mesh generation in rings example errors

[fedeinthemix]

Description

The file examples/rings/mesh/mesh.jl defines the function generate_ring_mesh. Among the keyword parameters there are rot_center, rot_axis and rot_\theta. However, the function body uses rc, ra and \theta instead. Therefore, calling the function errors out.

Even correcting the names, the gemerated mesh doesn't gemerate the same mesh as the one included in palace 0.13.0. In particular, differently from the distributed rings.msh file, the generated one does not include any physical group.

Environment

OS: NixOS 24.05
Palace: 0.13.0
Gmsh: 4.13.1
Julia: 1.11.1
Julia-Gmsh: 0.3.1

[fedeinthemix]

Regarding the physical groups. I did a mistake and did set the mesh file extension to ".mesh" instead of ".msh". After I corrected this, the physical groups do show up. Sorry

[hughcars]

Hi @fedeinthemix,

Thanks for the catch on the variable names, you're right they don't match the arguments, there must've been rebasing issue at some stage. These mesh generation scripts aren't part of CI and are mainly for demonstration of how to create these relatively simplistic geometries in case they're of use to users. Consequently we are at risk of things like this creeping in.

Aside from noting that bug (again, thank you), rerunning the script will not generate an identical file to the original msh file, because the HXT algorithm gmsh.option.setNumber("Mesh.Algorithm3D", 10) is not deterministic. Namely rerunning will give a different answer each time. The suspicion is this relates to their approach to thread parallelism, but this remains just a suspicion. If you want full determinism, algorithm 1 uses a serial Delaunay algorithm which is more predictable (if significantly slower).

I'm not sure why changing the extension like that would alter the presence of physical groups, as the mesh format ought to be fully specified by gmsh.option.setNumber("Mesh.MshFileVersion", 2.2) but I am glad it is working for you now. Feel free to try the fix in #306

[fedeinthemix]

Hi @hughcars

Thanks for the explanation about the meshing algorithm. I'm new to Palace and Gmsh and the mesh generating scripts are very useful.