[
https://issues.apache.org/jira/browse/KAFKA-3006?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15064478#comment-15064478
]
Gwen Shapira commented on KAFKA-3006:
-------------------------------------
I'm not seeing how stream processing frameworks are impacted.
Stream processing frameworks ship with the Kafka Client as part of their
libraries / dependencies, their users don't interact with Kafka directly (i.e.
I depend on Spark libraries which bring whatever Kafka they need, and I only
use Spark APIs, not Kafka ones).
We are changing nothing on the broker side, so no matter which API they choose
to go with, it will work.
> Make collection default container type for sequences in the consumer API
> ------------------------------------------------------------------------
>
> Key: KAFKA-3006
> URL: https://issues.apache.org/jira/browse/KAFKA-3006
> Project: Kafka
> Issue Type: Improvement
> Components: clients
> Affects Versions: 0.9.0.0
> Reporter: Pierre-Yves Ritschard
> Labels: patch
>
> The KafkaConsumer API has some annoying inconsistencies in the usage of
> collection types. For example, subscribe() takes a list, but subscription()
> returns a set. Similarly for assign() and assignment(). We also have pause()
> , seekToBeginning(), seekToEnd(), and resume() which annoyingly use a
> variable argument array, which means you have to copy the result of
> assignment() to an array if you want to pause all assigned partitions. We can
> solve these issues by adding the following variants:
> {code}
> void subscribe(Collection<String> topics);
> void subscribe(Collection<String> topics, ConsumerRebalanceListener);
> void assign(Collection<TopicPartition> partitions);
> void pause(Collection<TopicPartition> partitions);
> void resume(Collection<TopicPartition> partitions);
> void seekToBeginning(Collection<TopicPartition>);
> void seekToEnd(Collection<TopicPartition>);
> {code}
> This issues supersedes KAFKA-2991
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
