Skip to main content
Version: 0.0.1

Type Checking with MyPy

The Abstract Backend ships with a minimal types.py module and py.typed marker to show how to distribute type information (PEP 561) from the start. This page explains how to adapt that example for your project, keep your annotations healthy, and run MyPy with uv.

PEP Support Out of the Box​

The template configuration aligns with common typing standards:

  • PEP 484 – Type hints
  • PEP 585 – Generic type hints for built-ins
  • PEP 561 – Distributing type information via py.typed
  • PEP 695 – Modern type statement alias syntax (Python ≥3.12)

Where Types Live​

  • Default location: src/<your_package>/types.py
  • Marker file: src/<your_package>/py.typed
  • Public exports: declare symbols in __all__ so type checkers know what’s available.
src/your_package/types.py
type Percentage = float
type JSONDict = dict[str, "JSONValue"]
type JSONValue = str | int | float | bool | None | JSONDict | list["JSONValue"]

__all__ = ["Percentage", "JSONDict", "JSONValue"]

When you rename the package directory, update the paths in pyproject.toml, .github/workflows/type-check.yml, and any imports to match your chosen package name.

Running MyPy with uv​

The template keeps tooling simple:

# Install dependencies declared in pyproject.toml
uv sync

# Run mypy on the entire package
uv run mypy src/

# Target a submodule or run strict mode on demand
uv run mypy src/your_package/types.py
uv run mypy --strict src/your_package/

pyproject.toml contains default MyPy settings — adjust mypy.ini or pyproject.toml if you prefer a different configuration style.

Pre-commit Hook Integration​

A MyPy hook is already listed in .pre-commit-config.yaml. After running uv sync, enable the hooks:

pre-commit install
pre-commit run mypy --all-files

This mirrors what happens in CI and prevents untyped or mis-typed code from landing in the main branch.

Publishing Types (PEP 561)​

To ensure consumers receive your type information when installing from PyPI:

  1. Keep py.typed in the root of your package.
  2. Export symbols through __all__ in types.py (or via your package root __init__.py).
  3. If you restructure files, update .github/workflows/type-check.yml so the workflow monitors the correct paths.
  4. Uncomment and adjust the artifacts entries in pyproject.toml (wheel/sdist targets) if you rely on Hatch to copy additional files.

Customizing the Example​

  1. Rename the package – Replace your_package everywhere with your actual package name.
  2. Expand types.py – Add Protocols, TypedDicts, Literal types, etc., that match your domain.
  3. Document Types – Reference your new definitions in docstrings or docs so teammates know they exist.
  4. Add tests – Validate runtime helpers (type guards, validator helpers) just like any other logic.

Type Distribution Workflow​

The template includes .github/workflows/type-check.yml, which validates that types.py and py.typed end up inside distribution artifacts. After you rename your package, update the paths in that workflow (e.g., change src/your_package/types.py). See the PEP 561 Type Distribution Workflow for a full walkthrough.

Troubleshooting Checklist​

  • Imports fail in CI – Ensure the package name in MyPy command matches your src layout (e.g., uv run mypy src/).
  • py.typed missing – Confirm the file exists and, if you changed build tools, that it’s included in wheel/sdist artifacts.
  • Pre-commit slows down – Enable MyPy caching by adding cache_dir to configuration or leverage uv run mypy --dirty locally.
  • Third-party stubs missing – Install stub packages via uv add types-requests (they’re automatically tracked in uv.lock).

Helpful References​

Keep this file tailored to your real codebase once you flesh out the template. Treat the existing content as scaffolding and expand it as your project grows.