civicus-build
c7e32d8e07
Reference implementation of the strap-beam generators at revision 8.0.0, kept so the acceptance oracle can be regenerated. Not a live target; the running application has no OpenSCAD dependency. The oracle holds 123 frozen cases, 113 accepted and 10 rejected, produced by OpenSCAD 2021.01 with BOSL2 at 92d697c2. The ten rejections are part of the contract: a port that accepts them is wrong. tests/test_oracle.py specifies the port API and was written before the port, so the interface follows from what must be verified rather than what is convenient to implement. Proven by adversarial stub: a build() that rejects everything passes all 10 rejection tests and fails all 226 acceptance tests.
Reference implementation — OpenSCAD, revision 8.0.0
This is not a live target. It is kept so the acceptance oracle in
fixtures/strap-beam-8.0.0/ can be regenerated, and so the port has something
to be checked against. The running application has no OpenSCAD dependency, and
none of this is installed on an instance host.
What it produces
Eleven profiles across two generators. A profile is a cross-section swept along a longitudinal axis; a "member" is the end view of one strap bundle, so its length in the cross-section is the strap's width, never the beam's length.
| Generator | Profiles |
|---|---|
strap-beam-3x.scad |
Equilateral Triangle, General Triangle, A Frame, Y, T, Three-Fin |
strap-beam-4x.scad |
Square, Rectangle, Diamond, Cross, Four-Fin |
4x is the end of the N-strap family. Higher polygon counts compose from triangles and squares rather than getting their own generators.
The shared library
Everything reusable is in lib/, written for N members and used unchanged by
both generators:
| File | Contents |
|---|---|
sb-geom.scad |
GEO and MEMBER records, path construction, exact polyline distance, the monotone solver |
sb-join.scad |
butt joints, hull caps, concave-corner fillets, ring envelopes and bores, polygon fitting |
sb-report.scad |
value-based checks, section metrics, the SB_KEY=value report |
sb-profiles.scad |
three complete N-generic arrangements: ring, spokes, fins |
sb-core.scad |
umbrella include and the PROFILE record |
strap-beam-4x.scad is 210 lines and contains no geometry — five one-line
calls into the shared library. The same three calls with N=3 produce the
triangles, the Y and the Three-Fin. That is the evidence the split is real.
Design rules the library enforces
- Declared webs are cavity-to-cavity. A stated 1.2 mm web is 1.2 mm of plastic.
- Placement that cannot be derived exactly is solved numerically against the measured quantity, not approximated with a closed form.
- Junctions are structural before they are decorative: members butt through their neighbours, and fillets are applied on top of that overlap.
- A member with no enclosed side takes the outside wall on both faces, so asymmetric wall settings never make a symmetric profile chiral.
- Validation measures the finished section. Connectivity is necessary and never sufficient — revision 7 passed cross-sections joined by 0.13 mm.
Running it
Requires OpenSCAD 2021.01 and BOSL2 at 92d697c2. Use the pinned image:
docker run --rm -v "$PWD:/repo" mechcomp/reference-toolchain:8.0.0 \
openscad -o /dev/null --export-format=asciistl \
-Dprofile_type='"Three-Fin"' /repo/legacy/openscad/strap-beam-3x.scad
Every render emits a flat SB_KEY=value block on stderr, terminated by
SB_END=1. render_mode="Section" emits the 2D region instead of the swept
solid, which is what --export-format=svg consumes.
Three BOSL2 behaviours worth knowing
Found the hard way; the library works around all three:
- Boolean functions return a bare
[]rather than an empty region. offset()on a single-part region returns a bare path rather than a region.round_corners()raises rather than reporting when a radius will not fit.
sb_as_region(), sb_area()/sb_nparts() and sb_path_max_round() exist for
exactly these.