Test Coverage Implementation Plan — 2026-04-28 (v2)¶

Cycle: Test strategy execution — re-run after 20260428_test.md. Target window: 2 weeks (Phase A: 1 day · B–D: ~1 week · E–F: ongoing). Source of truth: Re-inventory by skill-test-coverage-audit on 2026-04-28. Related docs: - docs/cicd/testing-strategy.md - docs/status.md - docs/quickstart.md - tests/contract/test_pipeline_contracts.py - .github/instructions/tests.instructions.md

Current state (verified pytest --collect-only -q on 2026-04-28):

274 tests collected, 5 errors in 35.67s

docs/status.md and docs/quickstart.md claim "~200 tests" → stale.

New since v1: root cause of tests/test_api.py and tests/service/test_prediction_service.py collection errors identified — production import bug in src/app/config/mlflow.py (line 4): from data.params import load_params is missing the src. prefix → ModuleNotFoundError. This is not a test bug — it is a production-code bug masked by the test failure.

Gap matrix¶

Inventory: find src airflow/dags -name "*.py" | grep -v __pycache__ | wc -l → 94 files (including __init__.py). Non-init production modules: 70.

Legend: ✅ covered · ⚠ partial / broken collection · ❌ no tests · 🚧 placeholder · N/A __init__.py/config.

src/data/¶

src/data_quality/¶

src/features/¶

src/models/¶

src/pipelines/ (orchestrators)¶

Covered only by tests/contract/test_pipeline_contracts.py (asserts stage existence + I/O contracts in dvc.yaml). No invocation tests for any pipeline entrypoint. | File | Status | |---|---| | __main__.py, cli.py, source.py, preprocess.py, features.py, validate_raw.py, validate_features.py, validate_finished.py, validate_future.py, validation.py, classification.py, tune.py, final_train.py, register_model.py, inference.py, ablation.py | 🚧 contract-only |

src/app/routers/¶

src/app/services/ tasks/ data/ connections/ validation/ scraper/¶

src/app/schemas/ src/app/config/¶

src/ui/app/¶

airflow/dags/¶

src/utils/, src/make/¶

Aggregate¶

Collection errors (P0)¶

Phase A — Restore test signal (Day 1, P0)¶

Goal: green pytest tests/, honest test counts, runnable make test.

T1 — Fix production import bug in src/app/config/mlflow.py (~10 min)¶

• Replace from data.params import load_params → from src.data.params import load_params.
• Verification: python -c "from src.app.config.mlflow import mlflow_settings" succeeds.
• Side effect: unblocks tests/test_api.py and tests/service/test_prediction_service.py.

T2 — Resolve 3 missing-symbol collection errors (~1.5 h)¶

• For each of add_h2h_features, add_rest_days/rest_days_feature_meta, _select_best_run:
• git log --all -S '<symbol>' -- src/ to find when/why removed.
• Choose: (a) restore symbol, (b) rewrite test to current API, or (c) delete test + update docs/status.md.
• Verification: pytest --collect-only -q reports 0 errors.

T3 — Add [tool.pytest.ini_options] to pyproject.toml (~15 min)¶

[tool.pytest.ini_options]
pythonpath = ["src"]
testpaths = ["tests"]
markers = [
  "slow: tests that take >1s",
  "integration: requires live services",
  "load: locust load tests (excluded by default)",
]
addopts = "-q --strict-markers --tb=short"

T4 — Add make test* targets to Makefile (~15 min)¶

test:           pytest -m "not load and not integration"
test-fast:      pytest tests/unit tests/property -q
test-contract:  pytest tests/contract -q
test-coverage:  pytest --cov=src --cov-report=term-missing --cov-report=html

T5 — Update doc claims (~30 min)¶

• docs/status.md:61: replace "~200 tests" with current count.
• docs/quickstart.md:168: same.
• docs/cicd/testing-strategy.md: align "Suggested test coverage map" with this gap matrix.
• Verification: grep -n "~200" docs/ returns nothing.

Phase A DoD¶

• [ ] pytest --collect-only -q → 0 errors.
• [ ] pytest tests/ exits green (or marks failures as known with issue links).
• [ ] All make test* targets work.
• [ ] No stale ~200 claim remains.
• [ ] src/app/config/mlflow.py import is correct.

