Skip to content

TESTING.md

Latest commit

 

History

History
34 lines (26 loc) · 1.38 KB

TESTING.md

File metadata and controls

34 lines (26 loc) · 1.38 KB

Testing Strategy

Philosophy

  • TDD first: write a small failing test → implement → refactor.
  • Keep the solver APIs source-agnostic; use adapters in tests when you want to verify parsing.
  • Prefer small, deterministic tests; add smoke/integration tests for real datasets.

Types of tests

  • Unit (fast): pure functions, e.g. solve_lp on toy problems; solve_knapsack_01 with tiny arrays.
  • Adapter tests: parse external formats → check shapes, bounds, feasibility, and objective consistency.
  • Integration smoke: end-to-end for a specific instance (adapter → solve → checks).
  • Performance (later): quick sanity on large problems with time/iter thresholds.
  • Property-based (optional): random feasible instances to check invariants.

Current coverage

  • tests/utility/test_lp_utils.py: toy LPs + LPnetlib MAT smoke via load_lpnetlib_mat.
  • tests/utility/test_knapsack_utils.py: baseline DP cases.
  • tests/utility/test_io_knapsack_param.py: auto-discovery of Burkardt-style instances.

Commands

# run all
python -m unittest -v

# run a module
python -m unittest -v tests.utility.test_lp_utils

Reproducibility

  • Pin Python & SciPy in requirements.txt / environment.yml.
  • Keep adapters deterministic; avoid hidden randomness.
  • Store small datasets under tests/data/...; larger ones documented with links and checksums.