pure-di 0.0.1

Simple yet powerful dependency injection framework!

Pure DI

Source Code: https://github.com/lionsoon/pure_di Features:

• Lightweight - no external dependencies (only typing_extensions)
• Based on type hints - no configuration required, less boilerplate
• Pytest integration - writing unit tests was never so easy!
• yeild powered - just yeild your component and then write closing steps
• Async support - sometimes we need await something in build process

Install

pip install pure-di

Example

Let's assume we have following services.py file, witch contains components we want to construct.
For example to be more "real-world" we will use httpx and sqlalchemy packages:

# services.py
import ...

@dataclass
class Settings:
    db_uri: str

@dataclass
class Database:
    engine: AsyncEngine

@dataclass
class Application:
    client: httpx.AsyncClient
    database: Database
    
    async def run(self): ...

The main concept of pure_di is Provider (and AsyncProvider if we need async).
Providers consist of factories - callables that build components. Factories are included to the provider using provider.include_factory method.
All factory arguments should be annotated. And return type should be specified either by return annotation, or by service_type argument (see example below).

Let's create an AsyncProvider for our services in provider.py file:

# provider.py
import httpx
from pure_di import AsyncProvider
from sqlalchemy.ext.asyncio import create_async_engine, AsyncEngine

from services import Settings, Database, Application

provider = AsyncProvider()


@provider.include_factory
def provide_settings() -> Settings:
    return Settings(db_uri="sqlite://...")


@provider.include_factory(service_type=AsyncEngine)
async def provide_engine(settings: Settings):
    engine = create_async_engine(settings.db_uri)
    yield engine
    await engine.dispose()


provider.include_factory(Database, service_type=Database)
provider.include_factory(Application, service_type=Application)

Now in main.py we can use provider.build_async method to construct Container. Container is similar to dict, where keys are services types, and values are services. It also contains container.partial_solve method that works similar to functools.partial and solves those arguments whose types it contains.

# main.py
from provider import provider
from services import Database, Application

async def do_some_work(database: Database, some_data):
    ...

async def main():
    async with provider.build_async() as container:
        partial_solved_do_some_work = container.partial_solve(do_some_work)
        await partial_solved_do_some_work(some_data="some_data")
        application = container[Application]
        await application.run()

Test

Let's write some tests using same example and pytest

from pure_di.pytest import pytest_provide_async

from provider import provider
from services import Database
from main import do_some_work


@Provider.from_factory(service_type=Database)
def mock_database():
    return ...


@pytest_provide_async(provider | mock_database)
@pytest.mark.asyncio
async def test_do_some_work(mocked_database: Database, any_usual_fixture):
    await do_some_work(mocked_database, some_data="some_data")
    assert mocked_database.something_to_assert()