29 lines
1.1 KiB
Markdown
29 lines
1.1 KiB
Markdown
|
# Deprecated: Triangle Library Mesher
|
||
|
|
|
||
|
|
**Status:** Archived (Feb 2026)
|
||
|
|
**Replaced by:** `src/brain/triangulation_gmsh.py` (Gmsh Frontal-Delaunay)
|
||
|
|
|
||
|
|
## Why Deprecated
|
||
|
|
|
||
|
|
The Triangle library approach had these limitations:
|
||
|
|
|
||
|
|
1. **Boundary conformance issues** - Required aggressive post-filtering and manual workarounds for complex boundaries (notches, L-shapes)
|
||
|
|
2. **Iterative refinement** - Needed 3+ passes to achieve density field compliance (slow, complex)
|
||
|
|
3. **Random triangle orientations** - Pure Delaunay doesn't bias toward equilateral/regular patterns
|
||
|
|
4. **PSLG workarounds** - Hole keepouts required manual seeding and validation
|
||
|
|
|
||
|
|
## Replacement: Gmsh Frontal-Delaunay
|
||
|
|
|
||
|
|
Gmsh provides:
|
||
|
|
- ✅ **Single-pass meshing** with background size fields
|
||
|
|
- ✅ **Natural boundary conformance** (advances from boundaries inward)
|
||
|
|
- ✅ **Better triangle quality** (min angles 30-35° vs 25-30°)
|
||
|
|
- ✅ **Cleaner hole handling** via boolean operations
|
||
|
|
- ✅ **Faster** (~1-2 sec vs 1-3 sec for same geometry)
|
||
|
|
|
||
|
|
## Historical Reference
|
||
|
|
|
||
|
|
This code is preserved for reference only. Do not use in production.
|
||
|
|
|
||
|
|
**Last working version:** See commit `5c63d877` (Feb 17, 2026)
|