confluent-kafka-python icon indicating copy to clipboard operation
confluent-kafka-python copied to clipboard

async SchemaRegistryClient

Open pavelschon opened this issue 1 year ago • 1 comments

Description

SchemaRegistryClient is based on requests, which is basically sync. Therefore it is not suitable for async applications.

I suggest to implement AsyncSchemaRegistryClient based on aiohttp.

pavelschon avatar Jul 03 '24 11:07 pavelschon

Actually we do not need this anymore.

We've implemented our own SchemaRegistryClient based on fastavro and httpx, which just calls REST endpoints.

It was pretty easy, I would recommend anyone doing the same, if you want absolute flexibility.

pavelschon avatar Sep 30 '24 10:09 pavelschon

Does this even matter? I mean the schemas are cached so it's one sync call.

dada-engineer avatar Apr 26 '25 05:04 dada-engineer

From today's perspective, it does not make any sense to ship confluent-kafka-python with schema registry client at all.

There are many http libraries (httpx, aiohttp, requests, ...) so that anybody can write own REST client according to their needs, using libraries they like.

pavelschon avatar Apr 27 '25 19:04 pavelschon

Fixed by https://github.com/confluentinc/confluent-kafka-python/pull/1965

rayokota avatar Jun 03 '25 00:06 rayokota