Files
OpenNest/docs/geometry/lead-in-placement.md
T
aj 7127884584 fix(cutting): start tabbed lead-outs at the trimmed cut end
A tab trims the perimeter short of its entry, but the lead-out was still
generated from the nominal entry point. An arc lead-out therefore started
off its own radius (ExecutionMotionReader rejected it as inconsistent), and
a line lead-out ran diagonally back toward the entry.

Every lead-out style on a tabbed perimeter now leaves from the trimmed
cut's actual end, on that entity's normal, so arcs are tangent and the tab
gap stays uncut. Untabbed contours and the corner run-out rules are
unchanged. Malformed legacy output is still refused, never refit.

Red before the fix: the three tabbed arc cases threw "Arc has zero or
inconsistent radius" and the line case ended at y=5 instead of 4.8. Keeping
the entry's normal at the actual end fails the curved-perimeter case.
2026-10-04 23:13:41 -04:00

4.7 KiB
Raw Blame History

Lead-in placement at corners

Straight (LineLeadIn) lead-ins at closed internal contour corners use the inward angle bisector instead of the normal of whichever edge was selected. With the default 90-degree approach angle, a rectangular cutout therefore gets a diagonal lead-in into the scrap, rather than one lying along the other edge. This applies to automatic assignment and manual placement. The manual preview uses the same Core calculation, including when the lead-in style changes while the cursor is stationary.

ContourCuttingStrategy.ComputeLeadInNormal combines the two adjacent inward unit normals. It handles either winding, either selected edge, rotated parts, and line/arc junctions using their local normals. Summing vectors avoids angle wraparound and edge-length weighting. Open contours, disconnected endpoints, zero-length edges, and cusps without a unique bisector retain the entity normal.

The configured approach-angle offset is still applied relative to the computed normal; 90 degrees follows the bisector. Lead-in length is unchanged. Mid-edge points, external contours, circles, curved/composite lead-in styles, and lead-outs keep their existing placement rules.

This is a local direction correction, not a whole-path clearance guarantee. An excessively long lead-in or an approach angle rotated away from the bisector can still leave a small cutout. General non-circular containment/length clamping and sharp-corner handling for curved/composite lead-ins remain separate work.

Outside perimeter corners

A straight (LineLeadIn) lead-in at a convex corner of the outside perimeter (interior angle under 180 degrees) extends the edge cut first: the pierce sits on that edge's line, behind the corner, and the torch travels straight into the corner and keeps cutting along the same line. Which of the two edges was picked (auto assignment or the manual cursor) does not matter; the cut direction never changes. The approach angle is ignored at such a corner.

The straight lead is used only while its pierce stays at least CuttingParameters.PierceClearance from the contour and the lead crosses the contour nowhere but at the corner. Very flat corners (about 165 degrees and over for a 0.25 lead with 0.0625 clearance) fall back to the normal lead-in, perpendicular to the first-cut edge, so tessellated curves do not get straight leads. Reflex perimeter corners (the inside corner of an L) use the notch bisector, like cutout corners.

A LineLeadOut mirrors this: at a convex perimeter corner it runs straight on past the corner along the last-cut edge, with the same clearance fallback to the last-cut edge's normal and a bisector at reflex corners. A tabbed perimeter gets no run-out: every lead-out style leaves from where the trimmed cut actually ends, on that end's normal, so an arc lead-out starts on its own radius and the tab gap stays uncut. ContourCuttingStrategy.ResolveLeadIn/ResolveLeadOut own these rules; program generation and the manual preview share them. Other lead-in styles are unchanged.

Regression checks

Run dotnet test OpenNest.Tests/OpenNest.Tests.csproj --filter "FullyQualifiedName~CutoutCornerLeadInTests|FullyQualifiedName~PerimeterCornerLeadInTests". The cutout tests exercise generated part programs, default automatic placement, every rectangular corner with both adjoining edges and windings, part rotation, acute and obtuse angles, reflex corners, line/arc corners, preview agreement, and unchanged/fallback behavior. The perimeter tests cover every square corner under both windings and a rotation, auto and manual placement from either edge, preview agreement, approach-angle handling, the flat-corner clearance fallback, reflex notches, and straight lead-outs.

A headless before/after import of 4980 A01 PT07.dxf (SHA-256 1535D77BC1EEEDD21A27E7CE91EA4C51055118D019C5A09C144F1F41740895B6) reproduced the issue on all five 0.282 × 0.532 rectangular cutouts. At the default 0.125 lead-in length, all five corrected pierce points are approximately 0.08838835 inside both adjacent edges. The straight segments stay inside the rectangles except for their contour endpoints; the other two generated lead-ins remain unchanged. Coordinates were compared with a 1e-8 tolerance to allow incremental program round-trip floating-point noise. The source drawing is not bundled.

Windows manual acceptance: import the part and use Plate > Assign Lead-ins. Confirm diagonal lead-ins at all five small rectangular cutouts. Under Plate > Place Lead-in, select the part, lock a cutout, and hover a corner: the preview should point into the cutout and the committed lead-in should match. Hover a mid-edge point and confirm perpendicular placement is unchanged. Windows visual interaction is not verified by the Linux cross-build.