Synchronous Pull

[antonmry]

Based in some comments in this PR with @dpcollins-google, I opened this Feature request. Pub/sub provides an option to retrieve records synchronously: https://cloud.google.com/pubsub/docs/reference/rpc/google.pubsub.v1#pullrequest

It isn't the recommended way but it works well in some scenarios.

Pub/Sub Lite doesn't have this feature so if you need to do polling, some buffer is needed in the middle to simulate that behavior. That already happen in the PubSub Lite kafka library and the Apache Beam connector. It will happen again in the Source kafka connector.

To avoid to repeat the same workaround in each place, it would be nice to provide that API as part of the Java SDK. It would help also to clients moving from PubSub Lite to PubSub and viceversa.

[dpcollins-google]

This code is actually not duplicated, just not exposed to end users. Both of these classes use BufferingPullSubscriber here https://github.com/googleapis/java-pubsublite/blob/master/google-cloud-pubsublite/src/main/java/com/google/cloud/pubsublite/internal/BufferingPullSubscriber.java which implements this pattern. Both beam and kafka need explicit access to the partition numbers however, so a Cloud Pub/Sub Pull emulation layer wouldn't work for either of these.

That being said, we will consider this feature request for the future (exposing a public api that looks like Cloud Pub/Sub pull in the client) because it could be independently useful to other users.