aioinject 0.24.0

Async dependency injection library

Async-first dependency injection library based on python type hints

Framework integrations:

• FastAPI
• Litestar
• Strawberry-Graphql

Installation

Install using pip pip install aioinject

Example

import aioinject


class Database:
    def __init__(self) -> None:
        self._storage = {1: "Username"}

    def get(self, id: int) -> str | None:
        return self._storage.get(id)


class UserService:
    def __init__(self, database: Database) -> None:
        self._database = database

    def get(self, id: int) -> str:
        user = self._database.get(id)
        if user is None:
            raise ValueError
        return user


container = aioinject.Container()
container.register(aioinject.Singleton(Database))
container.register(aioinject.Singleton(UserService))


def main() -> None:
    with container.sync_context() as ctx:
        service = ctx.resolve(UserService)
        user = service.get(1)
        assert user == "Username"
        print(user)


if __name__ == "__main__":
    main()

Injecting dependencies into a function

You can inject dependencies into a function using @inject decorator, but that's usually only necessary if you're working with a framework:

import contextlib
from collections.abc import AsyncIterator
from contextlib import aclosing
from typing import Annotated

import uvicorn
from fastapi import FastAPI

from aioinject import Container, Inject, Singleton
from aioinject.ext.fastapi import AioInjectMiddleware, inject


class Service:
    pass


container = Container()
container.register(Singleton(Service))


@contextlib.asynccontextmanager
async def lifespan(_: FastAPI) -> AsyncIterator[None]:
    async with aclosing(container):
        yield


app = FastAPI(lifespan=lifespan)

app.add_middleware(AioInjectMiddleware, container=container)


@app.get("/")
@inject
async def root(service: Annotated[Service, Inject]) -> str:
    return str(service)


if __name__ == "__main__":
    uvicorn.run("main:app")

Using multiple dependencies with same type

If you have multiple implementations for the same dependency you can use typing.NewType to differentiate between them:

import dataclasses
from typing import Annotated, NewType

from aioinject import Container, providers, inject, Inject


@dataclasses.dataclass
class Client:
    name: str


GitHubClient = NewType("GitHubClient", Client)

GitLabClient = NewType("GitLabClient", Client)


def get_github_client() -> GitHubClient:
    return GitHubClient(Client(name="GitHub Client"))


def get_gitlab_client() -> GitLabClient:
    return GitLabClient(Client(name="GitLab Client"))


container = Container()
container.register(providers.Scoped(get_github_client))
container.register(providers.Scoped(get_gitlab_client))

with container.sync_context() as ctx:
    github_client = ctx.resolve(GitHubClient)
    gitlab_client = ctx.resolve(GitLabClient)

    print(github_client, gitlab_client)

Working with Resources

Often you need to initialize and close a resource (file, database connection, etc...), you can do that by using a contextlib.(async)contextmanager that would return your resource. Aioinject would automatically close them when context is closed, or when you call container.aclose() if your dependency is a Singleton.

import contextlib

from aioinject import Container, providers


class Session:
    pass


@contextlib.contextmanager
def get_session() -> Session:
    print("Startup")
    yield Session()
    print("Shutdown")


container = Container()
container.register(providers.Scoped(get_session))

with container.sync_context() as ctx:
    session = ctx.resolve(Session)  # Startup
    session = ctx.resolve(Session)  # Nothing is printed, Session is cached
# Shutdown

Async Dependencies

You can register async resolvers the same way as you do with other dependencies, all you need to change is to use Container.context instead of Container.sync_context:

import asyncio

from aioinject import Container, providers


class Service:
    pass


async def get_service() -> Service:
    await asyncio.sleep(1)
    return Service()


async def main() -> None:
    container = Container()
    container.register(providers.Scoped(get_service))

    async with container.context() as ctx:
        service = await ctx.resolve(Service)
        print(service)


if __name__ == "__main__":
    asyncio.run(main())

Providers

Scoped

Objects provided with Scoped provider would be cached within context.

Transient

Transient provider would provide different instances even if used within same context

Singleton

Singleton works as you expect - there would be only one instance of a singleton object, bound to a specific container

Object

Object provider just returns an object provided to it, e.g. Object(42)