Nests, drawings and releases store material and grade as free text, so
there was no canonical list for clients to suggest from or to fix
spelling against. Schema v10 adds materials and material_grades keyed by
SharedListNames.Key, with GET/POST /api/materials. Adding is idempotent
by key and keeps the first spelling.
The tables start empty: the materials already stored did not import
cleanly, so the list is filled through the API instead. Existing
free-text columns are unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A release now ends in nests ready to cut rather than empty drafts, and a
merge no longer leaves the draft it cleared waiting for someone to run Auto
Nest. "Nest after release" (on by default) runs Auto Nest headless on every
released group, new or merged, before anything is uploaded.
NestRenest removes every plate and nests each drawing's required quantity
onto sheets from the nest's plate defaults and plate-size options. It keeps
a layout only when it is valid and keepable, so an invalid result never
reaches the server; the nest is saved without sheets and the operator is
told to run Auto Nest. Unplaced copies stay as remaining quantity, and the
status column lists sheets, utilization and what did not fit. A new Nesting
progress stage reports each group, and cancelling while nesting saves
nothing.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The release dialog now lists the sets to release instead of one set with a
multiplier. More are added by typing a name (the matching active sets fill
the list as you type) and each set has its own kit count. Its groups span
all the sets, and each group can make a new Draft nest or merge into a Draft
nest of the same material, grade and thickness, listed with its sheet count.
The status column says that a merge discards the draft's layout and that
Undo Release restores it.
A draft open in a tab of this desktop is refused, because nest tabs do not
track unsaved edits and the merge would leave them behind. The main window
tells the dialog which server nests are open.
The release history shows what each release did to each nest (created, or
merged from one revision to another, and what undo did) and gains Undo
Release. When nests changed after the release, it names them and offers to
undo the others, leaving those as they are.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The release plan now takes several sets, each with its own kit count. A
drawing listed by more than one set becomes one item with the quantities
summed and each set's share kept, so the release can record what every kit
put into every nest. FindDraftsAsync lists the Draft nests of a group's
material through the new nest query filters.
The session can merge a group into a chosen draft instead of making a new
nest. It downloads the draft at the revision it was chosen at, refuses it
unless it is still a Draft of the group's material, thickness, units and cut
type, adds the items, and removes every plate so the whole demand is nested
again. The save is guarded by that revision, so a draft someone saved
meanwhile fails its group instead of being overwritten. An optional renest
step, run before anything is uploaded, will back "Nest after release"; if it
throws, the nest is still released without a layout.
The release record now carries every set line, merged nests with their
revisions before and after, and each set's drawing quantities.
SetReleaseText.UndoRefusal words a refused undo for the release history.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A release that merges into an existing draft needs the drafts of the same
material group. The nest query can now filter exactly on status, material,
grade and thickness. Material and grade compare like the shared material
lists (case and spacing ignored) through a SQLite function, so no key columns
or backfill are needed; a blank grade asks for nests without one, and
thickness matches at the four decimals drawings are stored with.
The client sends the filters only when set, so plain browsing keeps working
against servers that do not know them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A release can now cover several sets, each with its own kit count, so parts
from different kits share sheets. It can also merge into an existing draft
nest instead of always creating one. Schema v9 replaces the single-set
release tables: releases, their set lines, their nests (created or merged,
with the nest revision before and after) and the drawing quantities each set
added to each nest, so a kit stays traceable once kits share a nest. The v8
migration turns each existing release into one set line with created nests.
Undo runs on the server in one transaction: created nests are soft-deleted
and merged nests restored to their pre-merge revision, but only while each
still holds what the release wrote. Otherwise nothing changes and the
changed nests are listed, so the operator can skip them. A nest restored to
that revision counts as unchanged, which lets releases sharing a nest be
undone newest first.
