Decide whether a search limit explains the reported count
A floor planner runs in the browser and must limit its work. When a large problem reaches that limit, the returned candidate is useful evidence of placements found, but not a proof that no more fit. The first review question should be whether a cap was reached and what mathematical bound remains. Do not interpret a long wait, a finished spinner, or a capped count as an impossibility proof.
The local engine uses deterministic mixed-strip enumeration and several MaxRects rules. Its placement budget is capped at five hundred. It does not expose an arbitrary timeout, iteration count, or “search harder” query parameter in the public tool. Adding such an unsupported parameter to a link does not modify the actual algorithm. The comparison below is a synthetic geometric stress exercise, not a container specification or practical load recommendation.
Compare two floors that both return five hundred
Use identical 100 by 100 mm square footprints, zero gaps and clearance, and allowed rotation. A 2500 by 2000 mm floor fits twenty-five squares along length and twenty across width: five hundred. That valid grid equals its area upper bound, so the engine returns five hundred with optimal status and an area proof. The cap flag is false because no higher area ceiling remains.
A 3000 by 2000 mm floor has room for thirty by twenty such squares, a direct six-hundred-square construction. The current engine nevertheless retains only five hundred placements because of its work cap. Its area bound stays six hundred, its status is best found, and its cap flag is true. The same returned number therefore has different evidential meaning in the two runs.
| Floor | Footprint | Returned count | Area bound | Status | Cap reached |
|---|---|---|---|---|---|
| 2500 by 2000 mm | 100 by 100 mm | 500 | 500 | Optimal | No |
| 3000 by 2000 mm | 100 by 100 mm | 500 | 600 | Best found | Yes |
Reproduce the work-cap example with supported inputs
Open the five-hundred-bound floor and the capped six-hundred-bound floor. The 20ft preset activates the planner, while the edited floor dimensions and tiny synthetic footprints define the stress test.
Review the floor result panel, not the main case-to-pallet preview. The main case dimensions in these links make that separate panel valid but do not control the floor count. The floor engine uses the pallet length and width fields as its repeated rectangle. Keep the object definition explicit whenever the same interface combines more than one planning stage.
Recognize what determinism does and does not provide
For unchanged numeric inputs and the same engine revision, the returned count, status, and coordinates should repeat. That allows a colleague to reproduce the candidate and inspect its validity. It does not imply that a finite heuristic search considered every possible arrangement. Repeating the same run several times is not accumulating independent proof of optimality.
In the larger square example, a manual thirty-by-twenty construction demonstrates six hundred without asking the browser for a larger budget. That external construction is separate evidence and should be identified as such. The public tool still returns its capped candidate. Do not rewrite its status to optimal merely because a different proof has been supplied elsewhere; retain both the tool record and the independent construction.
Preserve a mathematical ceiling when the search is capped
The area bound is not replaced by five hundred just because the work cap is five hundred. That distinction prevents a resource limit from masquerading as a capacity limit. In harder rectangle problems, the upper bound may be loose, so the unsearched difference does not necessarily correspond to missing feasible placements. The square stress test was deliberately chosen because its larger construction can be checked directly.
The cap flag describes this implementation's work limit rather than elapsed seconds or a hardware benchmark. A slower computer can take longer without changing the deterministic count. A responsive browser interface can also finish a capped problem without reaching an optimum. Record the flag and method, not an informal impression of how thoroughly the spinner searched.
Choose a next step suited to the actual decision
If a capacity decision depends on finding more than the valid returned count, obtain additional checked placement evidence or a suitable independent optimization review. Do not simply load the upper bound. If the returned candidate already satisfies the order, its valid coordinates can be reviewed without claiming maximality. Separately confirm real equipment dimensions, the completed load envelope, access, and weight distribution.
Do not split a difficult floor into arbitrary regions and sum their independent optima as though that proves the full-floor maximum. A disjoint partition with valid placements can supply a feasible construction if all boundaries and required separations are checked, but it may exclude better cross-boundary patterns. Clearly state the partition and the claim it supports.
Search-budget evidence record
Synthetic floor 3000 by 2000 mm; identical footprint 100 by 100 mm; rotation allowed; zero gap and wall clearance. Returned count 500, area upper bound 600, best-found status, placement cap reached. Save coordinates and engine revision. Independent hand construction: thirty columns by twenty rows gives 600 under the same zero-spacing geometry. Keep that construction separate from the browser result. No real container, carrier capacity, or loading operation is represented by this stress exercise.
The methodology provides the algorithm boundary. A resource-limited search can answer what was found; a capacity maximum needs a supported proof or independent evidence beyond a completed run.