Skip to content

Gmsh under the hood

This page tells you what apeGmsh delegates to Gmsh and where to learn the Gmsh-level details when you need them — so you know which manual to open when a question goes deeper than this site.

apeGmsh is a wrapper, not a re-implementation. Geometry lives in Gmsh's OpenCASCADE kernel; meshing is Gmsh's meshers behind g.mesh; physical groups are Gmsh physical groups; the .msh format is Gmsh's own. What apeGmsh adds is the structural-FEM workflow on top — names instead of tags, declare-then-resolve physics, the frozen snapshot, the solver bridge. The wrapper is deliberately thin and non-magical: no mesh option is changed behind your back, and the full gmsh.* Python API remains available inside any session for anything without a first-class apeGmsh name.

That thinness means Gmsh's own documentation stays relevant to you:

  • the Gmsh reference manual — every mesh option, algorithm, and file-format detail;
  • the Gmsh Python API — the surface you reach through when apeGmsh doesn't wrap a call;
  • the Gmsh tutorials — meshing concepts (transfinite, fields, partitioning) in their original habitat.

A practical rule for mixed code: do through apeGmsh what apeGmsh names (geometry, sizing, groups, selection — they feed the metadata the rest of the workflow depends on), and reach through to raw gmsh.* only for option tweaks and queries that change no topology. Raw calls that create or delete entities bypass the label registry, and names are what the whole downstream workflow runs on.

Next: Selection & queries.