At a pole or apex `EvalSurface::eval` takes the normal from a step beside the
point, trying (+h, +h) first. At the far end of an interval (a sphere's north
pole at u = 1) that step lands past the end, where the extrapolated surface
folds back over the pole, and the normal came out turned over: the pole vertex
shaded dark with a light ring around it (ladder 01/005, every sphere and dome).
Steps now stay inside the active interval of a spline or a revolution's
profile. The pole test checked only that the normal was radial, either way;
it now checks the side too.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
form::extent took the box of 32 points per edge. A face's NURBS form is built
over that box, so where the boundary peaks between two samples the form
stopped short of its edge: the re-imported rod's bore ended at v = 9.99977
while its intersection with the rod reaches 10, and SMLib found the edge
1.9e-4 off the bore (SmBrep #4, SmEdge #13). Each sample higher or lower
than its neighbours now brackets the edge's extreme there, found by golden
section.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0113gDVmStT5tCGKW9Fxgc5K
form::extent unwrapped each loop of a face from its own first point, so a
hole beside the seam (a cross bore through a rod at the rod's seam angle)
came out a period away from the outer loop. The box over both was more than a
turn, cut to one turn from the hole: the face's NURBS form ran over quarters
3..7 while its seam uses lay at 4 and 8, and SMLib found the re-imported
rod's edge ends outside its face domain (SmBrep #70-#72). Each loop after the
first now goes the whole periods along that keep the face's box smallest.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0113gDVmStT5tCGKW9Fxgc5K
With touching: separate (the default) a join groups the joined body's
pieces by face contact. The groups that hold the material of the body
before the feature stay in it (each unless it touches one already kept;
the first group where the lineage doesn't say), and so do the groups that
touch nothing: a virtual grouping later operations take as one body. Every
other group that touches another along an edge or at a point becomes a
body of its own, and the feature's notice says so. The joined body keeps
its name and never holds an edge or point contact. A join that only
grouped pieces apart gives no notice, so KeptSeparate::Apart is gone.
one_body still keeps every piece in one body.
SMLib's export copies an edge or vertex that touching solids share once
per solid, so the record never listed those contacts:
Brep::region_contacts now takes such copies (the same curve and interval
between vertices at the same points, a vertex at the same point) as the
edge or vertex they share. A contact SMLib leaves out of the topology (the
3612 spoke tangent to its hub) is found on the exact geometry, the least
distance within the record tolerance, and reported as KeptSeparate::
Touching ("along a line or at a point").
The pending format change join-touching-bodies stays: older designs kept
pieces touching along an edge or at a point in one body too. Its notes say
so, and the reopened legacy emboss still expects touching: one_body.
Tests restored to their form before cc49081df and 733d3bd19: kernel-cases
2896-join-arm-end-near-ring-wall-gap (one body of two shells), 3379, 3455
and 3526 (no touching: one_body); the engine tests of posts joined while
apart (twice), two cubes for interference, a ring with a cylinder, the
buried tab and a_join_adds_only_its_sketch's 0.1 mm gap.
join_meeting_nothing_keeps_the_first_body expects one body and no notice.
The #3378 spoke stays 1 mm inside the hub.
New tests: blocks touching along an edge are split with the along-an-edge
notice; blocks apart stay one body without one; a ring with an arm on its
corner edge and a hub apart gives ring and hub in one body and the arm on
its own (doc); the engine's notice for an edge contact; the record's
contacts of cubes sharing an edge or a vertex (kernel).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Brep::region_contacts lists each pair of material regions that shares a
face, an edge (used by faces of both) or a vertex, with the most it
shares; pairs not listed are apart. Body::region_contacts reports it for
a boolean's result (lump positions as in Body::lumps), and Body::parts
cuts a body into groups of lumps (lumps is parts of one each). Exact
facts from the record only: no geometry, no decision.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
(cherry picked from commit 64a1d688e68e5f0c5b1f9e4fe0d2facfaa4744be)
A degree-1 patch of four control points on a parallelogram (a plane as an IGES
or STEP B-spline writes it) is that plane at every parameter, so its pcurves
take the plane's closed forms instead of going to SMLib (which drew the rim
circle of 3806's base as a 385-point spline).
