quality checks

HSEM Static Quality Checks

This document describes the static quality tools available in HSEM and how to run them locally.

Tools

Tool	Purpose	Config
Pyright	Type checking (CI-friendly Pylance equivalent)	pyrightconfig.json
Vulture	Dead-code and unused-symbol detection	vulture_whitelist.py
mypy	Legacy type checking	pyproject.toml [tool.mypy]
ruff	Linting and formatting	pyproject.toml [tool.ruff]
black	Code formatting	tox.ini [testenv:lint]
isort	Import sorting	tox.ini [testenv:lint]

Local Commands

Run all lint/format checks (isort + black + ruff)

tox -e lint

Run Pyright type checker

python -m pyright

This reads pyrightconfig.json and checks custom_components/hsem and tests/.

Run Vulture dead-code detector

python -m vulture custom_components/hsem tests vulture_whitelist.py --min-confidence 80

The vulture_whitelist.py file suppresses false positives for Home Assistant lifecycle methods (e.g. async_setup_entry, config flow steps) that Vulture would otherwise flag as unused because they are called dynamically by HA.

Run all quality checks (Pyright + Vulture)

tox -e quality

Run tests

pytest tests/

Pyright Configuration

pyrightconfig.json is set to typeCheckingMode: "basic" — a safe starting point for an HA integration that uses many dynamic patterns. Do not upgrade to strict mode without first resolving the known false-positive list.

Severity levels

Rule	Level	Reason
reportMissingTypeStubs	none	HA stubs are incomplete
reportUnknownMemberType	none	HA uses Any extensively
reportUnknownVariableType	none	HA uses Any extensively
reportUnknownArgumentType	none	HA uses Any extensively
reportTypedDictNotRequiredAccess	none	HA flow results use TypedDict with all-optional keys

Known remaining warnings (~156 total)

Most remaining warnings fall into two categories that are HA framework limitations, not bugs in HSEM:

1. CoordinatorEntity generic invariance (~38 warnings in custom_sensors/*.py): All HSEM sensors inherit from CoordinatorEntity[HSEMDataUpdateCoordinator] but HA's generic CoordinatorEntity[DataUpdateCoordinator[dict[str, Any]]] is invariant. These are safe at runtime; the correct fix is to wait for HA to widen the generic.

2. Test mock patterns (~100 warnings in tests/): Tests use partial mocks, MagicMock, and stub objects that don't have full type annotations. These are safe and intentional.

Vulture Whitelist

vulture_whitelist.py documents all HA dynamic entry points. Before deleting any function that Vulture flags, check whether it belongs to one of these categories:

• HA integration lifecycle: async_setup_entry, async_unload_entry, async_migrate_entry
• Config/options flow steps: async_step_user, async_step_init, etc.
• Diagnostics: async_get_config_entry_diagnostics
• Platform setup: async_setup_entry in sensor.py, select.py, switch.py, time.py
• Entity properties used by HA: device_info, native_value, is_on, etc.

If in doubt, add to the whitelist rather than deleting.

CI Integration

Pyright and Vulture run in CI as the quality job in .github/workflows/lint-and-test.yml. Both are currently set to continue-on-error: true (staged rollout) to avoid blocking PRs until the warning baseline is fully resolved.

Next steps to harden CI:

1. Resolve the remaining CoordinatorEntity invariance warnings (requires HA framework fix or a type-ignore comment on each super().__init__() call).
2. Set continue-on-error: false in the CI workflow once the warning count is zero.
3. Add tox -e quality to the local pre-commit checklist.