Commit Graph
35 Commits
Author SHA1 Message Date
jyvikoandClaude Opus 5.5 6700e6ef1f kernel: a body SMLib keeps in the record's order is mapped, whatever order its export lists
An imported solid's faces were mapped to the file's only when SMLib's export numbered them as the
record did (native.rs, numbering). The export is in canonical order, which can list edges
otherwise: the two periodic-knot tubes of #1374 came back with their seam and rim edges listed
[2, 0, 1, 5, 3, 4], so the sheet, a valid B-rep, was refused as "SMLib renumbered the solid's
faces". The export's order says where SMLib's own face and edge i went; put back in SMLib's
order it numbers them as the record does, and SMLib's i is the record's i (an import lists them in
the record's order). numbering now takes that too, checked the same way; an edge split into
pieces is no renumbering and stays unmapped.

The tube test asserts what it should: one surface body, not a mesh, its faces the file's two
faces in order, every face meshed, in under a minute, and the area exact: each face is the file's
quadratic circle (corner weights sqrt(2)/2 to 12 digits) moved straight along its axis, a
cylinder's side, so 2 * 2 pi r h, which SMLib's measure meets to 1e-6.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-30 15:19:38 -04:00
jyvikoandClaude Opus 5.5 f37b954988 Native B-rep: a body is written as it serializes, not through a JSON value
encode_native_body turned the whole saved record into a serde_json
Value (a map with an owned key for every field of every face, edge,
curve and pcurve), checked it, counted it, then wrote it out. Building
that tree was most of the cost: on the Menger level 4 sponge (3 tools,
--save) the 101 bodies the model cache encodes, 535 MB of encodings,
took 24.1 s of encode_native_body, 20.8 s of it in to_value.

binary::write_serialized is a serde Serializer for the tagged binary
format: it writes each value as it is serialized and applies the same
checks as it goes (nesting, string and array budgets, floats within
1e12, the value budget, the size limit). An object's entries are put in
key order when it closes, as the JSON map held them; floats that are not
finite are null, as to_value made them. The bytes are the same: a test
encodes plain, blended, cut, extruded, painted, split, open, round and
far-placed bodies both ways and compares them, and all 101 Menger bodies
(6 to 28 412 faces, 15 kB to 68 MB) encoded both ways in one run were
byte for byte the same. The JSON-value writer stays for the tests that
build hostile records.

Menger L4 3-tool --save, two runs each: encode_native_body 24.1 / 25.3 s
before, 3.0 / 3.0 s after (the longest call 1.34 s -> 0.29 s); the run
41.2 / 41.4 s wall before, 39.1 / 39.2 s after (the encodings run on
16 threads).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-30 05:55:40 -04:00
jyvikoandClaude Opus 5.5 e7e613b6fb Records: the item budget bounds input, not the kernel's own results (Refs #3992)
The portable record's 4,000,000-item budget refused a real result: the join of a level-4
Menger sponge's level-3 sponges ended with "SMLib export: the mapped B-rep does not validate:
portable B-rep: item budget". The budget guards records from outside the kernel, and a body
the kernel built in memory from inputs already bounded where they came in isn't hostile.

Brep::validate keeps the budget for records read from outside: native design caches
(decode_native_body, behind its own 256 MiB, 1M-per-array and 20M-value caps) and STEP import.
The new Brep::validate_built runs every other check, without the cap, on what the kernel
builds: SMLib's export of an operation's result, a body's record (Body::from_record, after
the decoder has checked a file's), a part of a validated record, the record handed back to
SMLib, and the validity check. Its checks are linear in a record already in memory.

SMLib's export bounded its counts by the same budget; a count is now bounded by SMLib's own
longest array (flat::MAX_LEN, checked on entry), which keeps every offset worked out from
the counts from overflowing, and the redundant topology-length cap goes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-29 22:15:38 -04:00
jyvikoandClaude Opus 5.5 f11c0c5712 Bodies: an operation's result is its record, not the SMLib body it was exported from
A body made by an SMLib operation kept that SMLib body beside its record (SMLib's
export). The export is not exact to the bit, so the kept body measured and operated a
little differently from the record's import, which is what a body read back from a
file has: a box measured 24000.000000000004 mm³ live and 24000.0 reopened, a drilled
box 21989.38056 against 21989.38013. A cut ring's volume therefore changed when it
was moved, the moved ring being measured on the SMLib body the move made
(measure_centre_tests::combine_cut_volume_does_not_depend_on_placement). Every body
now imports its record when an operation first needs it, so evaluated and reopened
bodies are the same.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-28 22:12:37 -04:00
jyvikoandClaude Opus 5.5 319c0bc9f7 Profiler: kernel operations and STEP/IGES import stages are spans
Booleans, blends, sweeps, lofts, shell/draft/offset, healing and sewing,
SMLib conversions, meshing and measuring; the import as parse, index,
per-solid read, placement and mesh check, convert batches and finish.
The existing per-step boolean/tessellate capture names its spans too.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-27 11:40:49 -04:00
jyvikoandClaude Opus 5.5 623f1b9c16 kernel: native and SMLib-import tests on the record; the adapter's tests go
native.rs's test module keeps its encode/decode, wire-record, hostile-input and appearance
tests. Deleted with the adapter they tested: pcurves_on_planes_are_the_curves_mapped_exactly,
pcurves_fit_curved_faces_within_tolerance, pcurves_of_seams_lie_a_period_apart and
pcurve_pictures, with their helpers (pcurves_of, check_pcurves, check_faces, curved_bodies,
seam_faces, draw). smlib_import_tests count shells on the record; their notes no longer
speak of truck.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
2026-10-27 02:28:50 -04:00
jyvikoandClaude Opus 5.5 252b2947a1 kernel: mesh bodies built by Body::of_mesh; fuzz booleans have no fallback counts
Mesh distance callers take geom's closest_on_triangle in place of the deleted guard's copy.