The old /api/sets/{id}/releases and /api/nests/{id}/release routes are
replaced by /api/releases; no tagged desktop used them. The desktop still
releases one set at a time until the release dialog is reworked.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Adding drawings to a set meant opening the library dialog, then the set
editor, then a second copy of the library in pick mode, then backing out
again. The set editor now has the library as its left pane (unfiltered,
searchable, sortable, paged): select drawings and Add to Set, double-click
or press Enter. Members already in the set are greyed. The library
dialog's pick mode had no other caller and is removed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The drawing set fixtures landed on master while uploads still accepted
any bytes; the server now rejects a file without a readable nest.json,
so the fixture's placeholder payload made every release test that
posts a nest fail with 400.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Write always stamped DateLastModified with the save time. That is right
for an ordinary save, but a converter carrying another system's date
cannot keep it: every PEP nest imported into the server recorded the
import time as its modified date.
PreserveDateLastModified (init, default false) writes the nest's own
date instead; every existing caller keeps stamping.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The server stored whatever plate count, part count, material and
thickness the client claimed, so a record could disagree with its own
archive, and a bulk importer had to read each written .nest back in full
just to describe it (about a quarter of the PEP archive import's time).
Uploads (POST and PUT /file) now read only the archive's nest.json -
never programs or geometry, and without referencing OpenNest.IO - and
overwrite those fields, following NestReader's fallbacks for thickness,
material, parts with a missing drawing and the trailing new-plate slot.
A file without a readable nest.json is 400 and stores nothing, so the
server can guarantee the record describes the bytes; nest.json
decompression stops at 64 MB. A metadata-only PUT keeps the stored
counts because the archive is unchanged; material and thickness stay
editable there, since the desktop applies the record's values on open.
Server tests that uploaded placeholder bytes now upload small
deterministic archives, and the container smoke's near-limit probe is a
padded real archive. New tests pin the server's values to a full
NestReader read of the same archive.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ToPolygon() used 1000 segments for every arc and circle regardless of
size: no more accurate for large arcs and needlessly dense for small
ones, which made every Drawing area and canonical-angle computation on
an arc-heavy part handle hundreds of thousands of vertices.
The parameterless overload now delegates to ToPolygonWithTolerance with
DefaultChordTolerance = 0.0001. That is ten times tighter than the
spacing and overlap checks' 0.001 because callers use this polygon for
area, which an inscribed polygon always under-reads: at 0.001, seven
area-sensitive cut-off, splitter and circle-packing tests drifted past
their precision; at 0.0001 all pass. ToPolygon(int) keeps the fixed
segment count for any caller that wants it.
With the one-pass Cleanup, a 300-nest PEP archive import on the forge
server went from 46.1 s to 15.5 s.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Cleanup removed each duplicate vertex with List.RemoveAt from the middle
of the list, so a polygon cost O(duplicates x vertices). Shape.ToPolygon
emits a duplicate at every entity junction and flattened arcs to 1000
points, so building the Drawing for a large arc-heavy part spent seconds
shifting memory: one 1 MB PEP nest took 12.7 s to convert and 11.2 s to
read back, a quarter of a 300-nest archive import's total time.
The single pass compares each vertex with its original predecessor, as
the back-to-front loop did, so tolerance-equal runs collapse identically
and the output is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Plan Cutting is now the only automatic lead-in assignment, and parts the
support grid cannot hold level are reported wherever they lack a
steadying tab.
- Remove Assign Lead-ins (plate and nest), LeadInAssigner, the lead-in
tool's Auto-Assign button and the Auto Tab size range; old settings
with autoTab sizes still load
- Manual Place Lead-in tabs the clicked entry when support tabs are on
and the grid cannot hold the part (SupportTabSettings)
- New PartSupport finding: Plan Cutting replay and best-effort plans
warn when a drawing pin, a lead-in that does not fit or no single tab
leaves a part unsteady; the pre-post review checks each placed
program's outside tab gap. These warnings replace the advisory
weak-tab notes; only the inside-corner tab note stays advisory, and
never for a pin
- Pinned perimeters keep their pin but carry the support analysis
- Draw grid points as pixel-snapped squares without anti-aliasing
- Post fixtures build lead-ins through a test-only TestLeadIns helper
- Docs: support tabs, pre-post verification and planner notes
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A tab cannot lift a part's margin past its contacts' own hull edges, so a
whole stretch of outline often ties for the best tab point, and the first
sample found won, leaving the tab off-center. Keep the full outline search
for the margin, then among samples within Tolerance.Epsilon of the best,
pick the one nearest the line through the center of gravity across the
long side of the outline's minimum-area bounding rectangle. A part held
only near one end is still tabbed at the far end.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Nest records gain a materialGrade field (server schema v8, blank for existing
rows) that is searched, sorted and versioned like the other metadata. The
Nest Info dialog gets a grade picker listing the machine configs' grades for
the chosen material; density survives only an unchanged name and grade. A
blank recorded grade keeps the archive's grade while the name matches, so
records saved before v8 do not lose it on open. Saved Nests shows a Grade column.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Replace the plain-text report with a colored status banner, per-check
count tiles and a findings list grouped by check type, with the selected
finding shown in full below it. A Copy Report button keeps the full text
report available. Consent, cancellation and failure blocking are unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Saved best-fit pairs are in the drawing's canonical frame, but the canonical
angle is recomputed on load and was not saved with the set. When that
computation changes (RotatingCalipers checking every hull edge, marks no
longer counted as material), old pairs were replayed in a turned frame and
could overlap.
