[
https://issues.apache.org/jira/browse/STORM-631?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14301350#comment-14301350
]
ASF GitHub Bot commented on STORM-631:
--------------------------------------
Github user harshach commented on the pull request:
https://github.com/apache/storm/pull/406#issuecomment-72469942
@ogorun There is work in progress to refactor kafka code at STORM-631. Can
you please add your requirements/issue there so that we can fix it as part of
STORM-631. Thanks.
> Simplifying kafka connector code
> --------------------------------
>
> Key: STORM-631
> URL: https://issues.apache.org/jira/browse/STORM-631
> Project: Apache Storm
> Issue Type: Bug
> Components: storm-kafka
> Reporter: Parth Brahmbhatt
> Assignee: Parth Brahmbhatt
>
> The current kafka connector was written when kafka did not support any APIs
> to expose its internal structures. We already have STORM-590 to modify the
> kafka connector to new kafka meta APIS.
> The current codebase uses some marker interfaces to support reading kafka
> partition/broker info from zookeeper or for users to specify the mapping them
> self. Adding one more layer that reads this info using kakfa APIs would
> reduce readability. Because we are using marker interfaces there are bunch of
> places that has if statements with instanceof checks and each one creates a
> separate code path to trace. We should delete these interfaces and their
> implementation in favor of a single way to get kafka topic information, using
> kafka APIs. This will be a backward incompatible change but should make
> future changes and usage easy.
> Current kafka connector also has 2 separate configs for trident and core
> topologies. This configurations extend from a common interface but the
> extensions are not interchangeable. I think it is intuitive to have a single
> config with sane defaults for both core and trident spouts.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