(not built: base in progress)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
2026-10-27 02:28:49 -04:00
jyvikoandClaude Opus 5.5 f6590bcaa6 kernel: a saved body is its record; SMLib bodies convert with no healer
Encoding writes the body's own record and decoding makes the body from it: the adapter to and
from the old kernel is gone. A record goes into SMLib as given (SMLib's own form); one SMLib
refuses is an error that names why, never sent through SMLib's healer. A body made by SMLib keeps
the SMLib body. The bounded preview of a decoded body is a coarse mesh from the planner, charged
to its triangle budget.

(not built: base in progress)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
2026-10-27 02:28:48 -04:00
Brennen HerbruckandClaude Opus 5.5 8fde178fe4 kernel: any schema 4 record into SMLib with its order, healed where refused
smlib_record_ordered is smlib_body_ordered for a record of any origin (a
STEP file's, not only a body's): SMLib's body of it, healed where SMLib's
check refuses it as given, with where each of SMLib's faces and edges is in
the record's order. A record SMLib refuses outright now goes through the
healer too before the import gives up.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-27 02:28:47 -04:00
jyvikoandClaude Opus 5.5 1d8299aedf kernel: an SMLib body keeps each edge's and vertex's tolerance through its record
SolidCraft's own record of a body SMLib made (the rig's record check, the saved body, the next
SMLib import) re-fitted every pcurve at the kernel's working tolerance, 1e-5 mm. SMLib's fillet
and chamfer surfaces lie some 2e-5 to 2e-4 mm off their faces, so 15 fillet and chamfer cases
failed that check although SMLib's own passed. And SMLib reported one tolerance for a whole body,
the largest any part of it needed, so one loose edge (a pocket cut left one at 0.23 mm) made the
whole body loose on its way back into SMLib.

Now the tolerances go per edge and per vertex, both ways:
- solidcraft-smlib reads and writes the fork's new SMLIB_FLAT_EDGE_TOLERANCE and
  VERTEX_TOLERANCE arrays. Export gives each edge and vertex record its own (never below the
  floor, which becomes the body's). Where those understate what an approximated edge needs on
  its face (SMLib holds a blend's end curve 2.3e-5 mm off its plane at 2e-5), the export falls
  back to SMLib's largest everywhere, as before. Import writes each record edge's and vertex's
  tolerance.
- `geom::brep::pcurve::EdgeUse` carries its edge record's tolerance, and a pcurve is fitted to
  it where it is larger than the face's.
- The kernel's decoded body keeps its record's edge and vertex tolerances
  (`Body::tolerances`, in its edge and vertex order) after an SMLib operation and a saved body's
  decode, and keeps them through a rigid placement. Its record writes them again; anything that
  builds a new solid drops them.

SMLIB_REV moves to the fork commit that adds the arrays (smlib-blends-abi2 4a67471, ABI 2).