Each saved set now records canonicalAngle, and BestFitCacheStore.Import
skips sets whose angle no longer matches the recomputed one (1e-9 rad,
modulo a full turn). Sets from older files have no angle and are dropped;
their pairs are recomputed when next needed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
TryImprove kept a swap when the trial run looked fuller than before, but
the trial charges each type's self-nesting gain again even when the run
already holds that type. Trading a placed copy for an identical unplaced
one therefore looked fuller by the gain and repeated forever: FlatBar ran
into the benchmark's 5-minute timeout on tabbed_cog_layout once the
all-edges MBR fix gave one part new rotations (it finished in 7 s before).
The swap is now judged on the merged run, so every kept swap really
leaves the run fuller and the loop terminates. Swaps the old test kept
that do fill the run are still kept.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MinimumBoundingRectangle(Polygon) evaluated only the first (n + 1) / 2
edges, assuming opposite edges give the same rectangle. That holds only
for centrally symmetric hulls: for the triangle (0,0), (10,0), (12,1) it
returned area 12 instead of 10, depending on vertex order.
Every edge is now a candidate. A second-half edge replaces the best only
when smaller beyond rounding, so symmetric hulls (rectangles, best-fit
pairs, whose MBR angle PairEvaluator uses directly) keep their previous
orientation instead of flipping 180 degrees on floating-point noise.
On the Desktop problem nests, 8 of 160 drawing entries now get a smaller
MBR (0.004% to 3.2%) and a turned canonical frame; none changed without a
real area gain. The problem-nest corpus is unchanged against f51e5aef.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Cut-off lines are drawn after the parts in their own red dashed pen, so they
stayed red over their muted parts. Draw them gray while the torch has not
started their part.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The cut planner status shows the run's elapsed time, and its inputs stay
enabled while planning: changing one cancels the run and starts a new one.
Cut-off settings open once the cancelled run releases the plates.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
While a Sequence by Path line is being dragged, the plate view shows the
numbers the finished line would give, without reordering the plate, so
reorder listeners (drawing list, overlap overlay, simulator) are not hit
on every mouse move. The plate is reordered once on release.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Plate > Sequence Parts > Sequence by Path: each left-drag adds a line,
cut in drawing order; the jump between lines is a rapid and claims no
parts. Each part belongs to the line passing closest to its center and
is ordered by where that nearest point falls along the line. Cut-first
parts stay ahead and cutoff dependency rules still apply. A small window
offers Undo Line, Clear (restores the starting order) and Done.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The best tab point used to land in inside corners: equally good points
run along a straight edge, vertices were tried first and the first tie
won. Operators cannot grind a nub flush there.
- SupportAnalyzer finds inside corners (inward-turning vertex runs of at
least 30 degrees, so chorded fillets count) and keeps tab points a
clearance from them; of equally good points it picks the one farthest
from an inside corner. Plan Cutting and the overlay use twice the tab
size; AcceptsTabAt rejects closer points
- When no steadying entry has a lead-in that fits, the search falls back
to entries clear of inside corners, then to every entry, each ordered
by how well the tab steadies the part, keeping only the steadiest that
fit
- Replay notes parts whose planned tab cannot hold them level or sits in
an inside corner; the notes are advisory and leave a Ready plan Ready
- Console/MCP plate summaries print a ready plan's notes
- docs/support-tabs.md and cutting-planner notes
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Model the machine's support bed as a grid of contact points measured
from the plate origin, and check each part's center of gravity against
the convex hull of the grid points under its material. The bed is
extensible so straight and CAD-imported curved slats can follow.
