[
https://issues.apache.org/jira/browse/FLINK-6550?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16092801#comment-16092801
]
ASF GitHub Bot commented on FLINK-6550:
---------------------------------------
Github user StephanEwen commented on the issue:
https://github.com/apache/flink/pull/4312
I think this is a good case where the new pull request template would have
been useful.
This is in fact a change with potential performance impact.
Part of the result of the last micro-benchmarks by Piotr was that we
already introduced overhead and performance regressions along the collectors
and contexts. Additional branch and method indirection did
eventually add up to make the code less efficient and less JIT friendly.
I am very worried that these types of changes continue to be made without
any consideration of potential impact.
> Calling ctx.collect with a null OutputTag should log a warning or throw an
> exception
> ------------------------------------------------------------------------------------
>
> Key: FLINK-6550
> URL: https://issues.apache.org/jira/browse/FLINK-6550
> