Files
OpenNest/docs
aj 277719a1d6 fix(planning): re-plan the last kept part when a learned rule changes what follows it
When a blocked approach teaches an order rule, the search backs up to the
earliest part that approach crossed, keeps the parts before it and
re-plans the rest. The last kept part was planned facing the part that
followed it in the old order. If the new order puts another part after
it, its entry and exit still face the old one, so the plan ends with a
stale exit that a kept-order re-plan of the same order would fix.

Now that part is planned again, in the same place, facing its new
successor; when its successor is unchanged nothing is re-planned.

The locked-among-regenerated fixture's plan changes: rapid 67.60 ->
60.22, expansions 528 -> 268 (digest updated). Learning "cut the
cutoff first" now takes 40 expansions rather than 22, because B is
planned again to face the cutoff; retrying every entry of B took 60.
2026-11-01 16:51:09 -05:00
..