Rig and ladder fillet and chamfer cases on SMLib: 43 pass before, 60 now, none lost. Tests: a
pcurve fitted to its use's own tolerance, and the ladder's 08/006 cylinder rim chamfer whose
record, saved body and placed copy all stay valid and SMLib's check takes back.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
2026-10-27 02:28:47 -04:00
jyvikoandClaude Opus 5.5 82865e2832 kernel: a body SMLib walks some loops of the other way keeps its order
smlib_body_ordered keeps SMLib's face and edge order for history only where SMLib's export
numbers the body as its record does, and compared each loop edge for edge in the same order
and direction. SMLib walks a face whose surface it turned (a hole's wall, a concave cylinder)
the other way round and may start the loop elsewhere, so a plate with a hole lost its history
through every SMLib boolean and shell although its faces and edges are numbered alike. A loop
now matches when it walks the same edges in the same cyclic order, or in the opposite order
with every edge reversed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
2026-10-27 02:28:47 -04:00
jyvikoandClaude Opus 5.5 ff0f7ccfa1 kernel: a body's record for SMLib takes a larger tolerance where an edge needs it
The record of a truck body is built to the body's working tolerance (1e-6 of its size, at
least 1e-5 mm), and truck's offsets and blends leave some edges a few times farther than that
from a face they bound: no pcurve fits, so the record and every SMLib boolean on the body
failed ("pcurve of loop 0 use 0 ...: the curve at 0 lies 3.2e-5 mm off the surface" in a
shelled revolve of the robustness generator). The record for SMLib now retries at ten and a
hundred times the working tolerance for the body, its edges and vertices before refusing; the
saved native body keeps the working tolerance. Robustness seed 64 (a filleted extrusion
joined to a cylinder), refused this way before, now builds as truck does to 1.6e-6 of its
volume.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
2026-10-27 02:28:46 -04:00
jyvikoandClaude Opus 5.5 00e43aacce kernel: an operand SMLib's check refuses goes through SMLib's healer
Truck's bodies are trusted into SMLib as given, but some miss SMLib's rules by a little (an
edge's vertex 4.4e-5 mm off it against a tolerance of 3.3e-5 after a truck operation), and
every boolean on them was refused by the check of its result. An operand whose trusted import
fails SMLib's check is imported again through SMLib's healer, and the healed body is used only
where SMLib then finds it valid; its numbering is checked against the record as before, so a
heal that renumbers leaves the result without history rather than with a wrong one.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
2026-10-27 02:28:46 -04:00
jyvikoandClaude Opus 5.5 d271add1ad kernel: revolutions go into SMLib as its cylinders, cones, spheres and tori
Truck holds every round surface as a placed revolution, and a body SMLib made comes back from
truck that way too, so the SMLib import refused spheres, tori and any second boolean on a
first one's result. The import now turns each revolution of a line along or through the axis,
or of a circle centred on it or in a plane through it, under a rigid placement, into SMLib's
analytic surface: the face's range mapped from the revolution's parameters (angles moved into
the first turn: SMLib's torus breaks on a tube range below zero), checked against the
revolution at 25 points, and the face turned over (loops reversed, shell uses flipped) where
the analytic normal runs the other way. Other revolutions are still refused.

A result body's order is checked by each face's boundary edges as well as its identity:
Body::new rephases revolved faces onto new faces with the same edges.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
2026-10-27 02:28:46 -04:00
jyvikoandClaude Opus 5.5 5238840bc8 kernel: booleans through SMLib behind SOLIDCRAFT_BOOLEANS, with SMLib's history
With the `smlib` feature, SOLIDCRAFT_BOOLEANS=smlib (read once per process, default truck)
sends every kernel::boolean through SMLib: each operand's schema 4 record imported and tagged
as operand 0 or 1, SmBrep's boolean, SMLib's check at every level, the result exported and
decoded (with its geometry check). A result either check refuses is an error. The setting
naming SMLib in a build without it, or anything unknown, fails every boolean instead of
quietly running truck, so a comparison never compares truck with itself.

The history is SMLib's tags renumbered into Body::history's index spaces. The export now
reports where each SMLib face and edge lands in canonical order (its arrays follow
SmBrep::GetFaces/GetEdges, the order the tags use). The result's order is checked against the
decoded solid by face and edge identity, the operands' by re-exporting the imported body and
comparing its numbering with the record's; where either differs the result has no history
rather than a wrong one. The result gets the operands' regularization protection and planar
roles as truck's does.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
2026-10-27 02:28:46 -04:00
jyvikoandClaude Opus 5.5 ce201eea1b kernel: SMLib spheres cut by a boolean decode oriented and meshed
Results of SMLib booleans on its own spheres came back wrong through
body_from_smlib (side-by-side runner: seed 13 cut −1611.9, seed 61 cut −252.4
against SMLib's 4056.9, seeds 46/61 union 80 % short). The schema 4 records were
right (check_geometry passes, outer loops counter-clockwise, shells out of the
material); three things on the truck side were not:

- Body::new orients a solid by the volume of truck's coarse mesh, and truck's
  mesher meshes a sphere bounded by its seam alone as nothing: the rest of the
  body set the sign and turned it inside out. A foreign record is oriented by
  its shells, so body_from_smlib now builds the body with Body::oriented
  (revolved faces turned as before, orientation kept, nothing meshed).
- rephase_revolved turns each revolved face so it sits around angle π. A face
  that uses its seam both ways goes all the way round; turning its surface ran
  the wrap through the face's hole, the display plan failed and the sphere
  meshed as nothing. Such faces now keep their wrap on their seam (truck's own
  primitives have none: their round faces are split).
- The display plan opened only a loop of one seam used both ways. A sphere face
  whose meridian a trim splits uses each piece both ways, turning back at each
  pole, and projected onto one parameter line ("boundary crosses itself"), so
  the whole body fell back to truck's mesher and lost its sphere faces. A loop
  through both poles that turns straight back at one now has its run between
  the poles moved one period along, whichever way keeps it within one period
  with a positive area (poles split as the planner splits them); the routing
  check finds such loops too.