SMLib may run a one-face sheet's loop against ours. A use's pcurve is the same
either way, so the loop's way round is read from an edge the face uses once and
every use is matched with it (3806 failed with 'a use SMLib did not keep').
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BnQawySUFx57iaJZq7z1Fm
exact_pcurves gives each use's pcurve where it has a closed form (or is given)
and None where it has none, without closing the loops; pcurves_within closes
them once every use has its pcurve. TrimFace, TrimEdge, TrimUse and the Trims
trait carry a face to a kernel that makes the pcurves of the uses with none.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BnQawySUFx57iaJZq7z1Fm
along took u or v the edge's own parameter at slope ±1. A rim SMLib makes (an
extruded or revolved circle, four quarters with scaled knots) runs the face's
quarter grid at another steady pace, and a closed rim's two ends are one point,
so neither its ends nor slope ±1 give its line. The pace is read from the edge's
points a quarter and three quarters along, unwrapped from the middle; the line
is taken only where it lifts onto the edge within the tolerance.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BnQawySUFx57iaJZq7z1Fm
Validation refuses a circle or ellipse edge, or a cylinder, cone, sphere,
torus or revolution face, without its NURBS form. Test fixtures are built on
forms; seam and domain defects use the form's parameter (one turn = 4).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BnQawySUFx57iaJZq7z1Fm
- A meridian of a sphere or a tube circle of a torus (a circle in a plane through the axis) has
its exact pcurve on the label: u fixed at the arc's azimuth, v its angle from the outward
radial, alpha + s t.
- A placed revolution has a NURBS form: its profile's control points turned about the axis on
the quarter grid from angle zero, weights multiplied, then its matrix; the record takes forms on
revolutions.
- `form::extent`: a face's parameter box from its boundary projected on its label (the collapsed
parameter at a pole bounds nothing; a boundary that never goes round covers the turn), what its
form is built over with no pcurve involved.
- A face that goes all the way round gets a form of exactly one turn, closed and so periodic.
- `along` projects inside the face's box (a spline extended past its knots, or a profile turned
past the face, has other preimages), and takes a closed form within the record's tolerance, as
the closed forms on analytic surfaces are.
Tests: fits are gone, so curves with no closed form on their face (a tilted circle on a sphere,
an ellipse on a cylinder, an intersection record, a curve weaving across a thread flank) are
asserted refused; the closed-spline fixtures are built as forms. geom 166/166.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BnQawySUFx57iaJZq7z1Fm
`On::fit` and everything it used (samples, least squares, breaks, filled poles) are deleted, and
the STEP reader no longer retries a face at ten times its tolerance. A pcurve is exact (closed
form, an affine image, or a straight line along a form's parameters) or the face is refused:
"not supported yet: no exact pcurve of <curve> on <surface>".
The rig and ladder are as with the fit (333/372, 101/116; none lost; 3806, already failing, now
refused on import). 16 unit tests that build sphere meridians, torus seams, closed spline
bands, intersection curves and surfaces of revolution on their labels now refuse: those paths
need meridians labelled on their face's meridian plane and revolutions on the quarter grid.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BnQawySUFx57iaJZq7z1Fm
`form::edge` is the one rule sew and the STEP reader share: a circle about a revolved face's
axis is labelled on that face's frame, and a circle or ellipse edge gets its form. The STEP
reader now gives every imported edge its label and form before any pcurve, and every
cylinder, cone, sphere and torus face its form over the extent its label's pcurves give; the
pcurves, the outer-loop choice and the turn-over test then run on the forms, and the record
carries them.