- Core: SupportBed/SupportGrid, SupportMaterial and SupportAnalyzer
(margin with registration tolerance, best single tab point)
- Nest-wide support bed that every plate follows, saved in
plateDefaults.supportBed; machine defaults in support-grid.json
- CuttingParameters.SupportTabs: Plan Cutting tabs only unsupported
parts and limits their perimeter entries to points where the tab
steadies the part, falling back to any fitting entry
- Desktop: Nest > Support Grid..., View > Draw Support overlay and a
"Tab parts the support grid cannot hold level" tab option
- docs/support-tabs.md, nest file format and planner notes
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Operators often build a nest first and only then want it kept as a
reusable kit. The drawing list's new "Add Library Drawings to Set..."
takes every drawing copied from this server's library, sums copies of
one library drawing, and opens the set editor on a chosen or new set
for review and Save. Unlinked, other-server and zero-quantity drawings
are named as skipped instead of silently dropped.
Members the set already lists keep the set's quantity: a nest released
from a set carries its kit multiplier, so copying its quantities back
would scale the kit on every round trip. Members are added at the
library's latest revision (S1), with a note when the nest holds an
older one or the drawing is archived.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A set is only useful on the floor once it becomes nests. Release on the
Sets tab previews one row per material group with a kit multiplier, an
archived-drawing toggle and a tick per group, then builds each ticked
group's nest as File > New would, in the group's units, uploads it as a
Draft, opens it in its own tab bound to the server record and records the
release. Every group's outcome stays on screen, and a release whose
nests exist but whose record failed can be recorded again.
Releases shows each release's nests and how far they have got ("2 of 3
cut"), and Nest Info names the release a nest came from, so a kit run can
be followed across its nests. Plate defaults convert between inches and
millimeters so a group drawn in other units still gets sensible plates.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Operators need to build kits from library drawings before they can be
released. The Drawing Library dialog gains a Sets tab beside Drawings,
so sets live where their drawings are browsed: list, search, archive,
restore, history, and an editor whose Add Drawings reuses the library in
a multi-select pick mode. The Drawings tab's menu shows which sets use a
drawing and adds it to a chosen or new set.
DrawingSetEditSession keeps a new set's id across a save whose reply was
lost, so a retry adopts the stored set instead of creating it twice, and
guards existing-set saves with the revision they were read at.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The desktop needs a cross-platform way to read and save sets and to turn
one into nests before any UI is built on it. RemoteDrawingSetStore covers
every /api/sets route with the library client's checks and conflict
handling. SetReleasePlan groups a set by material, grade, thickness, units
and cut type without I/O, so the release dialog can preview refusals
(duplicate names, drawings without material) before anything is created.
SetReleaseSession builds every group's nest through LibraryDrawingInsert
before uploading any, so a refusal leaves nothing on the server; uploads
then never cancel mid-flight, a lost reply is reconciled by the nest's
client-allocated id, and the release is posted only when every chosen
nest exists, with a same-id retry when that post fails.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Releasing a drawing set makes one nest per material group, so a kit run
spans several nests. The server now records which set revision was
released, the multiplier and the nests it made, so a run can be tracked
across its nests and a nest can say where it came from.
The client uploads every nest first and posts the release last: a failure
part-way leaves ordinary nests and no release, never a release pointing at
missing nests. For the same reason a release is accepted for an older set
revision or an archived set, since the nests already exist by then. The
release id is client-allocated and a nest belongs to at most one release,
so a retried post gets 409 with the release that already holds it.
Releases are append-only and read joined to each nest's current metadata,
so status and deletion are always current. The tables join schema v7, as
the set tables have not shipped yet.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A drawing set is a named list of library drawings with quantities (a kit,
an assembly, a recurring order) that will later be released into one nest
per material. This is the server half: sets get the same revision,
If-Match and archive contract as /api/drawings so the client and tests
follow a known shape.
