N26-295's PT42 pairs saved spacing in slice 1 but every shared edge was
still cut twice, costing a pass per edge and risking a freed part
tipping under the head on the second pass.
Plan Cutting now cuts a plate's common lines once unless the plate is set
to cut them twice. Each group is cut member by member, owner before
partner (a captured dependency that search and replay enforce, so a kept
order cutting the partner first is a constraint conflict): the member's
holes, then the shared cuts it owns as open CommonLine centerline cuts
entered through the notch on the span's axis, then its remainder pieces,
each ramped in from the kerf the shared cut opened. Capture marks the
shared spans as tabs on the owned clean copy, so replay's contour
accounting proves each perimeter is cut exactly once; a lead that fits
at neither end refuses as UnsupportedGeometry naming the span, never
shortened. Material still comes from the drawing, so utilization,
quantity and overlap results do not change. Members are never
support-tabbed; locked or excluded members and broken groups refuse.
The CI Fiber post writes shared cuts in the cut-off form with a lead-in
and no G41/G42, and remainders in the contour form on their perimeter's
compensation side, named by a remainder comment. Cincinnati CL (whose
PreApplied mode offsets nothing) and Gravograph refuse such nests, as do
the desktop and console for any post without ICommonLinePostProcessor;
nothing falls back to a silent double cut. Plate > Arrange > Cut Common
Lines Twice keeps slice 1's behavior, persists with the plate and makes
an open plan stale.
Pre-post review now checks each span is cut once, on its centerline, by
the member cut first, and the rapid check treats a finished member's
remainders as an obstacle. Compensation start in void and cancel in
kerf are unconfirmed on the CI FIBER 4020; release waits on a 3/16
stainless coupon trial.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
An arc lead-in meeting a straight window edge tangentially has a double
root at its joint. Rounding split it into two contacts a few 1e-8 apart,
or dropped it so the exact ray fallback reported the joint as an
uncertain contact. Either refused a valid entry, so N26-27U's windows
lost most of their entries and the plate never planned.
Treat a contact within 1e-6 of a joint its boundary passes through as
the joint, and let the exact line fallback skip ray hits at contacts
already recorded. The pinned-hole test now expects the other hole's
nearest entry rather than the fallback both runs used to share.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A part with about 1,700 holes (P241017-01) never planned: material
capture paired every boundary with every other one and lead checks
scanned every boundary per lead, so both hit the 1M native query limit
and the plate fell back to an unverified best-effort plan.
- Material capture sweeps boundary extents along X and queries only
pairs the lead checks' per-pair clearance cannot skip, in the full
scan's order.
- Material with 64+ boundaries keeps a grid of boundary extents. Lead
checks, contour-joint matching and replay's contour accounting query
only nearby boundaries; ray parity reads only holes on the point's row.
- Only native queries that run are charged to the budget; skipped far
material no longer costs its full boundary count.
- The search looks up chosen holes and route ranks instead of scanning.
With that the plate plans Ready, but in 189 s: every step re-emitted
and re-checked all holes already cut. On parts with 64+ contours a
middle hole step now checks only the contour it adds, continuing the
completed-contour state its prefix left (ReleasedContourState.Continue).
It falls back to the full prefix whenever it cannot continue exactly,
and the last contour still checks the whole part from before it.
Isolated contour emission clones only the chosen contour.
P241017-01 now plans Ready in 5.4 s with an identical route, expansions
and rapid travel. Core and engine suites pass; the problem-nest corpus
matches the unchanged commit's outcomes and expansions.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A lead check scanned every other part's material with native contact
queries. On the 5,742-washer plate that was about 80% of planning and
it never finished inside 120 s.
The earlier extent filter (reverted in ea1270e1) used a fixed 0.001 gap,
which native line/circle rounding can exceed for long leads beside small
circles. This skip adds 1e-6 times the span of both extents. The
discriminant's rounding is bounded by about 1e-7 times that span, and
about 2e-9 is seen. So a skipped material has no native contact, and the
900000-long lead beside a small circle is still checked. Each skipped
scan is charged to the native query budget as if it ran. Ray parity is
still asked natively unless the lead's midpoint is outside the
material's Y range. The pierce zone skips only material farther than its
radius by the same gap.
A randomized test checks that skipped material never has a native
contact. It fails with the span allowance at zero and passes with it as
low as 1e-8. Plate 1 of the washer nest now plans Ready in about 54 s.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Thick plate pierces leave a crater wider than the kerf. CuttingParameters.PierceDiameter
(Cutting Settings > Pierce Diameter, 0 = off) requires every lead-in pierce to keep half
of it from all part material: its own contour, its holes and every other part.
- LeadPathValidator checks the zone during Plan Cutting candidate checks and final
replay; fixed emissions warn as usual. No-fit findings explain the zone.
- PierceZoneReview reports Pierce zone warnings before posting for leads from any source.
- Holes whose configured pierce is inside the zone move it toward the hole's PolyLabel
pole (round holes: the centre), as close to the contour as the zone allows; a hole
too small for the zone is pierced at its pole and the shortfall is still reported.
- Straight corner leads and round-hole clamping use the larger of PierceClearance and
the zone radius. The manual lead-in preview shares the Core rules.
- Persisted through the DTO, owned copies and state fingerprint.
