Melhores práticas de Python 2026: qualidade do código, estrutura do projeto e conjunto de ferramentas

# Modern Python toolchain (2026)
pip install ruff  # fastest linter + formatter (replaces black + flake8 + isort)

# Format all Python files
ruff format .

# Lint and auto-fix
ruff check . --fix

# pyproject.toml configuration
# [tool.ruff]
# line-length = 100
# select = ["E", "F", "I", "N", "UP"]  # rules to enable
# ignore = ["E501"]

# GOOD: meaningful names, type hints, docstrings
def calculate_compound_interest(
    principal: float,
    annual_rate: float,
    years: int,
    compounds_per_year: int = 12,
) -> float:
    # FV = P * (1 + r/n)^(n*t)
    rate_per_period = annual_rate / compounds_per_year
    periods = compounds_per_year * years
    return principal * (1 + rate_per_period) ** periods

# BAD: cryptic names, no types, magic numbers
def calc(p, r, y, n=12):
    return p * (1 + r/n) ** (n*y)

myapp/
  pyproject.toml          # project metadata + tool config
  README.md
  src/
    myapp/
      __init__.py
      main.py             # entry point
      api/                # FastAPI routes
        __init__.py
        users.py
        posts.py
      models/             # SQLAlchemy/Pydantic models
        user.py
        post.py
      services/           # business logic
        user_service.py
      repositories/       # database access
        user_repo.py
      core/               # config, dependencies
        config.py
        database.py
  tests/
    unit/
      test_user_service.py
    integration/
      test_api.py
    conftest.py            # shared fixtures

# core/config.py — Pydantic Settings
from pydantic_settings import BaseSettings, SettingsConfigDict

class Settings(BaseSettings):
    model_config = SettingsConfigDict(
        env_file=".env",
        env_file_encoding="utf-8",
        case_sensitive=False,
    )

    # Database
    database_url: str
    database_pool_size: int = 10

    # API
    api_host: str = "0.0.0.0"
    api_port: int = 8000
    debug: bool = False

    # Security
    secret_key: str
    jwt_algorithm: str = "HS256"
    access_token_expire_minutes: int = 15

    # External services
    redis_url: str = "redis://localhost:6379/0"

# Singleton via functools.cache
from functools import lru_cache

@lru_cache(maxsize=1)
def get_settings() -> Settings:
    return Settings()

# Usage
settings = get_settings()
print(settings.database_url)

# FastAPI dependency
from fastapi import Depends

def get_db_url(settings: Settings = Depends(get_settings)) -> str:
    return settings.database_url

# Domain exceptions
class AppError(Exception):
    def __init__(self, message: str, code: str):
        super().__init__(message)
        self.code = code

class NotFoundError(AppError):
    def __init__(self, resource: str, id: int | str):
        super().__init__(f"{resource} {id} not found", "NOT_FOUND")

class ValidationError(AppError):
    def __init__(self, field: str, message: str):
        super().__init__(f"{field}: {message}", "VALIDATION_ERROR")

# Never swallow exceptions silently
# BAD
try:
    result = process(data)
except Exception:
    pass  # NEVER DO THIS

# GOOD: log and re-raise or handle specifically
import logging
logger = logging.getLogger(__name__)

try:
    result = process(data)
except ValueError as e:
    logger.warning(f"Invalid data: {e}")
    raise ValidationError("data", str(e)) from e
except Exception as e:
    logger.exception(f"Unexpected error processing data: {e}")
    raise

# 1. Use generators for large data
def process_large_file(path: str):
    with open(path) as f:
        for line in f:  # not f.readlines()!
            yield process_line(line)

# 2. Use slots for memory-efficient classes
class Point:
    __slots__ = ('x', 'y')  # saves 40-50% memory vs dict-based attrs
    def __init__(self, x: float, y: float):
        self.x = x
        self.y = y

# 3. Profile before optimizing
import cProfile
cProfile.run('your_slow_function()', sort='cumulative')

# 4. Use collections for appropriate data structures
from collections import deque, defaultdict, Counter

# deque for queue (O(1) append/pop from either end)
queue = deque()
queue.appendleft(item)  # O(1) vs list.insert(0, item) which is O(n)

# Counter for counting
word_count = Counter(text.split())
top_10 = word_count.most_common(10)

# 5. Prefer comprehensions over loops for transformations
squares = [x**2 for x in range(1000)]  # faster than for loop + append

# 6. Use functools.cache for expensive pure functions
from functools import cache

@cache
def fibonacci(n: int) -> int:
    if n < 2: return n
    return fibonacci(n-1) + fibonacci(n-2)

# 2026 standard: use uv (blazing fast, from Astral/ruff team)
pip install uv

# Create virtual environment and install deps (10x faster than pip)
uv venv && source .venv/bin/activate
uv pip install -r requirements.txt

# Or manage with pyproject.toml
uv add fastapi sqlalchemy pydantic  # adds to pyproject.toml
uv add --dev pytest ruff mypy       # dev dependencies

# Generate lockfile
uv lock

# Install from lockfile (reproducible)
uv sync