Synthetic corridor fixture

The synthetic corridor fixture is a public-safe, deterministic test case. It covers two corridors, five segments, mixed functional groups, one access gap, one monitoring gap, one structure, a plan, a Pareto archive, an operations timeline, a demo manifest, and a tiny known-optimum case.

Location: examples/synthetic-corridor/.

Generate

brushline fixture --out examples/synthetic-corridor
brushline fixture --out examples/synthetic-corridor --force

The generator is brushline.fixtures.synthetic.generate_fixture. It uses a fixed seed and a fixed creation time, so the committed fixture and a fresh build produce the same artifact digests.

Contents

Path

Meaning

dataset/manifest.json, dataset/segments.csv

The compiled dataset contract and the segment table.

corridor.json

The corridor graph: two corridors, five segments, two spans, two treatment units, two access points, one structure, and two service standards.

state.json

Five states, three treatment actions, one transition, and four observations.

plan.json

The 2026 plan with five actions and a CAD 5,300 total.

plans/plan-a.json to plans/plan-c.json

The three archive plans.

pareto.json

Three non-dominated cost/risk members and one balanced scalarization.

operations_timeline.jsonl

Five timestamped crew events.

demo.json

The public demo manifest.

tiny_optimum.json

The tiny known-optimum instance.

fixture.json

The fixture manifest: seed, expected counts, artifact digests, and content digest.

Expected values

Key

Value

Corridors

2

Segments

5

States

5

Treatment actions

3

Observations

4

Structures

1

Access gaps

1

Monitoring gaps

1

Plan actions

5

Pareto members

3

Timeline events

5

Known optimum

mow-2026

Known optimum penalty

0.46

Warnings

The fixture is valid, but it deliberately records two warnings:

  • segment s3 has no access point;

  • observation o2 covers only 40 percent of segment s2.

The unknown state on s3 and the absence of an observation for s3 are the monitoring gap.

Tiny known optimum

tiny_optimum.json defines a one-segment choice between no treatment, mowing in 2026, and mowing in 2027. The objective is

penalty = 0.4 * (cost / 1000) + 0.6 * risk

The options are:

Option

Cost

Risk

Penalty

none

0

0.8

0.48

mow-2026

1000

0.1

0.46

mow-2027

900

0.2

0.48

The optimum is mow-2026. The search validation uses this case to check that an exact solver and the metaheuristic adapter agree on a known answer.

Validation

brushline validate --dataset examples/synthetic-corridor/dataset

The test suite also validates the corridor, state, plan, archive, timeline, run, and demo contracts, and rebuilds the fixture to compare its digests with fixture.json.

Determinism

The fixture uses a fixed seed and a fixed 2026-11-14T00:00:00Z creation time. JSON is written with sorted keys and two-space indentation. The dataset table is CSV, not Parquet, so the table checksum does not depend on a Parquet library version. A change to the generator changes the content digest; regenerate the committed fixture and update the test expectation in the same change set.