
3 export a msh file (2d)

then continue from line 21 of this example
https://github.com/fusion-energy/cad_to_dagmc/blob/main/examples/surface_mesh/from_gmsh_mesh_file.py

[shimwell]

I wonder if the step file has something inside that I've not considered. Some sort of nested comp solid or something else,

original_ids and scrambled_ids should be the same length so I guess something is getting expanded in the process