Modern Python Development (3.12+)

Pure Python best practices for writing clean, idiomatic, and type-safe code. All guidance targets Python 3.12+ and covers only the standard language and stdlib — no third-party libraries or frameworks.

Naming Conventions

Prefix boolean variables and functions with is_, has_, can_, or should_.

Modern Syntax Essentials

Prefer modern constructs over legacy equivalents

Structural pattern matching

Use match/case for value dispatch, destructuring, and type narrowing — not as a simple switch replacement. See references/modern-syntax.md for detailed patterns.

match command:
    case {"action": "move", "direction": str(d)}:
        handle_move(d)
    case {"action": "quit"}:
        handle_quit()
    case _:
        handle_unknown(command)

The type statement (3.12+)

type Vector = list[float]
type Result[T] = T | Error
type Handler[**P] = Callable[P, Awaitable[None]]

Type Hints

Apply type hints to all function signatures, class attributes, and module-level variables. Omit return type only for __init__.

def calculate_total(items: list[float], *, tax_rate: float = 0.0) -> float:
    ...

Key rules:

• Use built-in generics: list[str], dict[str, int], tuple[int, ...], set[str]
• Use X | None instead of Optional[X]
• Use type statement for aliases, not bare assignment
• Prefer Protocol over ABC when only structural typing is needed
• Use @override decorator (3.12+) when overriding base class methods
• Use Self for methods returning the instance type

For comprehensive typing patterns including generics, Protocol, TypeGuard, TypeVar, and ParamSpec, consult references/type-hints.md.

Error Handling

Hierarchy

Define domain errors as a hierarchy inheriting from a project-level base:

class AppError(Exception):
    """Base for all application errors."""

class ValidationError(AppError):
    """Invalid input data."""

class NotFoundError(AppError):
    """Requested resource does not exist."""

Rules

• Catch the narrowest exception type possible. Never use bare except:.
• Raise from the original cause: raise NewError(...) from original.
• Use ExceptionGroup and except* for concurrent error aggregation.
• Use context managers (with statement) for resource cleanup — avoid manual try/finally.
• Document raised exceptions in the docstring, not with annotations.

Exception groups (3.11+)

try:
    async with TaskGroup() as tg:
        tg.create_task(validate_name(data))
        tg.create_task(validate_email(data))
except* ValidationError as eg:
    errors = eg.exceptions

Dataclasses and Data Modeling

Prefer dataclasses for data containers. Use __slots__ for memory-critical paths.

from dataclasses import dataclass, field

@dataclass(frozen=True, slots=True)
class Coordinate:
    x: float
    y: float
    label: str = ""
    tags: list[str] = field(default_factory=list)

• Use frozen=True for value objects and immutable data.
• Use slots=True (3.10+) to reduce memory and prevent attribute typos.
• Use kw_only=True when the class has more than 3 fields.
• Use field(default_factory=...) for mutable defaults.

Project Structure

Follow the src layout for any project intended to be packaged or tested:

project-name/
├── pyproject.toml
├── src/
│   └── package_name/
│       ├── __init__.py
│       ├── core/
│       ├── models/
│       └── utils/
├── tests/
│   ├── conftest.py
│   ├── unit/
│   └── integration/
└── scripts/

Key conventions:

• Single source of truth for version and metadata in pyproject.toml.
• The src/ layout prevents accidental imports from the working directory.
• Keep __init__.py files minimal — define the public API, not implementation.
• Organize by domain, not by technical role (prefer users/ over services/).

For complete project structure guidance including pyproject.toml conventions, module organization, and entry points, consult references/project-structure.md.

Idiomatic Patterns

Prefer stdlib over hand-rolling

• pathlib.Path for all file system operations.
• enum.Enum / enum.StrEnum for fixed sets of values.
• contextlib.contextmanager for simple resource management.
• functools.cache / functools.lru_cache for memoization.
• itertools for iterator composition.
• collections.abc for abstract container checks.

Generator expressions over list comprehensions when only iterating

total = sum(item.price for item in cart)  # generator — no intermediate list

Walrus operator for capture-and-test

if (match := pattern.search(line)) is not None:
    process(match.group(1))

__all__ for public API

__all__ = ["UserService", "create_user", "UserError"]

For comprehensive patterns including Protocols, ABCs, descriptors, context managers, enums, and more, consult references/patterns.md.

Additional Resources

Reference Files

For detailed guidance beyond this overview, consult:

• references/modern-syntax.md — Structural pattern matching, type statement, exception groups, @override, f-string details
• references/type-hints.md — Generics, Protocol, TypeGuard, TypeVar, ParamSpec, Self, Overload
• references/patterns.md — Idiomatic patterns: Protocols, enums, context managers, generators, pathlib, ABC
• references/project-structure.md — pyproject.toml conventions, src layout, module organization, entry points, __init__.py patterns