kerfmaster pin list · Nesting
recordStates what the code does today. Auditable against the code — this is the only type that can go stale.
Content last changed 2026-09-01 — computed from the item itself, not typed.
What it solves (Jordan 2026-08-31, his MetaCam Edit Array walkthrough): “I flip one part, place it in the correct spot so that the leadin space is shared, and then I copy the whole left column…” The ‘correct spot’ is now computed, not eyeballed: minimal (dx, dy) for the 180 column such that every ruled clearance holds against the 0-deg column — outline-to-outline gap, both columns’ pierce clear-air, both lead cuts — checked across every interacting row offset, then the bare pair-to-pair interface closed to the ruled gap the same way.
Why it wins: both leads point INTO one channel per pair (0-deg +X, flipped −X, interleaved in Y by dy), so the lead+standoff channel is paid once per pair instead of once per column; a recessed true outline at the pierce heights lets the columns tuck tighter still. With standoff = lead length and flat faces the two pierce lines coincide — the 2026-08-27 ‘pierce points align in Y’ observation drops out of the solve.
Where: nest_rules.tuck_alt_flip (solver) + fill_y_first(alt_flip=True, tuck=...) (consumer); test_tuck.py (22 controls). The three standing objectives all land: least material (pair pitch minimized), consistency for the person breaking out parts (strict pair pattern), shortest editable G-code (two regular sublattices = two P9997 arrays, one value each).