[Oliver-Johnson]

Requested by Oliver · project thread

What this changes

Two kinds of baseplate came out with a sliver missing from a flat face, so the downloaded mesh had a few open edges in every cell.

Before

• Baseplates: a screw bore of nearly the magnet pocket's width leaves 6 to 12 open edges a cell at 42 mm #92: a magnet pocket with a screw bore nearly as wide as the pocket, at 42 mm, for example a 6 x 2 magnet from beneath over a 3 mm hole with a 4 to 5.95 mm head 3 deep: 6 to 12 open edges a cell on the bottom face. 669 of the 3,332 designs in the issue's two families were open.
• Baseplates: corner pockets with screws leave 3 open edges a cell at 34.32, 34.73 and 38.95 mm #91: corner pockets with a 2 mm screw at 34.32 mm, or a magnet from above at 34.73 or 38.95 mm: 3 open edges a cell on a corner boss's top, 6 to 15 on a 2 x 2. The same on a boss's underside with a 4.9 mm magnet from beneath at 43.9 mm. On main as of A corner pocket's boss follows a rounded plate corner instead of standing out past it #90, a one-cell piece at those sizes builds closed (its corner boss is cut to the arc and closes on the second try); a 2 x 2 piece still comes out open.

After

• All of those build closed, except one head size (6 x 2 magnet, 3.909 mm head) whose open edges are on the pocket's roof, a different face (filed with four others as Baseplates: five designs the #91 and #92 sweeps left open or folded, on faces other than a flat cap #95).
• A design that already built closed is the same, byte for byte, with one exception that comes from A corner pocket's boss follows a rounded plate corner instead of standing out past it #90 (below): 66,497 of 66,497 across the sweeps on the base before A corner pocket's boss follows a rounded plate corner instead of standing out past it #90, and 766 random everyday designs.

How

• A flat face is a fan of triangles from one point. Where a fan line passes a few microns from a corner of a pocket's wall, the cut loses the sliver between them.
• Bottom face (Baseplates: a screw bore of nearly the magnet pocket's width leaves 6 to 12 open edges a cell at 42 mm #92): the fan point had to stay clear of every point on the pocket walls, including the extra points left where two bores were merged. With about 140 of them no point was clear and the fan stayed in the middle. Now a cell that still comes out open on its bottom face is fanned again clear of the corners only, and its cuts are taken again.
• Boss caps (Baseplates: corner pockets with screws leave 3 open edges a cell at 34.32, 34.73 and 38.95 mm #91): a boss's top and underside were fanned from a corner of its outline. Now a boss that still comes out open has that cap fanned again from a point clear of its pocket's corners (refanCaps), and its pocket cut again.
• Either way the new cut is kept only if it is closed, nothing is turned over, and it shares no more edges with its neighbours. Nothing that already came out closed takes this path. It costs about 10% more build time on a design whose bottom face is refanned and about 8% on one with a boss refanned, and nothing elsewhere.
• docs/ENGINE.md says what a refan does and when it is taken. test/plate-audit.js has 19 new rows. 13 of them fail on main 7c2103d (all 9 for Baseplates: a screw bore of nearly the magnet pocket's width leaves 6 to 12 open edges a cell at 42 mm #92 and the four 2 x 2 rows for Baseplates: corner pockets with screws leave 3 open edges a cell at 34.32, 34.73 and 38.95 mm #91). The other six, the one-cell Baseplates: corner pockets with screws leave 3 open edges a cell at 34.32, 34.73 and 38.95 mm #91 rows, failed on 1e1c48a, where the fix was made; A corner pocket's boss follows a rounded plate corner instead of standing out past it #90 has since cut their outer-corner bosses to the arc and they close on its second try, so they pass on 7c2103d too and stay as rows that pin it.

Merged with #90 and #93. A boss at a rounded plate corner is now cut to the arc (#90), and its cut is taken only if it comes out closed with nothing turned over. The refan runs inside that test, so a cut boss that came out open on its first try is fanned again before the test, and not only after it has failed. 7 of 2,266 designs that were closed on main (7c2103d) build other bytes for that reason: their corner boss was cut to the arc on both, and closed on main only after the second try with the cutter turned a 28th of a turn, and now closes on the first try. Same open, four-use, fold, wound and inverted counts on both, and the boss is cut to the arc on both (checked at the corner).

Two small ENGINE.md wording changes review asked for on #90 ride here: the chords' sag can put a bore up to 0.021 mm under MOUNT_WALL from the outline, and no audit row asserts the joint half of the corner test.

The full write-up is /mnt/project-files/issue-batches/issue-91-92-report.md.

Checks

• I edited the sources in src/, not the generated index.html / bins/index.html
• I ran node build.js and committed the regenerated pages alongside the sources
• npm run test:all passes (run here on 9b621c0, the merge with main 7c2103d: npm test exit 0 with 0 FAIL, and the plate UI specs plates-input, plates-robustness, export, keys and half-cells 186 passed. The two commits after it change only ENGINE.md and a comment in plate-audit.js; node build.js --check is clean and node test/ci-sim.js says "CI will pass." on the head. The whole UI suite is left to CI)

If this touches geometry

• Bins: node test/bin-audit.js reports all cases clean
• Baseplates: node test/plate-audit.js reports all plates watertight, and no case has picked up bad edges that were not already named as known (the new 2 x 2 rows keep the quarantined edges between bosses, the same count as main)
• I added a case covering the new shape: 19 rows, 9 for Baseplates: a screw bore of nearly the magnet pocket's width leaves 6 to 12 open edges a cell at 42 mm #92 (the other limits, built at their ends) and 10 for Baseplates: corner pockets with screws leave 3 open edges a cell at 34.32, 34.73 and 38.95 mm #91 (corner bosses beside a joint cut from beneath); 13 fail on main 7c2103d and all 19 pass here (the other six are explained above)

Sweeps, each design built on the head and on its base in one process, comparing every piece's triangles byte for byte:

• Before A corner pocket's boss follows a rounded plate corner instead of standing out past it #90 and Pin the sizes issues 77 and 80 reported in the plate audit #93 were merged (base 1e1c48a): 67,205 designs in nine sets. 708 open on main; 706 close, 2 stay open (the 3.909 mm head above and a random design with a 1 mm magnet and a snap joint, in Baseplates: five designs the #91 and #92 sweeps left open or folded, on faces other than a flat cap #95); 0 worse in open edges, folds, inverted or wound shells; 0 of 66,497 closed designs change bytes.
• After the merge (base 7c2103d): 2,969 designs (every design from those sets that was open or folded, 920 random designs, and about 1,300 spread across the Baseplates: corner pockets with screws leave 3 open edges a cell at 34.32, 34.73 and 38.95 mm #91 and Baseplates: corner pockets with a 2 mm screw hole leave open edges under the bosses at 34.04 mm #80 grids). 703 open on main; 698 close, 5 stay open (the same two, and three below the page's 1 mm minimum that it refuses); 0 worse in open edges, folds, inverted or wound shells; 0 with more four-use edges; 7 of 2,266 closed designs change bytes with the same counts (above).

Closes #91.
Closes #92.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Uqqb1cvhXxrpb3P6y6ykBY