MagicMock kills your IDE autocomplete. We fix that.
Type-safe mocking for pytest with full IDE support - autocomplete for method names, parameters, and mock assertions.
lib.mp4
When using pytest with unittest.mock, type inference becomes problematic:
from unittest.mock import MagicMock
from typing import cast
class UserService:
def get_user(self, user_id: int) -> dict:
return {"id": user_id, "name": "John"}
# Problem 1: MagicMock loses original type hints
mock_service = MagicMock(spec=UserService)
mock_service.get_user(1) # No autocomplete for get_user parameters
# Problem 2: cast() loses mock method hints
service = cast(UserService, mock_service)
service.get_user.assert_called_once() # assert_called_once has no type hint!typed-pytest provides type-safe mocking with:
- Catch typos at lint time - Misspelled method names are caught by mypy/pyright before running tests
- Original class method signatures - Full auto-completion for method names and parameters
- Type-checked mock assertions -
assert_called_once_with()and other assertions have full type hints - Type-checked mock properties -
return_value,side_effect,call_countare properly typed - IDE auto-completion for both original methods and mock methods
from typed_pytest_stubs import typed_mock, UserService
mock = typed_mock(UserService)
mock.get_usr # ❌ Caught by type checker: "get_usr" is not a known member
mock.get_user.assert_called_once_with(1) # ✅ Type-checked!from typed_pytest import TypedMock, typed_mock
mock_service: TypedMock[UserService] = typed_mock(UserService)
# Mock methods have type hints
mock_service.get_user.assert_called_once_with(1) # Type-checked!
mock_service.get_user.return_value = {"id": 1} # Type-checked!# Using uv
uv add typed-pytest
# Using pip
pip install typed-pytest- Python 3.10+
- pytest 8.0+
- pytest-mock 3.11+
from typed_pytest import typed_mock, TypedMock, TypedMocker
# Method 1: Direct mock creation
def test_user_service():
mock_service: TypedMock[UserService] = typed_mock(UserService)
mock_service.get_user.return_value = {"id": 1, "name": "Test"}
result = mock_service.get_user(1)
assert result == {"id": 1, "name": "Test"}
mock_service.get_user.assert_called_once_with(1)
# Method 2: Using pytest fixture
def test_with_fixture(typed_mocker: TypedMocker):
mock_service = typed_mocker.mock(UserService)
mock_service.get_user.return_value = {"id": 1}
mock_service.get_user(1)
mock_service.get_user.assert_called_once_with(1)typed-pytest-generator generates stub files for IDE auto-completion support. This allows your IDE to provide method signatures and type hints when using typed_mock().
The generator must run in your project's virtual environment because it imports your classes to inspect their method signatures. This means:
# ✅ Correct: Run in project environment (has access to your dependencies)
uv add typed-pytest --dev
uv run typed-pytest-generator
# ❌ Wrong: uvx runs in isolated environment (no access to your dependencies)
uvx typed-pytest-generator # Will fail if your classes import sqlalchemy, etc.Why? The generator uses Python's inspect module to extract method signatures, which requires actually importing your classes. If your classes depend on sqlalchemy, pydantic, or other packages, those must be installed in the environment where the generator runs.
Common errors:
No module named 'sqlalchemy'- Your class imports a package not in the isolated environmentcircular import- Your codebase has circular imports (fix withTYPE_CHECKINGblocks)
# Generate stubs for your classes
typed-pytest-generator -t myapp.services.UserService myapp.repos.ProductRepository
# Specify custom output directory
typed-pytest-generator -t myapp.services.UserService -o my_stubs
# Include private methods (starting with _)
typed-pytest-generator -t myapp.services.UserService --include-private
# Verbose output
typed-pytest-generator -t myapp.services.UserService -vThe generator creates the following structure:
typed_pytest_stubs/
├── __init__.py # Re-exports all stub classes
└── _runtime.py # Runtime class definitions with method signatures
# Import typed_mock from stubs package for full auto-completion
from typed_pytest_stubs import typed_mock, UserService
def test_user_service():
# typed_mock returns UserService_TypedMock with full IDE support
mock = typed_mock(UserService)
mock.get_user # ✅ Auto-complete for method names
mock.get_user.return_value # ✅ Auto-complete for mock properties
mock.get_user.assert_called_once_with(1) # ✅ Type-checked!
