Consumer batch mode

[hinxx]

Going through the rdkafka.h I noticed that rd_kafka_consume_batch() is considered legacy. Is there a similar non-legacy method in improving consumer performance?
On a related note, there seems to be quite a number of functions marked as legacy. How should I interpret the 'legacy' in such cases?

[edenhill]

The standard consumer_poll() API is typically fast enough, the bottleneck is typically fetching messages from Kafka, not returning the messages to the application.
Are you seeing significant performance improvements with consume_batch() over consumer_poll() (et.al)?

Deprecated means two things:

• we're no longer putting our efforts into these APIs.
• the API will probably be removed if we ever do a major version bump (nothing is planned).

[hinxx]

Thanks for clarifying!
I'm not that far down the line of being able to report any performance differences as I'm still in a process of coding up the (first iteration of) data consumption pipeline.
I was defaulting to rd_kafka_consumer_poll() until I noticed a whole set of functions that might be more performant (as per the comments in the header file).