Add support for Fast Stream Depends

[AndrianEquestrian]

Inspired by PR 854. A slightly cleaner solution, where FastApi separated from Fast Stream properly. Even though they were together at some point, now Fast Stream - is a separate library after all.

[coveralls]

coverage: 94.584% (-0.04%) from 94.62%
when pulling b414ef7 on AndrianEquestrian:better-fast-stream-depends-integration
into f2da51e on ets-labs:develop.

[cursor]

Bug: Falsy Dependency Extraction Fails

The _get_marker_from_parameter function fails to extract falsy but valid dependencies (e.g., False, 0, "", or a falsy _Marker) from Depends wrappers. The if _marker := marker_extractor(marker): line's implicit truthiness check prevents marker from being updated when the extracted dependency is falsy. This leaves the original Depends object, causing the subsequent isinstance(_Marker) check to fail, unlike the previous implementation.

src/dependency_injector/wiring.py#L616-L620

python-dependency-injector/src/dependency_injector/wiring.py

Lines 616 to 620 in b414ef7

	for marker_extractor in MARKER_EXTRACTORS:
	if _marker := marker_extractor(marker):
	marker = _marker
	break

Fix in Cursor

---

Was this report helpful? Give feedback by reacting with 👍 or 👎

[ZipFile]

Thanks for contribution!

[Totorokrut]

With raw Depends it doesn't work in FastStream.
Such test:

@broker.subscriber("hello_world") @inject async def on_hello_world( input: Input, service: Service=Depends(Provide["service"]), ): print(f"Service: {service}")

Crashes with an error:

pydantic_core._pydantic_core.ValidationError: 1 validation error for ResponseModel response Input should be a valid dictionary or instance of ResponseModel [type=model_type, input_value=<dependency_injector.wiri...bject at 0x78deeb6cdf70>, input_type=Provide] 

But with cast=False all works fine:

@broker.subscriber("hello_world") @inject async def on_hello_world( input: Input, service: Service=Depends(Provide["service"], cast=False), ): print(f"Service: {service}")

[AndrianEquestrian]

With raw Depends it doesn't work in FastStream. Such test:

@broker.subscriber("hello_world") @inject async def on_hello_world( input: Input, service: Service=Depends(Provide["service"]), ): print(f"Service: {service}")

Crashes with an error:

pydantic_core._pydantic_core.ValidationError: 1 validation error for ResponseModel response Input should be a valid dictionary or instance of ResponseModel [type=model_type, input_value=<dependency_injector.wiri...bject at 0x78deeb6cdf70>, input_type=Provide] 

But with cast=False all works fine:

@broker.subscriber("hello_world") @inject async def on_hello_world( input: Input, service: Service=Depends(Provide["service"], cast=False), ): print(f"Service: {service}")

Thank You for the observation! Please, include source where you import broker decorator from, I think I should investigate it, in order to determine the cause of the issue.

[Totorokrut]

There was a kafka broker in my test: broker = KafkaBroker("localhost:32833")

Actually, this issue can only be repeated with the FastDepends only by adding @fast_depends.inject:

@fast_depends.inject @inject async def apply_coefficient( a: int, b: str, coefficient_provider: CoefficientService = Depends( Provide[Container.service] ), ) -> float: return a * coefficient_provider.get_coefficient()

And now it crashes with the same error: pydantic_core._pydantic_core.ValidationError: 1 validation error for ResponseModel ...