Part-list nesting — Qty / Rotate / Bridge / Filler columns (Jordan 2026-08-31)

kerfmaster pin list · Nesting

rulingmixedA decision by Aristide or Jordan. True because it was decided; it can be superseded, but it cannot be stale.

Content last changed 2026-09-21 — computed from the item itself, not typed.

Source (Jordan, 2026-08-31): two MetaCam screenshots — the Parts list (Math_Nest_2.JPG: 21 rows, MSO7 0.375, Qty/Rotate/Bridge circled, Priority crossed out) and the Patterns tab (Math_Nest_1.JPG: 5 sheets 120x60, usage 75.6/69.3/66.8/55.3/44.3%). 'There is a third part of the equation called quantity. The fourth part is rotation, and the fifth part is bridge... you look at the quantities, you look at the rotations, you look at the bridges, and then come up with the best possible nesting where all of these are applied.'

The columns, as ruled

• Qty — ordered count per part; qty 0 keeps the part in the list but out of the run (row 21, PSI09-05QA-01). Quantities split across sheets however the pack works best.
• Rotate — rotation STEP: '360 is locked, 180 is 0/180, etc.' (so 90 = 0/90/180/270). This is the storage form of the grain-direction / orientation-lock click already ruled (free-rotation toggle, rotation-follows-the-print).
• Bridge — seeds from the table min_gap for the material; operator overrides per part (his example: the two 83-inch plates dropped to 0.09 while the rest ran 0.15). Semantics ruling: in MetaCam it's .15+.15=.3 (each neighbor contributes half); 'this is confusing and we should just have the total gap “around” that part.' So our column holds the FULL clearance that part demands around itself. CONFIRMED (Jordan, same day): 'Larger governs, however you have to keep track per item still. And if your sheet count goes up because everything is “larger governs” then you need to start hunting for areas where you can improve by applying the smaller bridges (part gaps).' So each pairwise gap = max(bridge_i, bridge_j), tracked per adjacency — and the asymmetry is an optimization lever: grouping small-bridge parts next to each other recovers the space that larger-governs costs. If larger-governs pushes the sheet count up, that hunt is required, not optional. Edge clearance too (Jordan, same day: '1. Yes, also the sheet edge.'): the part's bridge value governs its clearance to the SHEET EDGE as well — a 0.09 plate may sit 0.09 from the edge while a 0.30 part keeps 0.30. Under part-list nesting the margin is therefore PER PART, refining nest-margin-from-min-gap (whose table-derived margin is the seed each bridge value starts from).
• Filler — KEEP: 'we will use it when we can indicate which part can be duplicated more just to fill the sheet (this is going to depend per customer, if nothing is clicked we don't want you to fill the sheet, just maximize Y and keep the most amount of X as a remnant).' Per-part flag, per-customer habit; the no-flag default restates the standing remnant-slab rule. Sits beside nest-finish-even-columns (that one rounds qty up to even columns on operator click; this one authorizes open-ended duplication of flagged parts). Confirmed (Jordan, same day: '2. Yes, agreed.'): filler extras are recorded asked-qty beside filled-qty on the delivery sheet — extras visible, never silent — and obey every per-part rule (bridge, rotate) exactly like ordered parts.
• Priority — crossed out: 'Ignore, we will not need that.'

Terminology guard

MetaCam's 'Bridge' column is a PART GAP — it is NOT our bridge (permanent island-holding feature, PINLIST 2b). Treat like 'wirejoint': their word recorded, ours kept.

Benchmark

'Not necessarily beat it, because there are so many different ways that you can organize these files, however, do not come up with a plan for 7 sheets.' 340 parts (sum of Qty column; the Patterns-tab Parts column reading 663 was Jordan's confirmed error), 5 sheets of 60x120 3/8, MetaCam usage falling 75.6→44.3% toward the tail — the tail is where pocket-backfill (nest-pocket-backfill) earns its keep. All always-apply rules on per no-rule-dropping-to-beat. Scope (Jordan, same day): 'the target is only for this specific job' — 5-not-7 is not a standing sheet-count formula. And this job posts NOTHING: 'We are not going to ask you to actually create a dxf for this job. it was simply for education purposes. We will give you a real example that we can use as testing' — the LH-2.5 list was chosen because it exercises all the variables at once.

FILL RUNS LAST + FILL EVEN (Jordan, 2026-09-21 -- his 26342 catch: Fill ticked turned 50 asked into 96 parts on 2 full sheets, 'when "Fill" is selected ... very confusing. My goal was to fill the sheet with nice even rows'; his 'Go' on fill-runs-last + the fill-even option): the filler pass used to run right after first placement, BEFORE the spill rescue -- an asked quantity that momentarily spilled two parts onto a second sheet saw fill top BOTH sheets to capacity. Fill is elective extras, so it now runs as ONE unified pass after consolidation and the pocket rescue, before the frontier tuck (fill re-spends the remnant the tuck would free, and topping the asked partial column keeps the grid whole): extras only top up sheets the ASKED quantity kept -- fill can never multiply sheets -- and a coarse judge pack never fills at all (the strategy selector ranks the asked layout; extras ride only the winner's full pack). Row-strategy fill places through try_rows and every accepted spot is ALSO verified against the rescue's pocket placements (row bookkeeping cannot see them) through the same _pkt_ok arithmetic -- a miss undoes the spot whole and stops fill on that sheet, conservative, never an overlap. NEW fill_even (pack(fill_even=), select_pack passes it through, report echoes it -- the finish-even-columns ruling applied to filler extras): extras land only as COMPLETE columns -- the asked partial column tops to full height when it is wholly the same part, then whole new full-height columns, so every column holds the same count. UI: 'fill even rows' checkbox on the Part list tab; run_partlist persists the pick, run_partnest hands it to the pack. Honest limits stated in the docstring: freeform fill stays first-fit whatever the flag; a foreign partial frontier column stops that sheet's even top-up. Measured old-vs-new on the 26342-reproducing kit: OLD 2 sheets, 120 filled; NEW 1 sheet, filled == asked (the spill rescued into sheet 1's remnant -- exactly his screenshot-1 layout); fill_even on the control kit = 22 even columns x 11 each. Proven test_part_list 28 -> 37 controls (fill-runs-last on the saddle rescue kit, coarse never fills, fill_even even grid + audit, free fill still fills, select_pack hand-through) + pl_select 21, pl_macro 25, pl_sync 23, pl_free 42 zero edits + verify_plsel, plrefresh, refusal, pltabs, plhand. Zero dev O-numbers.