Skip to content

VALIDATION.md

Latest commit

 

History

History
98 lines (73 loc) · 3.91 KB

VALIDATION.md

File metadata and controls

98 lines (73 loc) · 3.91 KB

Validation

Test Matrix

Suite Count Scope
Python unit/integration 1305 pytest tests/ across 80+ files
Rust unit/integration 203 cargo test --workspace across 5 crates
FFI parity 20 Python vs Rust parity: UPDE, Stuart-Landau, Petri net, RuleEngine
Notebook execution 10 nbclient runs all .ipynb cells under Python 3.12
Domainpack validation 31 Each domainpack exercises binding_spec.yaml → run.py end-to-end

CI runs tests on Python 3.10–3.13 and Rust on Linux, macOS, Windows.

CI Validation Gates

All gates must pass before ci-gate allows merge.

Gate Tool What it enforces
lint ruff check + ruff format + check_version_sync.py Style, formatting, version parity
typecheck mypy src/ Static type safety on public API
test pytest --cov (4 Python versions) Functional correctness + coverage XML
coverage-guard tools/coverage_guard.py Per-module thresholds from coverage_guard_thresholds.json
module-linkage tools/check_test_module_linkage.py Every source module has ≥1 test reference
security bandit -r src/ + pip-audit OWASP + supply-chain vulnerabilities
rust-check cargo fmt/clippy/doc/test (3 OS) Formatting, lints, docs, correctness
ffi-test maturin develop + pytest (3 OS × 2 Python) Rust↔Python binding correctness
cargo-audit cargo audit Rust dependency CVEs
rust-msrv Rust 1.75.0 build Minimum supported Rust version
benchmark bench/run_benchmarks.py → bench/compare_baseline.py ≤20% regression vs baseline

Phase Dynamics Correctness

Kuramoto coupling

The UPDE stepper implements dθ_n/dt = ω_n + Σ_m K_nm sin(θ_m − θ_n) with phase-difference coupling (not matrix-vector product of absolute sines). Tests verify:

  • Identical phases remain synchronised under zero coupling
  • Strong coupling drives R → 1 (Kuramoto synchronisation)
  • Zero coupling preserves natural frequency drift
  • Phase wrapping stays in [0, 2π)

Adaptive integration (RK45)

Dormand–Prince RK45 with embedded error control. Tests verify:

  • Phases stay bounded after 500 steps
  • Adaptive dt grows for smooth dynamics
  • Synchronisation tendency under strong coupling matches RK4
  • Python and Rust produce identical trajectories (atol < 1e-6)

Order parameter

R = |mean(exp(iθ))| computed identically in Python (numpy) and Rust. Parity test asserts |R_py − R_rust| < 1e-10.

Performance Benchmarks

bench/run_benchmarks.py measures μs/step across a 30-point grid:

  • N ∈ {8, 16, 64, 256, 1024} oscillators
  • Methods: euler, rk4, rk45
  • Backends: Python (numpy), Rust (spo_kernel)

Output includes system fingerprint (Python/numpy/scipy versions, platform) for reproducibility. bench/compare_baseline.py fails CI if any config regresses

20% vs bench/baseline.json.

Domainpack Integration (25 domains)

Each domainpack provides a binding_spec.yaml + run.py exercising the full pipeline: binding validation → oscillator instantiation → UPDE stepping → coherence monitoring → boundary enforcement.

Domains span: autonomous_vehicles, bio_stub, cardiac_rhythm, chemical_reactor, circadian_biology, epidemic_sir, firefly_swarm, fusion_equilibrium, geometry_walk, identity_coherence, laser_array, manufacturing_spc, metaphysics_demo, minimal_domain, network_security, neuroscience_eeg, plasma_control, pll_clock, power_grid, quantum_simulation, queuewaves, rotating_machinery, satellite_constellation, swarm_robotics, traffic_flow.

tests/test_domainpack_validation.py validates all specs parse and execute without error.

Regeneration

# Full local validation
pytest tests/ -v --tb=short
cargo test --workspace
cargo clippy --workspace -- -D warnings
python bench/run_benchmarks.py --json
python tools/coverage_guard.py
python tools/check_test_module_linkage.py