Test: SMLib's sphere less a box cap beside its seam, across it (seam split) and
a bar bored into its side away from the seam decode without meshing to orient,
oriented as recorded, every face meshed, within 1e-3 of the exact volume.
Before, the decode meshed the body to orient it; with only the plan's pole
handling undone the −x cap measures 50.3 of 469.1 mm³, with only the
rephase undone the bored ball 252.4 of 4059.5.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
2026-10-27 02:28:45 -04:00
Brennen HerbruckandClaude Opus 5.5 91e482fc8a solidcraft-step: refuse loops across a closed spline surface's seam; conformance parts and truck volumes
A face winding around a closed (non-periodic) spline surface without a seam edge has loops
that jump across the surface's domain; it is refused for now instead of making a record whose
loops do not close. The step-conformance parts are read in a test, and the kernel's ignored
corpus comparison builds each record back into a truck solid to compare volumes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-27 02:28:44 -04:00
jyvikoandClaude Opus 5.5 144b569246 solidcraft-smlib: SMLib bodies into SolidCraft through schema 4
SMLib B-reps leave the C API as its flat arrays, read through the crate's
one flat wrapper (crate::flat::Flat, copied out as import::Arrays), and
become schema 4 records: shell pairs merged, sheets collapsed, radial cycles
normalised, analytic surfaces as their kinds (narrowing cones kept), loops
reversed where SMLib's surface is inside out, every pcurve from
brep::pcurve, canonical order, then validate() and check_geometry(). Edges
arrive as their own kinds; NURBS edges that are lines or circular arcs (the
primitives' circles) are recognised as a fallback and counted. With the
kernel's `smlib` feature, body_from_smlib() decodes them into a Body.

Adds the cylinder, sphere, cone and torus primitives. Each primitive passes
SMLib's own check at every level, its volume from the exact mass properties
matches the closed form, and it goes round SMLib -> schema 4 -> SMLib with
its counts, volume and a clean check. Truck now meshes a whole sphere
bounded only by its seam (SMLib's sphere) through the display plan.

SMLIB_REV moves to the fork's capi-primitives (the primitives on capi-flat
40bb855).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
2026-10-27 02:28:44 -04:00
Brennen HerbruckandClaude Opus 5.5 2a22f4641b solidcraft-smlib: schema 4 records into SMLib through the flat arrays
SMLIB_REV moves to the fork's capi-static 40bb855 (flat arrays, C ABI 2).
flat.rs wraps them (Flat: arrays by id, Brep::export, Flat::import,
Brep::is_manifold_solid; FFI in ffi/flat.rs). import.rs lays a schema 4 record
out as those arrays: a closed shell becomes one SMLib shell per region it
separates, a sheet's shell holds both sides of each face, the infinite region
is marked void as SMLib has it, pcurves become UV trims on B-spline faces only
(reversed for edgeuses against their edge), planes with non-unit axes go as
their exact bilinear patch. Brep::from_record builds the SMLib body.

kernel: brep_record(body) gives a body's schema 4 record, and smlib_body
(feature smlib) its SMLib body, trusted (no healer: healing moved a
cylinder's volume by 9e-7). Box, cylinder and tilted cylinder keep their
analytic volume in SMLib (1e-9, 1e-6) and are manifold solids; truck's
spheres and tori are revolutions, refused with "not supported yet" for now.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-27 02:28:44 -04:00
jyvikoandClaude Opus 5.5 95b2295160 kernel: the truck adapter writes pcurves through brep::pcurve
The adapter now saves each edge's curve record and interval when the edge first appears and
computes every face's pcurves from those records and the surface record it writes, with the
shared plain-record routine; the truck-face version in native/pcurve.rs is deleted. Loops go
in as the record keeps them (relative to the surface), so a reversed face no longer has its
wires reversed and back again.

The native tests are ported (adapter::face_alone gives one face's record for them). The cut
body's intersection pcurves now fit at 1e-6 mm too: the routine follows the converged
intersection, and truck's own points are within its 1e-6 mm of it, which the lift check
allows for on intersection edges. docs/native-brep.md says where pcurves come from and what
a surface intersection's point at t is.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
2026-10-27 02:28:44 -04:00
jyvikoandClaude Opus 5.5 2f4d6acacb brep: working tolerance, golden v5 records, cache producer 4
Edges, vertices and the body carry the face tolerance (10^-6 of the
body's size, at least 10^-5 mm): truck's traced intersection edges sit
about 1e-6 mm off their surfaces, so the linear tolerance is too tight for
a pcurve to lift onto them. The box-and-fillet golden fixture is the
schema 4 envelope now (NATIVE_BREP_BLESS=1 rewrites it), and the adapter
tests cover a pair of lumps. The model cache's producer revision becomes
solidcraft-truck-adapter/4.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
2026-10-27 02:28:44 -04:00
jyvikoandClaude Opus 5.5 445fb23859 brep: schema 4 types, validator and truck adapter
The plain B-rep follows SMLib's flat form with STEP's names: regions
(infinite first), face/wire/vertex shells, oriented faces, faces with a
domain and edge or vertex loops, oriented edges with a pcurve each, edges
with an interval, a tolerance and their list of uses, vertices with a
tolerance. Curves and surfaces are unbounded; their bounds are the edge
interval and the face domain. Loops are relative to the surface's natural
normal. Schema 3's read-only intersection and revolution records go.

The validator keeps schema 3's caps and adds the structural rules:
references, loop shape and connectivity, regions and closed shells, the
pcurve interval, natural parameter ranges, every edge's uses (primary
first, upper = !forward) and the canonical order of shells, faces, edges
and vertices. A reversed face is walked backwards, as truck's edge_iter
walks it, so truck's face and edge indices are the canonical ones.

