Chapter Outline

Chapter 13: Python Type Hints and Generics

Python is dynamically typed — variables don’t require explicit types. This flexibility is powerful, but as codebases grow, it can also become risky: errors may appear only at runtime, and function signatures may be ambiguous.

To improve safety and readability, Python introduced type hints (PEP 484). These don’t enforce types at runtime but can be checked using external tools like mypy, IDEs, or runtime validators like pydantic.

In this chapter, we’ll explore:

Basic and advanced type hints
The typing module and its constructs
Generics and reusable type-safe patterns
Advanced typing features (Protocol, TypedDict, Literal, NewType, etc.)
Practical examples in workflows and data validation

13.1 Function and Variable Annotations

Function Annotations

You can annotate parameters and the return type of a function.

python
def greet(name: str, age: int) -> str:
    return f"Hello, {name}. You are {age} years old."

name: str → name must be a str.
age: int → age must be an int.
-> str → the function returns a str.

If you call it incorrectly:

python
greet("Alice", "thirty")  # mypy or IDE will warn: expected int, got str

Python won’t raise an error at runtime, but type checkers like mypy will flag it.

mypy needs to be installed and run on your codebase to identify annotation errors. Visit the mypy site for additional instructions.

Default Values

python
def increment(x: int, step: int = 1) -> int:
    return x + step

Optional Parameters

Use Optional[type] (or Union[type, None]) when a parameter can be None.

python
from typing import Optional

def find_user(user_id: Optional[int]) -> str:
    if user_id is None:
        return "Guest"
    return f"User {user_id}"

u_id = None
print(find_user(u_id))

Variable Annotations

Variables can be annotated independently of assignment.

python
no_value: int                     # `no_value` assigned a type but not initialized
name: str = "Alice"               # `name` initialized to the value "Alice"
age: int = 30                     # `age` initialized to 30
scores: list[int] = [95, 88, 76]  # `scores` is a list of integers

Without Initialization

You can annotate without assigning a value. Useful for class attributes:

python
from typing import Optional

class User:
    id: int
    name: str
    email: Optional[str] = None   # This allows proper stringification

    def __str__(self) -> str:
      return f"User(id={self.id}, name={self.name}, email={self.email or 'N/A'})"

    def __repr__(self) -> str:
      return f"User(id={self.id!r}, name={self.name!r}, email={self.email!r})"

Notice the !r → it calls repr() on each attribute, so strings are quoted properly.

Forward References

For recursive data structures, use strings as type hints:

python
class Node:
    value: int
    next: "Node | None" = None  # quotes allow forward reference

    def __str__(self) -> str:
        return f"Node(value={self.value}, next={self.next})"

node_1 = Node()
node_1.value = 100

node_2 = Node()
node_2.value = 200
node_1.next = node_2
print(node_1)

Output:

bash
Node(value=100, next=Node(value=200, next=None))

Introspection with `annotations`

Python stores annotations in a special dictionary:

python
def add(x: int, y: int) -> int:
    return x + y

print(add.__annotations__)
# {'x': <class 'int'>, 'y': <class 'int'>, 'return': <class 'int'>}
print(Node.__annotations__)
# {'value': <class 'int'>, 'next': 'None | Node'}

Only things that are annotated are listed here. Un-annotated items are not listed within the dictionary.

13.2 Collections and the `typing` Module

Python’s collections give you fast, battle‑tested containers; the typing module (plus collections.abc) lets you describe those containers precisely so tools like mypy can catch bugs before runtime.