Schema v7 adds drawing_sets, drawing_set_revisions and
drawing_set_members. Members reference drawings by id and mean each
drawing's current revision, so reads join the current drawing record:
the client can group members by material without one request per drawing,
and the joined record is ignored on save so a read record can be sent
back unchanged. Every change keeps its own member list, and
/api/drawings/{id}/sets answers "used in" from current revisions only.
The drawing query's enum, integer and history-paging parsing is now shared
so both APIs refuse bad input identically.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Nests released from drawing sets need a state before Quote: created but not
yet nested or checked. Draft is appended to NestStatus (value 3) because the
server still accepts integer statuses from existing callers, so existing
numbers must not shift; NestStatuses.WorkflowOrder puts it first wherever the
operator picks a status.
Older desktops read Draft as Quote (the documented unknown-status fallback),
and an older server refuses "draft", so the server must be upgraded before
desktops that can set it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The pallet-change setting and post doc said M50 moves the cut sheet out,
borrowed from the CL-series manual. Cincinnati's CNC Codes Handbook calls
M50 "Table In" and M51 "Table Out". On the CI Fiber shuttle, Table In brings
the waiting pallet into the laser, which sends the cut one out, so M50 is
still the swap. Say that, note that it expects the waiting pallet to be
loaded, and note that newer Cincinnati samples end without a pallet code.
Default (M50) and posted output are unchanged.
The Cincinnati sample used by the CI Fiber tests is a customer production
part. Its geometry and NC were never tracked, but comments named its part
number and fixture file names and quoted its date and a pierce coordinate.
Replace them with generic wording and placeholder values; the fixture stays
outside the repository, configured only in the gitignored test-config.json.
Comment-only change.
The sample regression fixture now carries each part's pierce, lead-in and
cut moves from the Cincinnati machine program as its saved cutting program.
A new test posts those saved programs without reassigning lead-ins and
requires the program body to equal the sample exactly: L2/L4 macro, G41/G42
side, lead-ins and every cut move (21,911 lines).
Mutation checks: dropping the largest-contour perimeter rule from the kerf
fix fails it at the first perimeter (L4 posted as L2), and forcing G41 fails
it at the first perimeter's G42. The existing tests still replace the saved
programs with OpenNest's own lead-ins first; their comment now says so
instead of describing the sample's leads wrongly.
The CI Fiber post selected G42 for every perimeter. Cincinnati's CNC Codes
Handbook defines G42 as "product remains on the left", and imported
perimeters run clockwise with the part on the right, so the controller
offset every outside edge into the part: parts came out a kerf undersize.
Cincinnati's LaserNST samples only use G42 because they cut perimeters
counter-clockwise.
Winding is normalized only at DXF import, so the post can't assume a
direction either. It now takes the largest contour as the perimeter (as
ShapeProfile does) and picks G41 or G42 from each contour's actual winding.
That also stops a reversed drawing from swapping the L2/L4 lead-in macros.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The safety raise (M47 P<SafetyHeadraiseDistance>) used to fire only when
the next part had a different drawing name, which has nothing to do with
crash risk: a long rapid between copies of the same part got a plain M47,
while a short hop between neighbouring different drawings got the full
raise. Sheets are rarely flat (5/8" plate can be ~1.5" out of flat within
supplier tolerance), and longer rapids cross more of that bow.
Add SafetyHeadraiseRapidDistance (program units, default 12"): any rapid to
the next pierce longer than it takes the safety raise, regardless of
drawing. Unknown travel never does.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Run_NeverExceedsMaxParallelism_ButReachesIt asserted two 200 ms solves
overlapped, which depends on the thread pool adding a second worker in
time; hosted Windows release run 39400265218 saw only one. Each fake solve
now waits (up to 10 s) until two solves have entered, so a parallel runner
always reaches the limit and a serial one fails once after the timeout.
Forcing MaxDegreeOfParallelism = 1 fails it (Expected 2, Actual 1).
Project Memory b534b51c.
Console --plan-cutting and MCP plan_cutting run the planner at the default
effort; only the desktop exposes the setting. Adding a corpus nest must keep
the corpus server selected until the save, and the command is Nest > Edit.