With the pcurve fit turned off, the STEP and IGES circle cases pass (8 of the 9 import cases
that needed it; 3208's ellipse on an extruded spline surface remains).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BnQawySUFx57iaJZq7z1Fm
The bend's rim arcs carried frames of their own (x toward the arc's start), a fraction of a
quarter from the bend cylinder's, so their forms were off the cylinder's quarter grid and no
straight pcurve lay along them. `form::on_frame` relabels such a circle on the face's frame (the
same circle, its interval shifted); its form is then on the face's grid, its pcurve u = ±s + c.
`along` takes u or v as the edge's parameter either way along, offset. The volume test steps to
a pcurve's ends exactly (t0(1-f) + t1 f), as an interval's end may now carry rounding.
With the fit off, 38 of the 44 sew cases pass (from 29). With it on, the rig is unchanged
(333/372, ladder 101/116, no case changed).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BnQawySUFx57iaJZq7z1Fm
Every circle and ellipse edge sew builds now carries its NURBS form, its interval in the form's
parameter; every cylinder, cone, sphere and torus face its form over the extent its label's
pcurves give; and every pcurve is computed on the forms, so a bend's rim circle is a straight
line on the bend and an affine image on the flange, exactly. The volume test evaluates the faces
on their forms. A form interval's end within rounding past its last quarter stays on it.
With the pcurve fit turned off, 29 of the 44 sew cases that needed it pass (from none); with it
on, all 44 pass as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BnQawySUFx57iaJZq7z1Fm
On a plane, a B-spline or NURBS edge's control points go through the plane's affine parameters
and the result is cut exactly (bspline::segment) to the edge's interval, so a circle's form gives
its exact pcurve on any plane. On a spline face (a NURBS form), an edge that runs along the face's
parameters at its own pace gets the straight (u, v) line it is: between its ends' parameters, or
with u or v its own parameter (a rim on the same quarter grid). Each is taken only where it lifts
onto the edge to within 1e-9 of the geometry's size; nothing is fitted.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BnQawySUFx57iaJZq7z1Fm
A form's parameter is now the angle's quarter on its frame plus the arc's parameter in it, and a
form spans the quarters its interval needs (up to five for a turn). Every form on one frame has
the same parameter at the same angle, so a circle edge and the cylinder it bounds agree along it,
and an arc across the frame's x axis, or starting anywhere, has a form. Traces its label to 1e-14
on such arcs, negative angles and a turn from 30 degrees.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BnQawySUFx57iaJZq7z1Fm
`brep::form` builds each analytic kind's exact B-spline the way SMLib holds its own: a circle
or ellipse as four rational quadratic quarter arcs from its frame's x axis on [0, 4] (angle and
parameter related in closed form), a cylinder or cone as that circle times a line along the axis,
a sphere as the circle times a half circle of latitude, a torus as the circle times its tube's
circle, a line as the degree-1 B-spline on its own parameter. A circle and every face it bounds
then share one parameterisation. Each traces its label to 1e-14.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BnQawySUFx57iaJZq7z1Fm
A shell's `forward` is about the surface the face's (u, v) are on, its NURBS form when it has
one, which may run against its cylinder, cone, sphere or torus. The kernel read it against the
analytic surface, so a fillet's bore whose form runs the other way was a shaft and an M10 thread
was refused in its 8.5 mm hole. `Face::form_turns` says when the form turns the normal.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Edge::curve_interval (and brep::curve_interval for assemble's edges) gives the interval on the
edge's curve itself: a form circle's angles of its ends, its frame turned by the SMLib conversion
so the angle grows along the NURBS. lip and the ruled surfaces read arcs by it; the hull, the
blend tangents, the patch pcurves, point paths, IGES output and the prism runs evaluate on
Edge::parameters. assemble carries an edge's form through.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The edge mirror of a face's NURBS form: Edge::nurbs keeps the NURBS SMLib holds a conic as, and
the edge's interval and its pcurves' parameter are its. Edge::parameters is the curve every
evaluation at a parameter uses; curve says what the edge is. Nothing writes the form yet.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Every SMLib analytic surface is a B-spline surface, and its trims are in that NURBS's
parameters. Face::nurbs keeps that NURBS: the face's domain and pcurves are its (u, v), and
`surface` says what the face is. Face::parameters is the surface every (u, v) evaluation uses.
