Files
jyviko 95817e647a Sketch: a control point spline can be given its own knots and weights (#19525)
* claim: #19337

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DtL5LzWFYXTcbQ43JHhvpP

* Sketch: a control-point spline can be closed into a loop (#19337)

A control-point spline with "periodic": true is the closed uniform B-spline
over its control polygon: the wrap takes the first degree points again after
the last. It is refused on open before this change ("damaged design: invalid
periodic spline or revolve convention"), so a smooth closed outline given by
its control vertices could not be drawn.

The closed spline becomes one cubic (or higher-degree, up to the kernel limit)
Bezier piece per control point, joined C(degree-1) around the closing point. The
pieces come from the existing knot insertion (bezier_spans), now run over the
domain ends too, so the open knot vector gives Bezier pieces. The seam point is
set exactly. The kernel and profile finder see the same exact pieces as a
periodic fit spline.

- sketch: closed_control_beziers, closed evaluation in curves, validate_closed_control
  (degree 3 to 9, at least degree + 1 points), solver and point editing guards.
- loader: a closed control spline is accepted by its own rule; pending format
  change closed-control-splines.rs (the last release refuses such a design).
- command: sketch.spline.control_point and control_point_5 take periodic.
- UI: the closing click works for control splines (degree + 1 distinct points).

Test change: format.rs asserted control:true on a periodic spline is malformed;
it is now the valid closed control spline, so the malformed case is control:"yes"
and a closed control spline is asserted to load. sketch_modify_tests refusal of a
periodic control spline on four points becomes three points (still too few).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DtL5LzWFYXTcbQ43JHhvpP

* claim: #19524

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DtL5LzWFYXTcbQ43JHhvpP

* Sketch: a control point spline can be given its knots and weights

A control point spline could only be drawn with uniform knots, so a B-spline
with a non-uniform clamped knot vector had to be refitted. The model already
stored a spline's own knots and weights (projected free-form edges and DXF
splines), but no command created one.

- sketch.spline.control_knots: degree (1 to 9), knots and optional weights
  over the points. The curve is exactly that B-spline, drawn and extruded as
  given (Bezier spans by knot insertion, conics for a rational quadratic).
- A weighted spline must have degree 2: a higher degree with unequal weights
  has no exact profile form, so it is refused rather than drawn as a polyline.
- Knots that do not fit the points are refused as bad parameters, naming the
  rule (count, clamping, repeated knots, weights).
- A control point of such a spline cannot be removed; dragging one keeps its
  knots. The saved form is the existing spline-knots form, so no format change.

Test change: none.

Co-Authored-By: Claude Haiku 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DtL5LzWFYXTcbQ43JHhvpP

* Sketch: translate the knot spline command into the complete catalogs

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DtL5LzWFYXTcbQ43JHhvpP

* Kernel cases: the knot spline command is tested in its area's file

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DtL5LzWFYXTcbQ43JHhvpP

---------

Co-authored-by: jyviko <2501648+jyviko@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-19 17:26:03 +00:00
..