Core collections you’ll actually use

list: ordered, mutable sequence
tuple: fixed-size, immutable sequence
set, frozenset: unique items, fast membership tests
dict: key–value mapping
deque (collections.deque): fast appends/pops from both ends
defaultdict / Counter (collections): counting & grouped data

python
from collections import deque, defaultdict, Counter

dq = deque([1, 2, 3]); dq.appendleft(0)

freq = Counter("bananas")        # Counter({'a': 3, 'n': 2, 'b': 1, 's': 1})
group = defaultdict(list)
group["errors"].append("E001")

Type hinting collections (modern syntax)

In Python 3.9+, use built‑in generics: list[int], dict[str, float], etc. Prefer protocol/ABC types for function parameters (accept more inputs), and concrete types for returns (promise what you return).

python
# Good parameter types: accept any sequence/mapping; return concrete types.
from collections.abc import Sequence, Mapping

def average(nums: Sequence[float]) -> float:
    return sum(nums) / len(nums)

def invert(d: Mapping[str, int]) -> dict[int, str]:
    return {v: k for k, v in d.items()}

values: list[float] = [1.0, 2.0, 3.5]
print(average(values))  # 2.166...

Mutable vs. read‑only views

Use abstract bases from collections.abc to communicate mutability expectations:

Sequence[T] (read‑only indexing/iterating) vs MutableSequence[T]
Mapping[K, V] (read‑only) vs MutableMapping[K, V]

python
from collections.abc import MutableMapping

def touch(cache: MutableMapping[str, int], key: str) -> None:
    cache[key] = cache.get(key, 0) + 1

Nested / compound types

The following code defines type aliases using Python’s typing syntax.

python
Matrix = list[list[float]]               # 2D list
AdjList = dict[str, set[str]]            # graph adjacency list
Records = list[tuple[int, str | None]]   # union in tuples (3.10+ `|` syntax)

my_matrix: Matrix = [[1.2, 3.3, 9.5], [8.7, 4.5, 5.6]]
adjacency: AdjList = { 'one': { '1', '2', '3'}, 'two': { '1', '2', '3'} }
my_rec: Records = [(1, 'one'), (2, 'two'), (3, 'three')]

Deques, default dicts, and typing

Runtime types live in collections, but you annotate with built‑ins or typing aliases when needed:

python
from collections import deque, defaultdict

work: deque[str] = deque()
index: dict[str, list[int]] = defaultdict(list)

work.append("task-1")
index["x"].append(42)

defaultdict: A dictionary subclass that automatically provides a default value for missing keys.
Since "x" doesn’t exist yet, defaultdict creates index["x"] = [] automatically.

Typed dictionaries and lightweight records

When a dict has a fixed shape, use TypedDict. For tuple‑like records with names, use NamedTuple or dataclass.

python
from typing import TypedDict, NamedTuple

class UserTD(TypedDict):
    id: int
    name: str
    email: str | None

class Point(NamedTuple):
    x: float
    y: float

def make_user() -> UserTD:
    return {"id": 1, "name": "Alice", "email": None}

Iterators, generators, and streams

Use Iterable[T] when you only need to iterate, Iterator[T] when you also consume step‑by‑step, and Generator[Y, S, R] for full generator contracts.

python
from collections.abc import Iterable, Iterator

def only_ints(xs: Iterable[int]) -> list[int]:
    return [x for x in xs if isinstance(x, int)]

def countdown(n: int) -> Iterator[int]:
    while n:
        yield n
        n -= 1

for i in countdown(10):
    print(i)              # Prints 10 → 0

Validating with Pydantic

When “dicts of stuff” come from the outside (JSON, APIs), use Pydantic to validate and coerce.

python
from pydantic import BaseModel, EmailStr, ValidationError
from typing import Optional

class UserModel(BaseModel):
    id: int
    name: str
    email: Optional[EmailStr] = None
    tags: list[str] = []

try:
    u = UserModel(id="1", name="Alice", email="@example.com", tags=("a", "b"))
    # id coerced to int; tags coerced to list
except ValidationError as e:
    print(e.json())

The above code raises a validation error because of the email address format.

13.3 Callable and Type Aliases

When building larger Python applications, two recurring challenges come up:

Functions as parameters: You may want to pass a function into another function (e.g., a strategy or callback). How do you type hint such callables?
Repetitive, complex type hints: You may have a long, repeated type annotation like dict[str, list[tuple[int, float]]]. How do you avoid clutter and improve readability?

The solutions: Callable and Type Aliases.