The truck adapter writes truck's absolute wires as the loops, numbers
everything as the kernel does, and builds regions from the shells (a
tessellated nesting test only when there are several closed shells).
Pcurves come from pcurve::pcurves only; until it supports a face, encode
fails and the body is a cache miss. The envelope signature becomes
SCBREP05.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
2026-10-27 02:28:44 -04:00
jyvikoandClaude Opus 5.5 4afbe3b909 kernel: pcurve test for a face round its surface without a seam
A band of a cylinder bounded by two full circles and no seam cannot close
in (u, v): each loop's pcurve runs one period. The test records that such
loops come back continuous and a period open (the caller needs a seam
there), and that a hostile tolerance is refused.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
2026-10-27 02:28:44 -04:00
jyvikoandClaude Opus 5.5 c7088fac9c kernel: pcurve tests for seams, poles and loop orientation, and pictures
Faces with seams as imported faces have them (the kernel's own
revolutions come in pieces): a cylinder whose seam line is used twice
between two full circles, each a closed edge on one vertex, and a sphere
bounded by one meridian used twice from pole to pole. Their seam uses come
out a period apart, in the base period. The checks now also require every
face's outer loop to turn counter-clockwise in (u, v) for a face along its
surface's normal and clockwise against it, holes the other way: the rule
the pole junctions rely on, which every test body follows.

With SOLIDCRAFT_PCURVE_SVG=<dir>, a test writes each test body's face
loops in (u, v) as SVG for review.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
2026-10-27 02:28:43 -04:00
jyvikoandClaude Opus 5.5 099026d94d kernel: fitted pcurves on curved surfaces, unwrapped and closed per wire
Every edge not exact on a plane gets a cubic 2D B-spline on its curve's own
parameter: through the parameters of the curve's ends, least squares
through samples between (banded normal equations), with spans halved and
sampled twice as densely wherever the lift misses the curve by more than
half the tolerance, at the samples or halfway between them. Knots start
where the curve's parameterisation is less smooth: a spline's own knots
and the vertices of an intersection curve's leader polyline, with the
multiplicity a cubic needs there (without them the cylinder-minus-box
intersection curves took 400 spans at 1e-3 mm). Spans and rounds are
capped; a fit that does not converge, or a curve farther from the surface
than the tolerance allows, fails naming the wire, edge and distance.

Samples are projected with the kernel's own surface parameter search (the
evaluator's form, parameterised as truck parameterises it; truck's search
for surfaces it does not take), each from the one before, unwrapped along
periodic directions. Where the surface collapses (a pole) the undefined
parameter comes from the nearest sample that defines it, so a pcurve runs
straight into the pole. Each wire's uses are then moved by whole periods so
each starts where the one before it ends, starting after a pole junction
if there is one; at further pole junctions the offset that gives the loop
a boundary's turn is taken. The wire then moves into the base period. A
seam edge used twice is fitted once.

Truck places intersection curves on their surfaces to about 1e-6 mm, so
their pcurves need a tolerance of 1e-5 mm or more; the test records that.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
2026-10-27 02:28:43 -04:00
jyvikoandClaude Opus 5.5 26ab8981d0 kernel: exact pcurves of lines and splines on planes
A plane's parameters are affine in space, so the pcurve of a line or a
(rational) B-spline on a plane is the curve's own knots and weights with
its control points mapped into (u, v): no fit. Curved surfaces and
intersection curves still return "not supported yet". A test helper checks
every pcurve the way the schema validator does, its interval against the
curve's range, the lift error and the wire's closure in (u, v); the box's
line pcurves lift onto their lines' ends bit for bit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
2026-10-27 02:28:43 -04:00
jyvikoandClaude Opus 5.5 ee5deffaa4 brep: Curve2 pcurve type and the truck pcurve entry point
The shared base for the schema 4 types and the truck pcurves: 2D B-spline
pcurves as SMLib keeps its UV trim curves, and the truck adapter's pcurve
function, not supported yet.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
2026-10-27 02:28:43 -04:00
93a2922c5e Display plans and the model cache for large STEP assemblies (#2814) (#4021)
* STEP import: mesh band faces that wrap around a closed surface correctly

Fixes two related bugs that caused SolidCraft to fail to mesh a face wrapping
all the way around a closed, non-periodic surface (a revolved ridge band, or a
tube-shaped B-spline surface): a missing face, a folded/self-overlapping
triangulation, or a hole in the rendered body.

1. Ring edges that run slightly off their surface. STEP files write a surface
   and the curves bounding it to the file's own declared uncertainty, so a
   ring edge can legitimately sit a few micrometres off the surface it
   bounds. The code finding where a ring edge crosses the planned split line
   compared against the edge's raw (off-surface) point using a hardcoded
   tolerance far stricter than the file's own precision, rejecting a valid
   cut. Now matches against the point on the surface directly below the ring
   crossing, respecting the file's declared tolerance.

2. Closed B-spline surfaces, their slits and their slivers. truck only treats
   the rotation of a revolved surface as periodic; a face on a B-spline
   surface whose control net closes on itself (a tube) wasn't recognized as
   wrapping around at all. This produced a folded face, a hole (a boundary
   loop landing a full period away from where it should cut), and one
   near-zero-width sliver face costing the entire body its tessellation
   plan. Adds periodicity detection for closed B-spline surfaces, period-
   aware hole-loop alignment, dropping of zero-width slit boundaries, and
   isolating a too-thin-to-triangulate face so it no longer takes the rest
   of the body down with it. The GPU shader's periodic wrap mirrors the CPU
   math so the two don't drift apart.

Regression tests: band_faces_split_when_ring_edges_run_off_their_surface,
band_face_with_a_zero_width_slit_is_still_meshed,
a_sliver_face_too_thin_to_triangulate_does_not_cost_the_body_its_plan.
Ablation-tested: disabling either mechanism in (2) reproduces its target
test's original failure message while leaving the other test unaffected.

* STEP import: mesh the band's split faces, not just split them

jyviko asked whether a band wall missing from the interactive app's render
was a shading issue or a genuinely missing face. It was missing: the ring-
edge split fix made the band's topology correct, but plan_face still
produced zero triangles for the two new sub-faces.

Two causes, both in plan_face (crates/kernel/src/tess/mod.rs):

1. The tolerance for how far an imported boundary may lie off its surface
   was tied to the requested mesh's chord tolerance instead of to the
   body's own size. A finer display tolerance could then reject a boundary
   a coarser one had just accepted, so the face would drop out exactly when
   zooming in on it.

2. Near a sharp corner of this flat-panelled band, several boundary points
   that are legitimately a few micrometres apart (within the STEP file's
   own declared precision) all projected onto the same surface point. The
   boundary loop then repeated a point in (u, v), which a triangulator
   can't follow, so the face's plan failed outright. One face without a
   plan costs the whole body its plan, and the fallback mesher can't
   handle a closed B-spline surface either, leaving both halves of the
   split band with no triangles at all: present in the topology, invisible
   in every renderer, and absent from the body's volume.

Fix: scale the off-surface tolerance to the body's own extent (new
Rules::extent), and collapse a run of boundary nodes that land on the same
surface point to one (drop_collapsed) before triangulating.

Regression test additions to band_faces_split_when_ring_edges_run_off_their_surface:
the body's volume now matches the analytic pentagon prism (was a third of
it - only the top cap - before this fix); display_plan leaves no face
without triangles from coarse to fine tolerance; and the actual tolerance
and facet angle the viewport draws with gives the split band its full
lateral area on both halves.

Confirmed on review_band_big.step, the fixture behind the review screenshots:
volume's relative error against the analytic pentagon prism drops from
1.4e-4 to 4.1e-6, and the band wall now occludes the floor grid behind it
in a snapshot the same way the top and bottom caps always did.

Full kernel suite: 739 tests, 0 failed.

* STEP import: hold an imported boundary to the file's own tolerance, not the part's size

jyviko: the off-surface acceptance added in the last commit
(rules.extent * OFF_SURFACE) scaled with the part, so a 2 m part
silently accepted a boundary 2 mm off its surface while a 20 mm copy
of the same part accepted only 0.02 mm. A STEP file's own declared
uncertainty does not grow with the part, so the bound shouldn't either.

Body gains boundary_tol: Option<f64>, Some(file's declared tolerance)
for a body read from STEP (crate::step_in::Units::tol, already used
by the ring-split fix in split.rs), None for anything this kernel
built. It's set before the import's own unmeshed-face check runs, and
carried over by the two face-split operations that re-cut an imported
surface (splitface.rs, outline.rs), since the file's uncertainty still
describes those boundaries afterward.

plan_face's accept now reads: the file's own tolerance when the body
declares one, or the old chord-tolerance/extent bound otherwise. The
old bound's tol * 10.0 term had to go too, not just the extent term:
at a 2 m part's typical display tolerance that term alone is ~20 mm,
nearly ten times the extent bound jyviko objected to, so keeping it
would have left the exact 2 m/2 mm example still silently accepted.

Two new tests in step_in/tests.rs:
- band_faces_split_when_a_scaled_copy_runs_off_its_surface: the same
  band fixture at a realistic 2 m scale with the same absolute slop
  and declared uncertainty as the small one, meshes the same way at
  every chord tolerance from coarse to fine.
- off_surface_budget_follows_the_file_not_the_part_size: the actual
  regression. The same off-surface error is accepted or refused the
  same way whether the part is 20 mm or 2 m; a 2 mm error against a
  1 µm declared uncertainty is refused at both sizes now (accepted at
  2 m under the old bound, refused at 20 mm - that asymmetry was the
  bug). Checked at both a coarse and a fine display tolerance.

Known gap, not fixed here: a STEP-imported body saved into a
.solidcraft design and reopened loses boundary_tol (native.rs's
decoder has no field to read it from) and falls back to the old
extent-based bound on reload. Today's design format has no concept of
this value at all; persisting it is a real format change, out of
scope for this fix.

Full kernel suite: 776 tests, 0 failed. clippy -D warnings clean.

* Display plans for STEP threads and pointed ends (#3847)

Three faces cost a threaded or pointed STEP part its whole display plan, so it
was drawn from the reference mesh: cracked, faceted, and never refined up close.

- A loop through a pole (a cone's apex on the boundary) seemed to wind around
  the surface: the jump at the apex, unwrapped to the nearest copy, was
  arbitrary. The loop is turned to end where it leaves the pole, so the step
  that closes it is an ordinary one (`seam::close_through_pole`).
- Projection onto a spline with hundreds of spans (a helical flank) restarted
  from a 16 x 16 grid and settled on another turn; it now starts from two
  points per span, capped.
- Sag refinement no longer inserts a point exactly on a boundary side, which
  split the side and failed the face.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* tess: a refused face's edges are divided finer and it is planned again (#2814)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LGeCa5jWWdroMWvXxN3LLM

* tess: sliver faces plan empty; a face no method plans is filled flat from its nodes (#2814)

A face whose boundary runs out and back along itself (two edges between the
same vertices, a rounding apart, as STEP files write slivers) encloses nothing
and is planned with no triangles. A face still refused after its edges are
divided finer (a cusp, a sliver whose sides sag across each other, a small cap
at a pole) is filled flat from its shared boundary nodes, so the rest of the
body keeps its display plan instead of falling back to the reference mesh.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LGeCa5jWWdroMWvXxN3LLM

* display mesh tests: a rim the chord tolerance leaves a line is divided again (#2814)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LGeCa5jWWdroMWvXxN3LLM

* portable B-rep: up to 8192 control points along a spline direction (#2814)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LGeCa5jWWdroMWvXxN3LLM

* solidcraft-cli: --debug logs what the engine reports to stderr

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LGeCa5jWWdroMWvXxN3LLM

* STEP import: a spline written over a tiny parameter range is read over one (#2814)

Some writers parameterize splines over a few millionths (a humanoid robot's
edges span 5.4e-6). The kernel tells parameters apart only to 1e-6, so over
such a range it evaluated the curve's end wrongly, and trimming an edge to its
vertices dropped a control point: the bodies could not be stored exactly and
the whole model cache was left out of the design. The knots are now stretched
to span one; the curve is the same and edges are trimmed by their vertices, so
nothing read from the file refers to the old parameters.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LGeCa5jWWdroMWvXxN3LLM

* model cache: a body keeps its file's declared accuracy (#2814)

A body restored from the cache dropped the accuracy the file it was read from
declares, so the planner held its boundaries to the kernel's own bound and a
reopened design drew differently from the import. It is stored with the body's
runtime values, and only when there is one.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LGeCa5jWWdroMWvXxN3LLM

* tess: a boundary within a hundredth of its body lies on its surface, at every quality (#2814)

#1721 held a body read from a file to the accuracy the file declares. A writer
declares what it aims for, not a bound its edges keep: a 251-body robot
assembly declaring 0.01 µm has boundaries 10 to 30 µm off their surfaces, and
95 of its bodies lost their display plans. The bound is a hundredth of the
body again, what Normal quality always allowed, and now the same at every
quality, so zooming in still never drops a face (#1721's point). A file
declaring more raises it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LGeCa5jWWdroMWvXxN3LLM

* STEP import: keep faces whose edge vertices sit slightly off their curves

Some exporters write an edge's vertex beside the end of its curve, further off
than the accuracy the file declares (a surface export declares
0.01 mm and puts a vertex 0.19 mm off two spline ends). The reader refused
such edges and left the face, and with single-face bodies the body, out.

A vertex may now sit up to the file's tolerance or 1/1000 of the model's size
(the diagonal around its VERTEX_POINTs), whichever is larger, never more than
the existing 1 mm ceiling. Spline and open-conic edges have the end control
point moved onto the vertex, so every such edge meets its vertices exactly.
The file's declared accuracy still governs closed-curve, seam and circle
checks.

* tess, io: review clean-ups (#2814)

- A sliver face is one whose loops are all slit: drop_slits finds it in the
  retry after a failed plan, so the pre-check on every face (folded) is gone.
- The flat fill uses cdt::triangulate (a lenient fill that leaves out sides
  crossing one already kept) and its flood-fill inside test, and the plane
  from geom::Plane.
- Refinement only builds its indexes when a face was refused, takes the
  faces to plan again from edge_faces, and re-samples edges on the pool;
  Rules::finer for the finer rules.
- One PackageWriter::fits for the model cache, the preview and the thumbnail.
- The band split's ring tolerance lives in split.rs; comments say what the
  off-surface bound is now.
- The CLI logger prints Level::as_str().

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LGeCa5jWWdroMWvXxN3LLM

---------

Co-authored-by: David Lackovic <lackovicdavid1234@gmail.com>
Co-authored-by: bherbruck <brennen.herbruck@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: jyviko <2501648+jyviko@users.noreply.github.com>
Co-authored-by: Ray <raykholo@users.noreply.github.com>
2026-10-21 10:37:12 +00:00
bherbruckandClaude Opus 5.5 349a6c6fc4 perf: a copy placed for showing doesn't turn its revolved faces again (#3044) (#3104)
Moving an occurrence places every body under it again, and each placed copy
searched every revolved face for where it sits to turn it away from its seam,
although its source had been turned already. On the Prusa Core One assembly
that search was most of the 12.7 s a move of the top occurrence took. A body
now records that its revolved faces are turned, and world placement
(`place_matrix`) skips the search for such a body. Modelling moves (pattern
copies, hole tools) keep `transform_matrix`, which still turns them in the
copy's own frame: where a face wraps all the way round the turn it finds
depends on those coordinates, and later booleans depend on it.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-20 11:01:56 +00:00
bherbruckandClaude Opus 5.5 d0afc31632 Save the evaluated model with every design: exact, stored once, scaled caps (#3053) (#3093)
* Model cache written by default: exact bodies, stored once, scaled caps (#3053)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* bench-open: say why bodies were not kept

* Exact planes and revolutions; budgets for large bodies (#3053)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Native B-rep: rebuild shells as the kernel made them (closed and seam edges) (#3053)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs: the model cache and the exact native B-rep (#3053)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* bench-open: time a save without the cache too

* Model cache entries are Deflate-compressed (#3053)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Model cache: one that does not fit the design file is left out, never a failed save (#3053)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-20 02:31:08 +00:00
02c77df5f7 inspect.clearance: mesh each body once per check instead of once per pair (#2867)
* inspect.clearance: pairs whose bodies did not change are not measured again

Each clearance check measured every close pair again, though between two
checks usually only a few bodies change or move. Results are now kept per pair,
keyed by both bodies' geometry and placement and the check's settings; a pair
whose two bodies are unchanged is taken from that cache. With 40 overlapping
parts, a check after moving one of them took 0.8 s instead of 23 s (741 of 780
pairs reused). Pairs that could not be worked out are tried again each time.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* inspect.clearance: boxes and placed keys exactly as before the reuse

The pair boxes are taken over the mesh's triangles with finite corners, as a
distance shape's are, so pruning is unchanged. A placed body's identity now
includes its model body's own frame, so two placements are never mistaken for
one another.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* inspect.clearance: a large body is meshed once for all its pairs

Each pair that touches cuts the larger body down to the pair's region with a
kernel boolean, and that boolean meshed the whole body twice: once for its
checks and once inside the intersection. A sheet with 600 faces paid about
2.5 s per pair for meshes that were the same every time.

A body can now keep its meshes for reuse (Body::keeping_meshes); its booleans
take the meshes from there, and the intersection takes its shell meshes
from there too (truck-shapeops: and_with_history_meshed). Clearance uses such
copies for one check. Results are unchanged to the bit. On a 3 x 2 m sheet
with bends and 300 holes, the clip took 0.23 s per pair instead of 2.6 s, and
24 rivets took 5.6 s instead of 16.5 s.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Co-authored-by: bherbruck <brennen.herbruck@gmail.com>
2026-10-19 16:14:24 -04:00
bherbruckandClaude Opus 5.5 9b0f028f13 STEP assemblies: mesh a repeated part once and place it for its instances (#1152)
The reader converts each item once, but every instance made its own display mesh
(tessellation, seam and tangent flags), the most expensive step of opening a large
assembly. A later instance of an item now shows its first instance's mesh, placed by
the difference of the two placements. Only rigid placements share; split lines and
any operation drop the link. Measurements and export read each instance's own
geometry, so eval output is unchanged.

Generated assemblies: mesh phase 2,215 -> 139 ms (901 bodies) and 1,028 -> 85 ms
(5,401 bodies), with the same triangle counts.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-14 14:23:33 -04:00
SahenandClaude Opus 5.5 2bb24d1d2a perf: prism pairs before the mesh, bodies measured once, and a STEP import that copies less (#714)
A cut or an intersection between prisms (a plate and the tools through it) is now decided in 2D before either operand is meshed. A body's size and volume are measured once, a placement reads its orientation from its matrix, the evaluation cache is bounded to 64 MiB, and STEP import copies fewer shells. Hole plates evaluate 20-30% faster with less memory, and models are unchanged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-13 10:55:56 -04:00
selune 913583bfee Show a bounded native-cache preview before authoritative reconstruction (#463)
Refs #303. file.save / file.save_as can store a warm native geometry cache (cache: true, off by default) keyed to the document hash, app version and adapter revision. Opening shows a bounded preview from it until the rebuilt session replaces it. Any cache failure counts as a miss.
2026-10-13 02:54:01 -04:00
selune 9f88537927 Define a kernel-neutral native B-rep format and guarded adapter (#461)
A SolidCraft-owned portable B-rep schema (geometry schema 2 with analytic curves and surfaces, explicit frames and ranges, fitted intersection curves) and a versioned body envelope (SCBREP03) with its own metadata records, a bounded binary codec, a validator with resource budgets, and an adapter that refuses what it can't convert exactly. Pinned by a box-and-fillet golden fixture. Not yet written by application saves. Refs #303.
2026-10-13 00:13:29 -04:00