Python Type Safety

Use guards and conditionals to narrow types within code blocks.

Quick Start

def get_user(user_id: str) -> User | None:
    """Return type makes 'might not exist' explicit."""
    ...

# Type checker enforces handling None case
user = get_user("123")
if user is None:
    raise UserNotFoundError("123")
print(user.name)  # Type checker knows user is User here

Fundamental Patterns

Pattern 1: Annotate All Public Signatures

Every public function, method, and class should have type annotations.

def get_user(user_id: str) -> User:
    """Retrieve user by ID."""
    ...

def process_batch(
    items: list[Item],
    max_workers: int = 4,
) -> BatchResult[ProcessedItem]:
    """Process items concurrently."""
    ...

class UserRepository:
    def __init__(self, db: Database) -> None:
        self._db = db

    async def find_by_id(self, user_id: str) -> User | None:
        """Return User if found, None otherwise."""
        ...

    async def find_by_email(self, email: str) -> User | None:
        ...

    async def save(self, user: User) -> User:
        """Save and return user with generated ID."""
        ...

Use mypy --strict or pyright in CI to catch type errors early. For existing projects, enable strict mode incrementally using per-module overrides.

Pattern 2: Use Modern Union Syntax

Python 3.10+ provides cleaner union syntax.

# Preferred (3.10+)
def find_user(user_id: str) -> User | None:
    ...

def parse_value(v: str) -> int | float | str:
    ...

# Older style (still valid, needed for 3.9)
from typing import Optional, Union

def find_user(user_id: str) -> Optional[User]:
    ...

Pattern 3: Type Narrowing with Guards

Use conditionals to narrow types for the type checker.

def process_user(user_id: str) -> UserData:
    user = find_user(user_id)

    if user is None:
        raise UserNotFoundError(f"User {user_id} not found")

    # Type checker knows user is User here, not User | None
    return UserData(
        name=user.name,
        email=user.email,
    )

def process_items(items: list[Item | None]) -> list[ProcessedItem]:
    # Filter and narrow types
    valid_items = [item for item in items if item is not None]
    # valid_items is now list[Item]
    return [process(item) for item in valid_items]

Pattern 4: Generic Classes

Create type-safe reusable containers.

from typing import TypeVar, Generic

T = TypeVar("T")
E = TypeVar("E", bound=Exception)

class Result(Generic[T, E]):
    """Represents either a success value or an error."""

    def __init__(
        self,
        value: T | None = None,
        error: E | None = None,
    ) -> None:
        if (value is None) == (error is None):
            raise ValueError("Exactly one of value or error must be set")
        self._value = value
        self._error = error

    @property
    def is_success(self) -> bool:
        return self._error is None

    @property
    def is_failure(self) -> bool:
        return self._error is not None

    def unwrap(self) -> T:
        """Get value or raise the error."""
        if self._error is not None:
            raise self._error
        return self._value  # type: ignore[return-value]

    def unwrap_or(self, default: T) -> T:
        """Get value or return default."""
        if self._error is not None:
            return default
        return self._value  # type: ignore[return-value]

# Usage preserves types
def parse_config(path: str) -> Result[Config, ConfigError]:
    try:
        return Result(value=Config.from_file(path))
    except ConfigError as e:
        return Result(error=e)

result = parse_config("config.yaml")
if result.is_success:
    config = result.unwrap()  # Type: Config

Advanced Patterns

Pattern 5: Generic Repository

Create type-safe data access patterns.

from typing import TypeVar, Generic
from abc import ABC, abstractmethod

T = TypeVar("T")
ID = TypeVar("ID")

class Repository(ABC, Generic[T, ID]):
    """Generic repository interface."""

    @abstractmethod
    async def get(self, id: ID) -> T | None:
        """Get entity by ID."""
        ...

    @abstractmethod
    async def save(self, entity: T) -> T:
        """Save and return entity."""
        ...

    @abstractmethod
    async def delete(self, id: ID) -> bool:
        """Delete entity, return True if existed."""
        ...

class UserRepository(Repository[User, str]):
    """Concrete repository for Users with string IDs."""

    async def get(self, id: str) -> User | None:
        row = await self._db.fetchrow(
            "SELECT * FROM users WHERE id = $1", id
        )
        return User(**row) if row else None

    async def save(self, entity: User) -> User:
        ...

    async def delete(self, id: str) -> bool:
        ...