Callable

The typing.Callable type lets you describe a function’s signature — the shape of its parameters and its return type.

Basic Example

python
from typing import Callable

def apply_twice(func: Callable[[int], int], x: int) -> int:
    """Apply a function to a value twice."""
    return func(func(x))

def square(n: int) -> int:
    return n * n

print(apply_twice(square, 3))  # (3^2)^2 = 81

Callable[[int], int] means:
- A function taking one int argument,
- Returning an int.

With Multiple Parameters

python
def operate(func: Callable[[int, int], float], a: int, b: int) -> float:
    return func(a, b)

def divide(a: int, b: int) -> float:
    return a / b

print(operate(divide, 10, 2))  # 5.0

Any Parameters (`...`)

python
def run_any(func: Callable[..., str]) -> str:
    """Callable with any number of arguments, but must return str."""
    return func("Hello", "World")

def join_strings(a: str, b: str) -> str:
    return a + " " + b

print(run_any(join_strings))  # Hello World

Using Callable in Classes (Strategy Pattern)

You can use Callable for strategies, hooks, and callbacks.

python
from typing import Callable

class Calculator:
    def __init__(self, operation: Callable[[int, int], int]):
        self.operation = operation

    def compute(self, a: int, b: int) -> int:
        return self.operation(a, b)


# Different strategies
add = lambda x, y: x + y
mul = lambda x, y: x * y

calc_add = Calculator(add)
calc_mul = Calculator(mul)

print(calc_add.compute(2, 3))  # 5
print(calc_mul.compute(2, 3))  # 6

Here Callable[[int, int], int] makes the class type-safe while still flexible.

Type Aliases

Sometimes type annotations become long and hard to read. A type alias gives them a readable name.

Simple Alias

python
UserId = int
Score = float

def record_score(user: UserId, score: Score) -> None:
    print(f"User {user} scored {score}")

record_score(101, 9.5)

This improves readability, even though UserId is just an int.

Complex Aliases

python
from typing import Callable

# Graph types
Node = str
AdjList = dict[Node, set[Node]]

# Function type
Transform = Callable[[str], str]

def apply_pipeline(data: list[str], steps: list[Transform]) -> list[str]:
    for step in steps:
        data = [step(x) for x in data]
    return data

# Usage
steps: list[Transform] = [str.strip, str.upper]
print(apply_pipeline(["  hello ", " world "], steps))
# ['HELLO', 'WORLD']

13.4 Generics

Generics in Python allow us to write flexible and type-safe code that works across many types without duplicating logic. They’re the foundation for reusable data structures (lists, queues, trees), frameworks, and APIs.

Why Generics?

Without generics, you either:

Hard-code a specific type:

python
def first_str(items: list[str]) -> str:
    return items[0]

→ Only works for list[str].

Use Any, which removes type safety:

python
from typing import Any

def first_any(items: list[Any]) -> Any:
    return items[0]

With Generics, you can write one function that works for all item types, while keeping type safety.

Generic Functions

python
from typing import TypeVar

T = TypeVar("T")  # placeholder for any type

def first(items: list[T]) -> T:
    """Return the first element of a list."""
    return items[0]

print(first([1, 2, 3]))        # infers T=int → returns int
print(first(["a", "b", "c"]))  # infers T=str → returns str

Here, T is a type variable that represents “the type of list elements.” mypy (or Pyright) checks that the return type matches the input type.

Generic Classes

python
from typing import Generic

class Box(Generic[T]):
    """A generic container that holds one item of any type."""

    def __init__(self, value: T):
        self.value = value

    def get(self) -> T:
        return self.value


int_box = Box(42)          # Box[int]
str_box = Box("hello")     # Box[str]

print(int_box.get())       # int
print(str_box.get())       # str

Box[int] → a box for integers.
Box[str] → a box for strings.

Frameworks like SQLAlchemy, Django ORM, and FastAPI use similar patterns internally.

Bounded Type Variables

You can restrict type variables to a base class.

python
from typing import Protocol

class Comparable(Protocol):
    def __lt__(self, other: object) -> bool: ...