The SMLib conversion reads and writes it (fork FACE_NURBS_FORM, SMLIB_REV c4483f6); the STEP
writer writes the analytic surface and the face's edges without pcurves.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
Loop2::circle and circle_from gave two half arcs, so everything built from a circle came out in
two: an extruded wall in two faces, a rim in two edges, and so every round, chamfer or sweep on
it in two B-spline halves that SMLib cannot join (two separately parameterised arcs meet at
different curve parameters, SmVertex::IsTopologicalVertex; different free-form surfaces are not
topological, SmEdge::IsTopological). A circle is now one arc of a whole turn from its start
angle: one closed edge, one wall with a seam, one round. Sketch circles, the cylinder builder
(circle_from at its seam angle) and every other builder take it through Loop2; SMLib's arc
curves take a whole turn (smlib_sweeps clamps to one) and ArcSweepCylinder makes the whole
cylinder.
Cases, singly on 9f5ad20c: fillets/1191, 2797, 3753 and sweeps/1269, 2295 fail before and pass
now; eight hole and six thread cases give what the base gives (12 pass, booleans/1279,
booleans/307 and threads/2038 fail as on the base).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
The ruled surface's far-curve solve kept its own copy of NURBS Book A2.3. The
geom crate now exposes bspline::basis_derivatives (the basis functions alive at
a parameter and their derivatives up to an order, each from its local knots via
local_basis, on the span tess::span_of picks); tess's second-derivative surface
evaluation and the ruled surface both use it, and the copy is gone. smlib's
mod list back in order.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
Intersection::point started its Newton solve from the last point found (a Cell holding both
surfaces' parameters) when that landed near the leader, as truck took a hinted point. The
answer then depended on what was evaluated before, so CurveEval was not Sync, and the
clearance check, which shares a body's DistanceShape (its edges' CurveEvals) across worker
threads, did not compile; with threads it would also have made the result depend on their
interleaving. Each point is now solved from the leader's own point, the path the hint fell
back to, and the unused trace tolerance goes with the hint.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
Each piece keeps the edge's curve on a sub-interval, a new vertex sits on the curve at each
split, every use becomes uses of the pieces with its pcurve cut at the same parameters
(bspline::segment), wire edges likewise; the result is in canonical order with each
edge's source edge. What a sheet needs so that another's vertex on one of its edges is a
vertex of both.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
part()'s walk becomes renumbered(): the record of given source regions in canonical order.
part() hands it the regions it cuts; canonical() hands it every region as it is, open,
wire and vertex shells included, for records builders put together in their own order.
Each edge's uses come from the walk; a source list naming exactly those uses keeps its
radial order (turned to start at the primary use), as part() did.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
A loop of N >= 5 edges is N four-sided patches about a centre point: each edge split at
the middle of its parameter interval, a spoke from each midpoint to the centre, patch k
on the second half of edge k-1, the first half of edge k and the two spokes beside them.
All patches share one compatible space, so each spoke is the same control points in both
patches it divides and the centre one point; the least-squares solve takes them once
and adds rows opposing the two patches' derivatives across each spoke (tangent there).