Pattern 6: TypeVar with Bounds

Restrict generic parameters to specific types.

from typing import TypeVar
from pydantic import BaseModel

ModelT = TypeVar("ModelT", bound=BaseModel)

def validate_and_create(model_cls: type[ModelT], data: dict) -> ModelT:
    """Create a validated Pydantic model from dict."""
    return model_cls.model_validate(data)

# Works with any BaseModel subclass
class User(BaseModel):
    name: str
    email: str

user = validate_and_create(User, {"name": "Alice", "email": "[email protected]"})
# user is typed as User

# Type error: str is not a BaseModel subclass
result = validate_and_create(str, {"name": "Alice"})  # Error!

Pattern 7: Protocols for Structural Typing

Define interfaces without requiring inheritance.

from typing import Protocol, runtime_checkable

@runtime_checkable
class Serializable(Protocol):
    """Any class that can be serialized to/from dict."""

    def to_dict(self) -> dict:
        ...

    @classmethod
    def from_dict(cls, data: dict) -> "Serializable":
        ...

# User satisfies Serializable without inheriting from it
class User:
    def __init__(self, id: str, name: str) -> None:
        self.id = id
        self.name = name

    def to_dict(self) -> dict:
        return {"id": self.id, "name": self.name}

    @classmethod
    def from_dict(cls, data: dict) -> "User":
        return cls(id=data["id"], name=data["name"])

def serialize(obj: Serializable) -> str:
    """Works with any Serializable object."""
    return json.dumps(obj.to_dict())

# Works - User matches the protocol
serialize(User("1", "Alice"))

# Runtime checking with @runtime_checkable
isinstance(User("1", "Alice"), Serializable)  # True

Pattern 8: Common Protocol Patterns

Define reusable structural interfaces.

from typing import Protocol

class Closeable(Protocol):
    """Resource that can be closed."""
    def close(self) -> None: ...

class AsyncCloseable(Protocol):
    """Async resource that can be closed."""
    async def close(self) -> None: ...

class Readable(Protocol):
    """Object that can be read from."""
    def read(self, n: int = -1) -> bytes: ...

class HasId(Protocol):
    """Object with an ID property."""
    @property
    def id(self) -> str: ...

class Comparable(Protocol):
    """Object that supports comparison."""
    def __lt__(self, other: "Comparable") -> bool: ...
    def __le__(self, other: "Comparable") -> bool: ...

Pattern 9: Type Aliases

Create meaningful type names.

Note: The type Alias = ... statement syntax (PEP 695) was introduced in Python 3.12, not 3.10. For projects targeting earlier versions (including 3.10/3.11), use the TypeAlias annotation (PEP 613, available since Python 3.10).

# Python 3.12+ type statement (PEP 695)
type UserId = str
type UserDict = dict[str, Any]

# Python 3.12+ type statement with generics (PEP 695)
type Handler[T] = Callable[[Request], T]
type AsyncHandler[T] = Callable[[Request], Awaitable[T]]

# Python 3.10-3.11 style (needed for broader compatibility)
from typing import TypeAlias
from collections.abc import Callable, Awaitable

UserId: TypeAlias = str
Handler: TypeAlias = Callable[[Request], Response]

# Usage
def register_handler(path: str, handler: Handler[Response]) -> None:
    ...

Pattern 10: Callable Types

Type function parameters and callbacks.

from collections.abc import Callable, Awaitable

# Sync callback
ProgressCallback = Callable[[int, int], None]  # (current, total)

# Async callback
AsyncHandler = Callable[[Request], Awaitable[Response]]

# With named parameters (using Protocol)
class OnProgress(Protocol):
    def __call__(
        self,
        current: int,
        total: int,
        *,
        message: str = "",
    ) -> None: ...

def process_items(
    items: list[Item],
    on_progress: ProgressCallback | None = None,
) -> list[Result]:
    for i, item in enumerate(items):
        if on_progress:
            on_progress(i, len(items))
        ...

Configuration

Strict Mode Checklist

For mypy --strict compliance:

# pyproject.toml
[tool.mypy]
python_version = "3.12"
strict = true
warn_return_any = true
warn_unused_ignores = true
disallow_untyped_defs = true
disallow_incomplete_defs = true
no_implicit_optional = true

Incremental adoption goals:

• All function parameters annotated
• All return types annotated
• Class attributes annotated
• Minimize Any usage (acceptable for truly dynamic data)
• Generic collections use type parameters (list[str] not list)

For existing codebases, enable strict mode per-module using # mypy: strict or configure per-module overrides in pyproject.toml.

Best Practices Summary

1. Annotate all public APIs - Functions, methods, class attributes
2. Use T | None - Modern union syntax over Optional[T]
3. Run strict type checking - mypy --strict in CI
4. Use generics - Preserve type info in reusable code
5. Define protocols - Structural typing for interfaces
6. Narrow types - Use guards to help the type checker
7. Bound type vars - Restrict generics to meaningful types
8. Create type aliases - Meaningful names for complex types
9. Minimize Any - Use specific types or generics. Any is acceptable for truly dynamic data or when interfacing with untyped third-party code
10. Document with types - Types are enforceable documentation