U = TypeVar("U", bound=Comparable)

def minimum(a: U, b: U) -> U:
    return a if a < b else b

print(minimum(3, 7))       # works (int is Comparable)
print(minimum("a", "b"))   # works (str is Comparable)

This guarantees that minimum only accepts types that implement <.

Multiple TypeVars

python
K = TypeVar("K")
V = TypeVar("V")

def get_or_default(d: dict[K, V], key: K, default: V) -> V:
    return d.get(key, default)

scores: dict[str, int] = {"alice": 90}
print(get_or_default(scores, "bob", 0))  # returns int

Aliases with Generics

python
from typing import TypeVar

T = TypeVar("T")
Matrix = list[list[T]]

def transpose(matrix: Matrix[T]) -> Matrix[T]:
    return [list(row) for row in zip(*matrix)]

print(transpose([[1, 2, 3], [4, 5, 6]]))
# [[1, 4], [2, 5], [3, 6]]

Using Generics with `mypy`

Run type checks:

bash
mypy your_file.py

Example file:

python
nums = [1, 2, 3]
print(first(nums))       # OK
print(first("hello"))    # Error: str is not a list

Mypy enforces that only lists are passed.

Generics with Pydantic

Pydantic models also support generics, useful for API responses.

python
from typing import Generic, TypeVar, List
from pydantic import BaseModel

T = TypeVar("T")

class ApiResponse(BaseModel, Generic[T]):
    success: bool
    data: list[T]

class User(BaseModel):
    id: int
    name: str

# Use generics in practice
resp = ApiResponse[User](success=True, data=[User(id=1, name="Alice")])
print(resp.json())

Output:

json
{"success": true, "data": [{"id": 1, "name": "Alice"}]}

Frameworks like FastAPI leverage this to define generic API response types.

13.5 Advanced Typing Patterns

Protocols: Structural Subtyping

python
from typing import Protocol

class Drivable(Protocol):
    def drive(self) -> None: ...

class Car:
    def drive(self) -> None: print("Car driving")

def start_journey(vehicle: Drivable):
    vehicle.drive()

car = Car()
print(start_journey(car))

TypedDict: JSON-like schemas

It's a special construct that adds type hints to a dictionary.

python
from typing import TypedDict

class CarSpec(TypedDict):
    make: str
    model: str
    year: int
    electric: bool

my_car: CarSpec = { "make": "Ford", "model": "Mustang", "year": 1967, "electric": False }

Literal: Restricted values

python
from typing import Literal

def paint(color: Literal["red", "blue"]): ...

NewType: Strong semantic types

python
from typing import NewType

UserId = NewType('UserId', int)
Email = NewType('Email', str)

class User:
  id: UserId
  email: Email

13.6 Tools

So far we have talked about mypy and pydantic in this chapter:

Use mypy to type check your code statically.
Use pydantic to enforce data types at runtime, unlike type hints.

13.7 Assignment

In this assignment, you will apply type hints, generics, and runtime validation to a simple workflow.

Requirements

Define a generic Box[T] class that can hold any type (int, str, dict, etc.).
Create a function load_records(file: str) -> list[dict[str, str]] that loads a CSV file and returns records. Add proper type hints.
Define a type alias Record = dict[str, str].
Create a callable pipeline:
```
python
from typing import Callable
Transform = Callable[[Record], Record]
```
Write two transforms:
- strip_whitespace(record: Record) -> Record
- uppercase_name(record: Record) -> Record
Use mypy to check your code for type safety.

Use Pydantic to validate a record schema:

python
class UserRecord(BaseModel):
    id: int
    name: str
    email: EmailStr

Hints

Use list[dict[str, str]] for collections of records.
You can simulate CSV input with a list of dicts if you don’t want to read files.
mypy file.py will report any type mismatches.
Try invalid inputs (e.g., {"id": "abc"}) to see how Pydantic enforces runtime validation.

Check your understanding

Test your knowledge of Python Type Hints and Generics

Python