Producers and Consumers should accept Topics as arguments

[krisajenkins]

I find myself writing a lot of code that looks like this:

    weather_topic = app.topic("weather_data")
    producer.produce(topic=weather_topic.name, ...)

...and the ergonomics of that aren't great. I know there's a good reason to create Topic objects, but to a beginner it's going to feel like they're wrapping up strings just to unwrap them again.

It's a small thing, but I think it would feel nicer to program if we changed the produce and consume APIs to accept either strings or Topics:

# In Producer:
def produce(topic: Union[str, Topic], ...)

# In Consumer:
def subscribe(topics: List[Union[str, Topic]], ...)

WDTY?

[stereosky]

I agree with this suggestion. Introducing a Topic class concept and discarding it when you use it in the Producer/Consumer can be confusing for beginners. Supporting string or topic would make it work as expected

[daniil-quix]

Hi @shrutimantri , Thanks a lot for taking a stab at this issue. It is much appreciated!

We plan to take a slightly different direction with Producers, Consumers, and Topics.

The overall idea is to keep the interfaces of kafka.Producer and kafka.Consumer as close as possible to the underlying confluent_kafka classes. Instead, we want to introduce separate SerializingProducer and DeserializingConsumer classes that will accept Topic objects and use them also to serialize & deserialize data.

But to do that, we first need to refactor certain things in the library, so the path is not completely clear yet :)

I will close the #347 PR for now, but keep the issue open.

Cheers!