The one- and three-/four-sided fill is the one-patch case of the same model. The result
lists each patch's sides (edge pieces with their parameter ranges, or spokes) and the
spoke curves; every pcurve is a straight line in (u, v).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
The loop's edges become the sides of the unit square (three sides: the fourth collapses
to the first corner, a pole), each mapped affinely onto [0, 1]. Opposite sides are made
compatible exactly (split into Bezier pieces at every break, raised to the higher degree,
rejoined), the bilinearly blended Coons patch is the start, and with the boundary rows
held the interior control points solve the G1/G2 rows, the points to pass through and a
vanishing multiple of the thin-plate energy by least squares, refined by halving spans.
Every edge lies on the patch exactly and its pcurve is a straight line in (u, v), affine
in the edge's own parameter: no pcurve is fitted.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
compound concatenates the records and takes the part of all their material, so the result is
in canonical order with each face, edge and vertex numbered from its input. part now keeps an
island inside a void when both its region and the void's material are taken.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
Closed forms on cylinders, cones, spheres and tori; on a spline the basis functions' own
second derivatives (local_basis) through the quotient rule for the weights; zero on a
plane. A revolution has none (its profile evaluator carries first derivatives only).
The seam continuity measure reads normal curvatures from them.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
A lump of a body is the part of one of its material regions: its outer shell, its open and wire
shells and a void for each closed inner shell, renumbered in the record's canonical order, with
the source index of each face, edge and vertex for history. Islands inside those voids are parts
of their own. Read from the regions the record already holds, with no mesh or ray test.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
A rotation and a translation move every point, axis and control point and leave parameters,
pcurves, tolerances, orientation and the canonical order as they are: the same B-rep in other
coordinates. Placing a body (assembly instances, patterns) no longer needs a kernel round trip.
Reflections and scalings change parameters or geometry kinds and go to the kernel.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
EvalSurface::wrap asked uv::closed whether a spline closes on itself every
time it evaluated a point outside the knot interval, and that test walks the
whole control net (its extent, then the first and last rows). A tessellation
of a closed spline unwraps its parameters across the seam, so about half its
evaluations are outside: on a coil's helical tube (thousands of control
points) the test was 80% of the tessellation's time, 30 s of the coil
command on SMLib (sweeps/3832, sweeps/793).
EvalSurface::Nurbs now carries `closed`, set by EvalSurface::nurbs from the
same test; wrap and uv::domain read it. An affine map of the surface keeps it
(closedness does not change under one). The answers are the same as before. The constructor's callers in
crates/kernel/src/tess (size.rs, surface.rs) change with it.
Not built: base in progress (rebased onto kernel-smlib/no-truck 41adb372).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
A drawn 3D spline reached the kernel only as its sampled polyline (Wire.pts), so SMLib refused
to sweep along it ("not supported yet: SMLib sweep along a smooth path of points"). The doc
holds the spline's fit points; the curve is the centripetal Catmull-Rom spline that
fit_polyline samples. Each span of it is a cubic (Barry-Goldman), and its derivative at a fit
point depends only on that point, its neighbours and the two knot intervals, so neighbouring
spans share it: the spline is exactly one cubic Bézier per span, C1 across the spans.
- geom: catmull_rom holds the curve: samples() (fit_polyline's arithmetic, moved verbatim;
sketch::fit_polyline now calls it) and beziers(), the exact spans; ArcLength on CurveN
(Gauss-Legendre per knot span, Newton for the parameter at a length).
- kernel: PathSeg::Cubic, a cubic Bézier in space (the 3D counterpart of PathSeg::Curve, which
is planar), with its arms wherever PathSeg is matched (start/end, tangents, length, point at
a length, turn, SMLib curve as a degree-3 NURBS, halving). PointPath / PathPiece carry a path
through 3D sketch curves: polylines as points, drawn splines as their fit points (reversed
when the path runs them backwards); PointPath::segs gives SMLib the spans and the lines;
sweep_point_path_with sweeps it (SMLib through sweep_with, so corners, twist, scale, rails and
orientation take the existing routes). A smooth polyline piece (a projected or linked curve,
which the kernel only has as points) is still refused on SMLib.
