Clean Geometry Solids

clean-geometry-solids

Diagnose the currently selected polysurfaces/extrusions. Sibling to clean-geometry (curves). One script next to this file, cleangeosol_diagnose.py, via the usual one-line loader. v1 is REPORT-ONLY — reports counts and stashes a plan in sc.sticky["cleangeosol_plan"]; a v2 apply is deferred.

Checks

• A. Open vs closed — Brep.IsSolid → closed. Else count naked edges (Edge.Valence == EdgeAdjacency.Naked) + total naked length, and classify burst (fixable) vs open (intentional) by testing whether JoinNakedEdges(JOIN_TOL) on a duplicate closes it or shrinks the naked count. Burst = the surfaces touch but aren't joined (repairable); intentional = large isolated naked loops.
• B. Extrusion-convertibility — for closed solids: find two parallel planar caps + confirm every side face normal is ⊥ to the sweep direction (prismatic). Right prism → convertible; a sheared cap offset → oblique, flagged NOT convertible (Extrusion.Create is perpendicular-only). Objects already Rhino.Geometry.Extrusion are skipped (already light).

Procedure

1. Resolve path. Glob **/clean-geometry-solids/cleangeosol_diagnose.py. Do NOT hardcode.
2. Diagnose.

   exec(open(r"<ABS>\cleangeosol_diagnose.py").read())
   

   Prints: closed/open/already-extrusion counts; burst-fixable vs intentional-open (with examples: id, naked-edge count, naked length, burst?); extrusion convertible vs oblique.
3. Report. v1 stops here. For a fix now, do it explicitly per object with confirmation.

Gotchas (baked in)

• Selection excludes locked/hidden. Only selectable objects are scanned (e.g. 50 of 340 breps on the test model — the rest were on locked layers). Unlock/show layers to include them.
• Burst test is on a DUPLICATE (DuplicateBrep().JoinNakedEdges) so nothing is modified. JOIN_TOL = 1000×tol is a generous detection threshold — the v2 fix should use a tighter, confirmed value so it never welds truly-separate shells.
• Oblique gate is essential. A sheared prism looks convertible but Extrusion.Create (perpendicular extrude only) would silently build a different solid — so oblique prisms are reported as skip, not convertible. v2 must re-validate any rebuild by volume+centroid before replacing.
• Complex breps skipped for extrusion analysis (Faces.Count > 400) — a prism has few faces.
• EdgeAdjacency.Naked and Surface.TryGetPlane (tuple [0]=bool [1]=plane) are RhinoCommon; Vector3d * Vector3d is the dot product in IronPython.

Not yet implemented (deferred to v2)

• Apply. v1 reports only. v2: JoinNakedEdges(tol) / CapPlanarHoles to close confirmed bursts; rebuild right prisms as Extrusion.Create(profile, signedHeight, cap) (holes via SetOuterProfile
  • AddInnerProfile) with post-convert volume/centroid validation before Objects.Replace.
• Distinguishing intentional-open more finely (paired-naked-edge fraction, planarity of the missing cap).
• Non-brep surfaces / meshes.

Reference

Built 2026-07-01 (task 2 non-curve half of the 6-task suite). Diagnose validated on the live model 260629 main ms.3dm (50 selectable breps → 28 solid / 22 open-intentional / 23 extrusion-convertible / 1 oblique). See memory rhino-clean-rhino-method and the plan.