Perimeter fallback is not included; see docs/geometry/pierce-zone.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A converted outline whose seam closes within the admitted 1e-6 gap could
only be entered at its seam: entered anywhere else, the arc after the
seam started off its own circle and every emission was refused as an
inconsistent radius. The search then spent its whole budget on that
part and the plate never planned. Emission now bridges the gap with a
straight move and cuts the arc as drawn; the verifier already admits
that slack. An entry on the seam vertex is left as before.
Each prefix of a part re-checked its earlier contours' leads against
every other part's material. One search now remembers each verdict by
the lead curve's exact bits and the material's reference, charging the
same native query budget; replay never uses it. On the reported plate
the memo alone gives a byte-identical result in 57 s instead of 12 min 47 s;
with the seam fix the plate plans Ready in about 17 s.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Breakoffs are now picked by their joining line, default to the smaller side as scrap
(flippable), and cut that line between tabs on a new Break layer instead of slots.
Plan Cutting and Assign Lead-ins enter each cut from the scrap with the nest's internal
lead-in, or pierce on the line without one; checks treat break cuts as intentional open
cuts that are not material.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Posting review now proves containment exactly as the cutting planner does
(shared NestedPartContainment) and warns when a nested part is cut after its
host (NestedCutOrder) or when a host's hole lead or the nested part's lead
meets the other's material (NestedLead), including subprogram hole poses.
Ambiguous containment is incomplete, never clear; fixed-lead motions are not
reported twice. A committed cutout preview warns until Plan Cutting cuts the
insert first.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
D1a (user, 2026-10-14): one locked cutout must not refuse the whole layout.
With confirmed parameters a locked program is now a fixed emission, accepted
as given. The replay checker and the lead validator take a per-motion owner:
a violation inside the fixed emission (a lead meeting another part's or its
own material, a rapid inside the locked program crossing a cut contour, an
uncertifiable tab gap or a partial/reversed/retraced cut) is recorded apart
instead of failing the candidate. The search never rejects or learns an order
rule from them, so fixed emissions never steer it; the planner's rapid into
the locked program and every regenerated lead keep every strict check.
Replay turns those violations into typed FixedLead findings that name the
part, contour and the part hit. A plate whose only findings are these is
BestEffort with IndependentlyReplayed set: never Ready, applicable only with
the per-proposal acceptance, and CuttingPlanService.Apply still rejects it.
Overlap, unreadable programs, stale input and cancellation still block, and
unlocked programs a caller left ineligible and the no-parameter route keep
refusing as before.
Pre-post review now re-checks locked leads against every part's clean
material, so the same violations reappear before posting.
Existing tests that encoded the old locked refusal now assert the warning;
two batch tests that used a locked crossing as their hard block use an
overlap instead. The 5.1 digests are unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The second delta review compared the filtered checks with the previous
implementation on 5,400 generated cases. Rapid checks matched in every
case, but 114 lead checks differed: a long lead passing a small circle
or arc well over 0.001 away was reported clear, while the native
line/circle query, through rounding in its squared terms, reports a
contact there. A coordinate limit and a fixed margin cannot bound that
cancellation.
Lead checks therefore examine every other part's material again,
exactly as before the filter; LeadMaterialSnapshot no longer keeps an
extent. Rapid checks, including the pre-post review's, keep skipping
completed contours more than 0.001 clear of the rapid. The 900000-long
lead beside a radius-0.0001 circle is a regression test.
Planning a dense 144-part grid now takes about 19 s again (lead checks
dominate); a new part order is still found where the old search gave
up.
The delta review found two more ways the extent filter could skip a
check a native query would have flagged:
- Native line intersections accept points 0.00001 outside each line's
bounding box, so a lead 0.000003 from another part touched it while
their extents were 0.000003 apart, beyond the 1e-6 margin.
- Around a circle of radius 5e11, rounding let the native query count
a rapid at x = -0.00001 as touching although the circle's extent
started at x = 0; the margin scaled only with the rapid's own size.
Rather than chase each tolerance, the filter now has one narrow rule:
Extent.IsClearOf skips only when both extents are finite, lie within
1e6 and are more than 0.001 apart on some axis. 0.001 is ten times the
widest absolute band of any native contact query (the 0.00001 box
allowance and the 0.0001 contact reach of tiny arcs), and within 1e6
rounding stays far below it. Larger geometry is always checked in
full, as before the filter. Planning speed is unchanged (144-part grid
about 0.5-0.8 s).
All three reproductions are regression tests, with a test of the rule
itself.
Every lead was checked against every placed part's material and every
rapid against every completed contour, so each check cost O(parts) and
planning a plate cost O(parts^2). Lead checks were 88% of planning time
on a 144-part grid.
LeadMaterialSnapshot and each completed contour now keep a conservative
extent (an arc counts as its whole supporting circle). A lead or rapid
skips material or a contour only when the extents are farther apart
than 1e-6 x (1 + coordinate size), far above the contact tolerance, so
results are unchanged: anything touching or closer still gets the full
native check. The rapid filter also applies to pre-post verification,
which shares ReleasedContourState.
New tests cover the cases just inside the skip: an arc lead and a
completed arc whose bulge reaches past their endpoints, and a lead and
a rapid that only touch another part's extent.
ajandClaude Opus 5.5