Domain refractor proposal: Optimize Mesh Creation Process with GMSH

Issue: Domain Refactoring Proposal – Optimize Mesh Creation Process with GMSH

Description

The current implementation of the domain library does not follow Python best practices. For example, CustomDomain inherits from BuiltInModel, even though it does not require any GMSH features.

We propose a refactoring of the domain submodule, to be introduced in the next major bvpy release (i.e., bvpy 2.0), to enhance coherence and maintainability.

Another inefficiency in bvpy is the current mesh creation process, which writes the mesh to disk twice—first in .msh format and then in .xdmf format. This workflow is suboptimal, as disk operations are time-consuming.

We propose rewriting the AbstractDomain.generate and AbstractDomain.convert_gmsh_to_fenics methods to eliminate unnecessary file I/O. This can be achieved using in-memory data structures and tools such as fe.MeshEditor(), which has already been employed in bvpy in other contexts.^[2]

Objectives of the Refactoring

• Ensure coherence and consistency with Python best practices.
• Optimize the process by avoiding unnecessary file operations. Specifically, eliminate the need for tempfile.TemporaryDirectory.
• Preserve domain attributes, such as GMSH physical groups, by storing them in a MeshFunction object to retain geometry labels.
• Implement changes in a way that facilitates a future transition from FEniCS to FEniCSx with minimal modifications.

References

1. GMSH in FEniCSx
2. fe.MeshEditor() in bvpy

assigned to @grevilla

created branch 130-optimize-mesh-creation-process-with-gmsh to address this issue

added enhancement label

Time improvements:

changed milestone to %BVPy 2.0

closed

changed title from Optimize Mesh Creation Process with GMSH to Domain refractor proposal: Optimize Mesh Creation Process with GMSH

changed the description