Injection de dépendances courantes

In FastAPI, dependencies are powerful. They help you share logic, manage resources, and keep your code clean.

Common dependencies are functions or classes that you reuse across many API endpoints. This prevents repetitive code and makes your application easier to maintain and test.

Think of them as reusable building blocks for your API.

Many applications need global settings, like API keys or database URLs. By injecting configuration, you can easily access these settings in any endpoint without resorting to global variables.

This keeps your configuration flexible and easy to update.

Let's create a dependency that returns a simple settings dictionary. In a real application, this might load from environment variables using Pydantic's BaseSettings.

A common task for backend APIs is interacting with a database. Each request typically needs its own unique database "session" or connection.

Managing these sessions manually in every endpoint is tedious and error-prone. A dependency can handle creating and closing them automatically for you.

Here's a simplified example of a get_db dependency. In a real application, this would connect to a database using an ORM like SQLAlchemy.

The yield keyword is key here!

The yield keyword in our get_db function is very important.

• Code before yield runs before the endpoint function. This is for setup (e.g., opening a database connection).
• Code after yield runs after the endpoint function and its response. This is for teardown (e.g., closing the connection).

This pattern is called a "dependency with yield" or a "context manager dependency".

Most APIs need to know who is making a request. Authentication dependencies extract user information from request headers (like an API key or JWT token) and make it available to your endpoints.

This keeps your authentication logic separate and reusable across many routes.

Here's a basic dependency that simulates getting the current user. In a real application, this would involve validating a token from the request headers.

If the user isn't authenticated, it can raise an HTTPException to stop the request early.

Dependencies can also depend on other dependencies! This allows for complex logic to be broken down into smaller, manageable pieces.

For example, a "current admin user" dependency might first depend on "current user", then check if that user has admin privileges.

This modularity is a core strength of FastAPI's dependency injection.

Consider the following FastAPI dependency:

from fastapi import HTTPException, status

def check_api_key(api_key: str = "SECRET_KEY"):
    if api_key != "VALID_API_KEY":
        raise HTTPException(
            status_code=status.HTTP_403_FORBIDDEN,
            detail="Invalid API Key"
        )
    return True

If an endpoint uses Depends(check_api_key) and a request comes in without the correct API_KEY, what will happen?

We've explored how to inject common dependencies in FastAPI:

• Configuration: Share application settings easily across endpoints.
• Database Sessions: Manage database connections per request using yield for proper setup and teardown.
• User Authentication: Centralize logic for identifying and validating users.

Dependencies promote clean, reusable, and testable code by keeping cross-cutting concerns separate from your main endpoint logic.