mock.get_user.return_value = {"id": 1, "name": "Test"}
result = mock.get_user(1)The generator supports two backends for extracting type information:
| Backend | Speed | Return Types | Use Case |
|---|---|---|---|
inspect (default) |
Fast (~10ms/class) | typing.Any |
Quick iteration during development |
stubgen |
Slower (~500ms/class) | Preserved (dict[str, Any], bool, etc.) |
Production, CI, better type hints |
Note: The
stubgenbackend requiresmypyto be installed:pip install mypy # or uv add mypy --dev
# Use default inspect backend (fast)
typed-pytest-generator -t myapp.services.UserService
# Use stubgen backend for better type information
typed-pytest-generator --backend stubgen -t myapp.services.UserService
# Short form
typed-pytest-generator -b stubgen -t myapp.services.UserServiceExample output difference:
With inspect backend:
class UserService_TypedMock:
@property
def get_user(self) -> MockedMethod[[int], typing.Any]: ...With stubgen backend:
class UserService_TypedMock:
@property
def get_user(self) -> MockedMethod[[int], dict[str, Any]]: ...You can configure typed-pytest-generator in your pyproject.toml file. This allows you to define targets, output directory, and other options without passing them via CLI every time.
[tool.typed-pytest-generator]
# List of fully qualified class names to generate stubs for
# Supports wildcard patterns:
# - "module.*" matches all classes in the module (non-recursive)
# - "module.**" matches all classes in the package recursively (includes submodules)
targets = [
"myapp.services.*", # All classes in myapp.services module
"myapp.repositories.**", # All classes in repositories and all submodules
"myapp.models.User", # Specific class
]
# Output directory for generated stubs (default: "typed_pytest_stubs")
output-dir = "typed_pytest_stubs"
# Include private methods starting with _ (default: false)
include-private = false
# Exclude specific classes from stub generation
exclude-targets = [
"myapp.internal.PrivateHelper",
"myapp.legacy.DeprecatedService",
]
# Backend for type extraction (default: "inspect")
# - "inspect": Fast, uses Python's inspect module
# - "stubgen": Slower, uses mypy's stubgen for accurate return types
backend = "inspect"Once configured, simply run:
# Uses configuration from pyproject.toml (auto-discovered)
typed-pytest-generator
# With verbose output
typed-pytest-generator -vCLI arguments take precedence over pyproject.toml configuration:
# Override targets from config
typed-pytest-generator -t myapp.services.NewService
# Override output directory
typed-pytest-generator -o custom_stubs
# Add exclusions (merged with config exclusions)
typed-pytest-generator -e myapp.services.SkipThis
# Use a specific config file
typed-pytest-generator -c /path/to/pyproject.toml
# Use wildcard to match all classes in a module (non-recursive)
typed-pytest-generator -t "myapp.services.*"
# Use recursive wildcard to match all classes in a package and submodules
typed-pytest-generator -t "myapp.repositories.**"
# Combine options
typed-pytest-generator -t "myapp.repositories.**" -e myapp.repositories.Internal -o stubs -vAdd typed_pytest_stubs/ to your .gitignore since these files are generated:
# Generated stub files
typed_pytest_stubs/For pyright/pylance users, add the stubs directory to your pyproject.toml:
[tool.pyright]
include = ["src", "tests", "typed_pytest_stubs"]Since typed_pytest_stubs/ is typically gitignored, some type checkers need extra configuration to find the generated stubs.
pyrefly (Meta's type checker)
Add the project root to search-path:
[tool.pyrefly]
project-includes = ["src", "tests"]
# Add project root so typed_pytest_stubs can be found
search-path = ["."]ty (Astral's type checker)
Add the project root to extra-paths:
[tool.ty.environment]
# Add project root so typed_pytest_stubs can be found
extra-paths = ["."]mypy and pyright
These work out of the box since they auto-discover packages in the project root.
The generated stubs should not be committed to Git. Instead, generate them:
-
Before running tests locally
typed-pytest-generator && pytest -
In your CI pipeline
# GitHub Actions example - name: Generate test stubs run: uv run typed-pytest-generator - name: Run tests run: uv run pytest
-
As a Makefile target (recommended)
test: uv run typed-pytest-generator uv run pytest
Why not pre-commit hook or Git?
- Generated code causes unnecessary Git conflicts
- Stubs should always reflect the current source code
- Keeps PRs clean and focused on actual changes
# Clone repository
git clone https://github.com/tmdgusya/typed-pytest.git
cd typed-pytest
# Install dependencies
uv sync --all-extras
# Run tests
uv run pytest
# Run tests with coverage
uv run pytest --cov=src/typed_pytest --cov-report=term-missing
# Type check
uv run mypy src/
uv run pyright src/
# Lint
uv run ruff check src/ tests/
uv run ruff format src/ tests/MIT