Python Backend Development Standards

This document defines coding standards and architectural patterns for Python backend development, with distinct approaches for brownfield and greenfield projects.

General

Code Structure and Style

• •Methods MUST be focused on a single responsibility.
• •Complex logic SHOULD be broken down into smaller, testable units.
• •Duplicated blocks of code MUST be removed by consolidating logic into shared methods.
• •Nested logic SHOULD be simplified, and deep if trees SHOULD be avoided to reduce cognitive complexity.
• •String literals SHOULD be replaced with constants or enums to prevent duplication.
• •Magic numbers MUST be replaced with named constants for better readability.
• •Exceptions MUST always be handled meaningfully; catch blocks MUST NOT be left empty.
• •Code MUST follow the PEP-8 standard, but lines can be up to 120 characters long
• •Code MUST be typed following the abilities of the current python version.

Dependency Management

• •Define versions with fixed version numbers
• •If the dependency manager in use supports lock files, they MUST be used to ensure predictable builds and to mitigate any security risks.

Testing

• •Unit tests MUST be written for all public methods.
• •Private methods MUST NOT be called directly in tests unless specifically requested.

Security

• •Parameterized queries MUST be used to prevent SQL injection.
• •Sensitive data MUST be stored securely, and passwords or tokens MUST NOT be logged.

General

• •Resources MUST be managed using context managers (with/async with) to ensure they are closed reliably.
• •Async Programming: You MUST use async/await for asynchronous operations and MUST NOT use blocking synchronous calls in async contexts
• •No Raw Dictionaries: You MUST NOT use raw Python dictionaries for structured data that requires validation. TypedDict can be used instead.

Brownfield Projects

Legacy Code Integration

• •Existing code SHOULD be updated to use modern syntax where applicable.

Testing Strategy

Pytest

• •Use pytest fixtures to avoid duplication of testing data
• •Use pytest.mark.parametrize to avoid repeating similar tests with slightly different test values.
• •Separate sections in the test between setup/execute/assert to delineate which parts of the test are for setup code, which is the code being tested in execute and what code is doing asserting.

Greenfield Projects

Architecture: Hexagonal Design Pattern

You MUST implement the Hexagonal Architecture (Ports and Adapters) pattern to maintain clear separation of concerns:

• •Application/Domain Layer: Contains pure business logic, domain models, entities, and business rules (located in application/domain/)
• •Application/Services Layer: Contains use cases, application services, and domain logic implementation (located in application/services/)
• •Ports Layer: Defines communication interfaces between the application core and external systems (located in application/ports/)
• •Adapters Layer: Contains framework-specific implementations of ports, including input adapters (controllers, CLI) and output adapters (databases, external APIs, message brokers) (located in adapters/)

Business domain logic MUST be completely isolated from vendor-specific and technology-specific code. Domain entities MUST NOT have dependencies on external frameworks or libraries.

Language Requirements

• •Minimum Python Version: 3.13+
• •Modern Language Features: You SHOULD use the latest available Python language features and best practices
• •Type Hints: All functions, methods, classes, and variables MUST have proper type annotations

Schema-Driven Development

• •Schema First: You MUST define OpenAPI schema before implementation
• •Validation: All data structures MUST be validated against their schemas

Data Validation and Models

• •Pydantic Models: You MUST use Pydantic models for input validation, data serialization, and API contracts
• •No Raw Dictionaries: You MUST NOT use raw Python dictionaries for structured data that requires validation
• •Strict Typing: You SHOULD enable Pydantic's strict mode for type validation
• •Field Validation: You SHOULD use Pydantic validators for complex business rule validation

Example:

from pydantic import BaseModel, Field, field_validator
from typing import Annotated
from datetime import datetime

class UserCreateRequest(BaseModel):
    name: Annotated[str, Field(min_length=1, max_length=100)]
    email: Annotated[str, Field(pattern=r'^[^@]+@[^@]+\.[^@]+$')]
    age: Annotated[int, Field(ge=0, le=150)]

    @field_validator('email')
    @classmethod
    def validate_email_domain(cls, v):
        # Custom business validation logic
        return v

Error Handling

• •HTTP Error Models: You MUST throw appropriate exceptions using the provided HTTP Error models
• •No Custom Response Logic: You MUST NOT implement custom error response serialization or HTTP response handling
• •Centralized Middleware: All errors SHALL be caught and handled by centralized error middleware defined in the provided core library
• •Appropriate Status Codes: You MUST use semantically correct HTTP status codes through the provided error models

Project Structure

Code MUST be organized following hexagonal architecture:

myservice/              # Root App Package
├── adapters/           # Framework specific implementation of ports
│   ├── input/          # App access ports, such as WebMVC Controllers
│   └── output/         # App external dependency ports, such as a WebClient
├── application/        # Portable business domain logic
│   ├── domain/         # Business Domain Models (Entities, DTOs, DAOs)
│   ├── ports/          # Communication Interfaces
│   │   ├── input/      # Input (domain access) Interfaces
│   │   └── output/     # Output (domain dependencies) Interfaces
│   └── services/       # Domain Logic Implementation