Public solver record · v1.0.0

Cut plans should be reproducible, not mysterious.

These fixed inputs make SawNest's boundary behavior inspectable. Each case states the stock, constraints, expected result, and the specific rule it protects.

Download benchmark JSON
4public cases
2D + 1Dplanning engines
18automated launch checks
CC BY 4.0dataset license
B01Cabinet sheet sample
Input
Seven mixed cabinet parts on 48 × 96 in stock
Constraint
1/8 in kerf · 1/4 in edge trim · cabinet-side rotation locked
Expected
1 sheet · 7/7 placed · 80.6186% finished-part yield · $74 stock
Why it exists
Tests multistart ordering, rotation rules, kerf, trim, cost, and a high-utilization single-sheet result.
B02Exact trim boundary
Input
One 46 × 46 in panel on a 48 × 48 in sheet
Constraint
1 in trim on every edge
Expected
Fits at 1,1 · 91.8403% finished-part yield
Why it exists
Proves that trim is unavailable margin and that a part larger than 46 in on either axis is rejected.
B03Terminal kerf exact fit
Input
Two 30 in rails from 60.125 in stock
Constraint
1/8 in kerf
Expected
1 board · 1 internal kerf · 0 in remainder
Why it exists
Checks that two finished pieces require one separating kerf, not two hidden kerfs.
B04Kerf boundary failure
Input
Three 32 in rails from 96 in stock
Constraint
1/8 in kerf
Expected
2 boards required
Why it exists
Shows why nominal lengths alone are insufficient: 96 in of parts plus two kerfs cannot fit in 96 in of stock.
Interpretation

A passing benchmark is a floor, not a ranking claim.

The sheet engine runs fixed sort and split strategies, retains the candidate using the fewest sheets after placing every possible part, and uses stable tie-breaking. The linear engine uses deterministic best-fit decreasing.

These cases verify known behavior; they do not prove global optimality for every large cutting-stock problem. Every displayed layout remains a planning estimate that must be checked against actual stock and a safe shop process.