Phase B — Server-side P1 gaps¶

T6 — Expand tests/test_api.py to all routers¶

Per router: 1 happy + 1 negative test using TestClient with _get_feature_lookup/_get_predictor overrides. - routers/healthcheck.py, livescores.py, sources.py, stats.py, monitoring.py. - Verification: pytest tests/test_api.py -q.

T7 — CORS env-driven test¶

Parametrize CORS_ORIGINS env, restart app, assert Access-Control-Allow-Origin header.

T8 — tests/unit/test_api_client.py — mock httpx for src/ui/app/api_client.py¶

Cover: success, 4xx, 5xx, timeout.

T9 — Helm make helm-test target — helm lint && helm template for chart.¶

T10 — (Optional, @pytest.mark.integration) MinIO smoke for src/app/data/storage.py.¶

Phase B DoD¶

• [ ] Each src/app/routers/*.py has ≥ 1 happy + 1 negative test.
• [ ] CORS env behaviour tested.
• [ ] src/ui/app/api_client.py ≥ 80 % covered.
• [ ] make helm-test exists and passes.

Phase C — ML / pipeline P1¶

T11 — Extend tests/contract/test_pipeline_contracts.py to all DVC stages¶

Currently 10 stages in EXPECTED_STAGES; verify against dvc.yaml and add any missing (tune, final_train, validate_finished, validate_future, ablation).

T12 — tests/unit/test_tuning.py — Optuna 1-trial smoke for src/models/tuning.py (deterministic seed, mock MLflow).¶

T13 — tests/unit/test_register_model.py — src/pipelines/register_model.py with mlflow.MlflowClient mock; assert tags, alias, model URI.¶

T14 — (Optional CI) dvc repro -P on a tiny fixture dataset.¶

Phase C DoD¶

• [ ] Every stage in dvc.yaml covered by contract tests.
• [ ] Optuna smoke test exists.
• [ ] Model registration unit-tested with MLflow mock.

Phase D — UI + Airflow¶

T15 — Streamlit smoke via streamlit.testing.v1.AppTest¶

• One smoke per page in src/ui/app/pages/.
• Verification: pytest tests/unit/test_ui_smoke.py.

T16 — Airflow DagBag validation¶

• tests/unit/test_dag_validity.py: DagBag(include_examples=False).import_errors == {} and each DAG has default_args.retries >= 1.

Phase D DoD¶

• [ ] Each src/ui/app/pages/*.py smoke-tested.
• [ ] airflow/dags/** validated via DagBag, no import errors.

Phase E — P2/P3 expansion (ongoing)¶

• T17 — Scraper snapshot tests for src/app/scraper/driver.py (recorded HTML fixtures).
• T18 — Units for src/data/source.py, src/data/storage.py.
• T19 — Unit for src/features/select.py.
• T20 — Property tests for src/data/preprocess.py invariants.
• T21 — Coverage gating: --cov-fail-under=70 once Phase B–D land.
• T22 — Mutation testing pilot (mutmut run --paths-to-mutate src/features/elo.py).
• T23 — Create canonical docs/testing.md consolidating conventions (link, do not inline rules from tests.instructions.md).
• T24 — Tests for src/data_quality/{finished,future}.py.
• T25 — Tests for additional schemas (models.py, validate.py).

Phase F — CI integration¶

• T26 — .gitlab-ci.yml matrix: test:fast, test:contract, test:helm, test:dvc-smoke.
• T27 — Coverage badge + branch protection on main.

Doc drift (D-Tn)¶

Definition of Done — overall¶

• [ ] DoD-T1 pytest tests/ green, no collection errors.
• [ ] DoD-T2 Every src/app/routers/*.py has ≥ 1 happy + 1 negative test.
• [ ] DoD-T3 Every stage in dvc.yaml is in EXPECTED_STAGES.
• [ ] DoD-T4 Every src/ui/app/pages/*.py collects via AppTest without exception.
• [ ] DoD-T5 airflow/dags/** validated via DagBag with 0 import errors.
• [ ] DoD-T6 helm lint + helm template green in CI.
• [ ] DoD-T7 make test exists and is referenced in docs/quickstart.md.
• [ ] DoD-T8 docs/testing.md exists with current inventory + link to tests.instructions.md.
• [ ] DoD-T9 Coverage ≥ 70 % on src/ (gated in CI after Phase E).

2-day sprint slice¶

If only 2 days are budgeted: T1, T2, T3, T4, T5, T6, T11, T16. This restores signal, fixes the production import bug, gives all routers basic coverage, completes the contract-test surface, and validates DAGs.