Skip to content

v0.0.2 — convex-corner density fix

Choose a tag to compare

View all tags
@domattioli domattioli released this 25 Apr 02:25
· 152 commits to main since this release
v0.0.2
dd31c79

Distribution: admesh2D on PyPI (deferred; see v0.0.1 notes). Import name: admesh.

What changed since 0.0.1

Notched-rectangle demo now refines the four convex outer corners. Previously the demo passed h0 = boundary_scale = 0.04 to triangulate(...), but the curvature stage drives fh down to ≈0.015 at sharp convex corners. With h0 > min(fh) the initial lattice is too coarse and the rejection method keeps only ~14% of points where fh asks for sub-h0 spacing — thin slivers near the corners.

Demo now samples fh on a coarse grid to estimate min(fh) and uses it as h0, mirroring MATLAB ADmeshRoutine.m:256 where hmin = min(h(:)) is computed after the size field is built.

Quality on the notched-rectangle demo:

  • min_q: 0.165 → 0.682
  • mean_q: 0.891 → 0.985
  • N: 222 → 3426

Install

pip install https://github.com/domattioli/ADMESH/releases/download/v0.0.2/admesh2d-0.0.2-py3-none-any.whl
from admesh.routine import triangulate
from admesh import domains
p, t = triangulate(domains.UNIT_DISK, h0=0.1)

PyPI

admesh2D is reserved on PyPI (the admesh namespace is occupied by an unrelated STL library). Upload requires your PyPI credentials:

python -m twine upload dist/admesh2d-0.0.2-py3-none-any.whl dist/admesh2d-0.0.2.tar.gz

🤖 Generated with Claude Code

Assets 4
Loading