Tools > Options > Cut planning effort (default 1.0, 0.5-10) multiplies the
desktop cut planner's per-plate expansion budget for both the free-order
search and the current-order retry. The budget counts expansions, not time,
so the setting is how faster hardware can be spent on hard plates.
The panel reads the setting on the UI thread and passes it to
CuttingPlanBatch.Capture; the worker never reads settings. The 1,000-
expansion first pass stays fixed: it is a cheap probe for easy plates, so a
higher effort only adds work where the full search runs. A plate refused
at the search limit now says the effort can be raised, without promising a
plan. The default keeps today's budgets and characterization digests.
Console, MCP and API do not run the cut planner, so they get no option.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Stopping a fill only requests cancellation; its worker keeps reading
Plate.Parts until it returns. The progress window is modeless, so a
click or key on the plate view in that window (clone placement, Delete,
arrow nudges, drops) changed the part list mid-enumeration and threw
"Collection was modified" in PlateFillService.TryAddClear.
PlateView now refuses editing input while IsFillInProgress, the same
way it does during an inspection. Panning and zooming stay live.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Plans every nest on a dedicated OpenNest server (OPENNEST_PROBLEM_NESTS_SERVER)
plate by plate, with a per-plate time limit, and fails when a plate ranks
below the "planning: 1=... 2=..." baseline in the nest's comments. Customer
nests and baselines stay on the server, never in the repository. Refuses a
server holding more than 200 nests, so a production archive is not planned
by mistake.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Console --plan-cutting and MCP plan_cutting/apply_cutting_plan run the
desktop's CuttingPlanBatch: capture, plan, then an all-or-nothing apply.
Settings come from an explicit JSON file or the plates' identical saved
settings, never defaults. Unverified plans need per-call acceptance,
which is not posting consent. Console --no-fill plans a loaded nest as
is; a blocked plan exits 2 and saves nothing. Scrap cut-offs stay
desktop-only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
An attempt along a part order stopped backtracking after 8 x entries x
contours expansions without progress, with contours taken as the most of
any part on the plate. One many-holed part therefore let every blocked
part, even a two-contour tab, replay its predecessors' entries for the
whole of that allowance; a plate whose rounds each stalled that long
ran out of budget before the recovery pass that would have finished it.
Along an order the stall now uses the larger contour count of the blocked
part and of the part cut just before it, whose departure a retry changes,
capped by the old limit. The sequence-free recovery pass keeps the plate
maximum. Planning characterization digests are unchanged.
The optional private-fixture test now also runs RecoveryCuttingNestPath
with default lead-ins, and requires that file's plates to plan within
60% of their budget.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The cut planner's directed-contour recovery pass gave up on plates the
full search could solve, so they ended "no plan found within the search
limit":
- Each hole was checked as an isolated probe that repeats the part's
scribes, so every hole-to-hole hop was checked as a detour through
the scribe start. Scribes and other open cuts are now checked once
from the arrival, and each hole as a direct hop from the one before.
- Hole entries were ranked only by pierce distance, so the departure
often ran back across the hole just cut. The nearest hole is still
tried first, but its entries that leave toward the holes still to
cut come first.
- The first-ranked outside entry could spend the whole recovery
allowance. Each outside entry now gets a share of what is left.
The optional private-fixture test now takes its config key, adding
RecoveryCuttingNestPath; it fails on the old search and passes now.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A plate of at least 200 parts with no cutoffs or prerequisites, whose
part centres bin into rows and columns within a quarter of the smallest
part extent, is toured line by line in alternating directions. Each row
and column gap must be wider than that tolerance, each cell may hold
only one part, and at least a quarter of the cells must be filled
(interleaved rows fill half). The route runs along whichever axis, from
whichever corner, is shortest from the start point.
This replaces the travelling-salesman 2-opt/Or-opt passes, which took
about 8 s on 5,742 parts. It is still only a proposal: the joint search
checks every rapid and lead, and replans after a learned rule use the
general order. Smaller plates, and the characterization fixtures, keep
their order unchanged.
The washer nest's plate 1 now plans Ready in about 44 s (was 54 s), with
86k expansions against 91k.
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>
ajandClaude Opus 5.5