- doc: path_points builds the PointPath (a wire with three or more fit points is a spline
piece); solid sweep and pipe call sweep_point_path_with; rails, surface sweeps and curve
networks take its points as before.
Checked in scratch (not committed): the spans pass through every fit point and every
fit_polyline sample at its parameter (1e-12), C1 at the joints and round a closed spline;
samples() is bit-identical to the old fit_polyline over 1180 random open and closed inputs;
SMLib sweeps of a radius-2 circle along an open spline and round a closed one match
pi r^2 L (L by independent quadrature of the curve's definition) to 1e-7 and 3e-8.
sketch/wire.rs: the existing closed-spline test reads SPAN_SAMPLES from its new home.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
uv::project, given no hint that converges within `accept`, evaluated a grid
of up to 16384 points of the whole surface for every point it placed, then
sorted all of them by distance. The grid is the same for every point of a
surface, and a face's pcurves place hundreds of points on one surface: on a
modelled thread's helical faces that grid was most of a boolean's time on
SMLib, in the operands' records, SMLib's export of each body and the
result's (threads/2526, the ladder's threads).
uv::Starts keeps the grid, filled on first use; uv::project_with takes it,
and a face's pcurve fit (pcurve::On) keeps one per face. The 32 nearest
starts are picked with a partial select instead of a full sort, in the same
order (nearest first, grid order where equally near), so every point gets
the parameters it got before. (The tessellation's boundary projection can
take the same grid; crates/kernel/src/tess is the mesh agent's.)
Not built: base in progress (rebased onto kernel-smlib/no-truck 41adb372).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
On a B-spline face the export now takes each edgeuse's UV trim exactly as
SMLib holds it (EDGEUSE_NURBS_*): the face's parameters are SMLib's, so the
trim is the pcurve, reversed exactly (knots t -> a + b - t) where the
edgeuse runs against its edge, its knots mapped onto the edge record's
parameter where that is an affine image of SMLib's (identity for spline
edges, as_line's map for NURBS lines, the C side's map for lines), and its
end knots set to the edge interval where they differ by rounding (within
the schema's 1e-9 slack, else an error). Changed: export.rs `curves`
(per-edge TrimMap), new `trims`, `mapped`, `Face::given`,
`Edge::trim_map`, the face loop in `native_ordered`, `pcurves`,
`outer_area`.
Analytic faces and circle/ellipse edges keep brep::pcurve: SMLib keeps an
analytic face's trims on its internal rational B-spline (SmCylinder.h:
"the map from one to the other is not linear") and a circle's on its NURBS
parameter, neither an exact B-spline in the record's angle parameters.
brep::pcurve::EdgeUse gains `given`, a producer's exact pcurve taken as it
is (structure and interval checked, moved by whole periods with its loop,
never refitted). Existing construction sites (STEP reader, tests, truck
adapter) pass `given: None`; the existing tests change only for the new
field.
Built, clippy-clean and tested on 4b2c0eaf before the rebase onto
kernel-smlib/no-truck (not built: base in progress).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
On a thread's root band (1.4 mm tall at a pitch of 1.5) a point on the band's lower rim lies
0.1 mm from the rim of the turn below. The search's starts are dense across the band and
sparse along it, so its four nearest starts all lay on the turn below, Gauss-Newton settled
there, and the pcurve fit refused the edge ("the curve at 0 lies 0.1008 mm off the surface",
ladder/06-threads/007). The search now goes on through the nearest 32 starts until one
converges within the accepted distance, and stops at the first that does.
The test strips now put their knots at other angles on each turn (7.3 control points a
turn), as SMLib's screw sweeps do.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
SMLib's intersection curves carry knots of full multiplicity where the curve is in fact
differentiable (its derivative the same either side). The pcurve fit took every such knot
for a corner and fitted the two sides apart; on a thread's flank, whose u runs hundreds of
times as fast as its v, the pieces met at up to 1.4 deg in (u, v), and SMLib's check refused
the record read back (Test #7 Bad CurveKink, offset 0.9997 against cos 1 deg).
A knot that leaves a spline only continuous by its multiplicity now gets a cubic with a
tangent there when the spline's one-sided derivatives (EvalCurve::sides) agree within 1e-9
of their size. Knots where the curve does turn are corners as before.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
A thread cut by SMLib's screw sweep leaves flanks that are cubic splines of hundreds of knot
spans along the helix (an M6 through 20 mm: about twenty turns over u in [0, 1]). With no
hint (an edge's first sample), the projection started from a 17 x 17 grid over the parameter
box: one start every turn and a quarter, so the nearest starts lay on the turns beside the
point's own, Gauss-Newton settled there, and the pcurve fit gave up with "the curve at 0
lies 0.87 mm off the surface" (pitch x cos 30 deg, the next turn's flank). Every modelled
thread failed its export that way.
The search now starts, along each direction where the surface has its own knots (a
spline's, a revolution's profile's), from each knot span's start and two points inside it as
well as the 16 even steps, capped at 16384 points of the surface, so every turn is searched.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017itGh5fpMEousnZ91cYcVt
Where two edge uses of a loop meet at a pole (meridians of a sphere), the periodic parameter's
jump there is free, and the pcurves took whichever jump made the loop counter-clockwise. Two
meridians a three-quarter turn apart bound a quarter of the sphere or three quarters of it, and
the loop alone can't tell which: a half disc revolved 270 degrees on SMLib came back as a quarter
ball turned inside out (sweeps/2667, V = -226 against 678.6).
pcurve::pcurves_within takes the face's parameter box, as SMLib gives it, and picks the jumps
whose extent in that parameter is nearest the box's; pcurves is unchanged.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
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
SMLib bounds an intersection edge on a piece of its spline (seed 32 intersect:
a block and a cylinder, the top disc's edges), and the truck decode refused
such an edge ("not supported yet: a spline or intersection edge on part of its
curve"), since truck bounds an edge by its curve's whole range.
bspline::segment cuts a B-spline to a sub-interval of its domain exactly:
Boehm insertion of both ends until each has multiplicity p, then the knots and
control points between them, clamped on the interval. The same points at the
same parameters, so the edge keeps its interval and the record its numbers.
load_edge_curve loads a spline or rational spline edge on part of its curve as
that piece; an intersection edge on part of its curve is still refused. Our
own records (interval = the curve's range) load as before.
Tests: segments of a rational cubic with a double knot, inside a span, across
knots, from or to a domain end, match the whole curve within 1e-12; both
spline kinds load on a piece with that range and the same points, and beyond
the domain are refused; SMLib's block ∩ cylinder decodes (it was refused) with
π r² h within 1e-3.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
The adapter loads an analytic torus as truck's revolution of the whole profile
circle, a rational quadratic closed on itself, so an SMLib torus is one face
bounded by its two seam circles, each used both ways. The display evaluator's
domain gave a revolution a period around its axis only: projected along the
profile the loop jumped from one end of the knot interval to the other at its
start vertex, the plan failed ("boundary crosses itself") and the body fell
back to truck's mesher, which meshes a sphere bounded by its seam alone as
nothing (seed 60 union, a torus beside a sphere: 925.0 of 990.7 mm³).
tess::uv::closed_curve detects a closed profile the way closed does a closed
spline surface (clamped knots, first and last control point and weight the
same within 1e-9 of the polygon's extent; the shared checks are factored out),
the domain then has period = the profile's knot interval along it, and
EvalSurface::wrap/derivs wrap that parameter back into the interval, so the
plan's projection, unwrapping and both evaluators treat it as periodic. A
cylinder's or cone's line profile is unchanged.
Tests: a revolved whole circle, either parameter order, has both periods and
evaluates a period away as inside; SMLib's torus, alone and beside a sphere,
decodes without meshing to orient it, plans every face and measures within
1e-3 of 2π²Rr² (+ 4/3 πr³). With the closed profile undone the torus plan
fails as above.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
truck lists some planar faces' hole loops before the outer loop, so the
rule that loops[0] is the outer loop never held for truck's records; moving
it in the adapter would renumber those faces' edges, which history and
names stand on. The outer loop is the one that runs counter-clockwise about
the natural normal (what check_geometry checks); the order is the
producer's, and SMLib's import puts the outer loop first, as SMLib needs.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
check_geometry compared pcurve ends with edge points only, so a loop whose
pcurves enclose nothing passed whenever its uses lift onto the same 3D points:
SMLib's hole-rim fillet with both seam uses on u = 0, or a whole sphere whose
pole-to-pole seam has both uses on one line (the continuity check skips the
pole). Such a face has no surface and meshes to nothing.
Each edge loop's signed (u, v) area is now taken over 16 points of each use's
pcurve: a loop within 1e-9 of the square of its extent of zero is refused
(Problem::Empty), and each face must have exactly one counter-clockwise loop
about the natural normal, the rest clockwise (Problem::Turn). Outer and inner
are told apart by area rather than position, because the truck adapter writes a
face's wires in truck's order, which puts a hole first on some planar faces of
the sample designs (the model cache test caught it).
Fixtures: the pole-to-pole seam on one line, and a sheet whose plane axes are
swapped so its loop runs clockwise; a seamed sphere passes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
A spline surface whose first and last control rows coincide in a direction
(points and weights, knots clamped there) closes on itself, but the display
evaluator's domain gave NURBS no period. The pcurve routine therefore never
moved a seam's second use by a period: SMLib's fillet round a hole's rim (a
rational spline closed in u) got both seam pcurves on u = 0, its outer loop
enclosed no (u, v) area, the decoded face tessellated to nothing and the body
measured 6839.73 mm3 against SMLib's 6989.00.
tess::uv::closed detects the closure from the record, within 1e-9 of the
control net's extent; the domain then has period = the active knot interval's
width, and the evaluator wraps a parameter outside that interval back inside
it, so pcurves, projection, sizing and the display plan all treat the direction
as periodic. The record stays as it is. The display plan hands both evaluators
wrapped parameters (the GPU does not wrap).
Tests: a rational torus band, a NURBS cylinder and a NURBS dome with a pole,
each bounded by a seam used twice, get seam pcurves one period apart; a record
of the band passes check_geometry and its seam uses on one line are refused;
SMLib's hole-rim fillet decodes with every face meshed and SMLib's volume
within 1e-3.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
Schema 4 counts every pcurve's control points as items, which schema 3 had
none of, so a body that fitted the old 1,000,000 no longer does (a STEP
plate of 50k faces). The binary envelope's own caps on bytes, values and
entries per collection still bound a body.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT
A whole sphere bounded only by its seam (SMLib's sphere: one meridian
used twice) came out with both uses on the same u line, a loop that
encloses nothing. Two causes in the per-loop closure:
- The pole-junction search took any positive signed area as the right
turn. With the two uses on one line the area is rounding, and in most
frames that rounding is a tiny positive number, so "no shift" won over
the period shift. A turn now needs more than 1e-9 of the loop's
period box.
- Junctions counted as poles only where the surface collapses exactly.
Ends within the tolerance of a pole (as a kernel's records have them)
were closed to the nearest period instead. A junction is now free
where moving its periodic parameter by half a period stays within the
tolerance, and fit samples within a quarter of it take their free
parameter from a neighbour, so the pcurve runs straight into the pole.
Tests: the sphere bounded by its seam in four frames, three meridian
angles, circle and NURBS meridians, both loop orders, ends at or a tenth
of the tolerance off the poles; and a torus bounded by its two seams.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QWhMwdqJYQKQq8RCLSL1LT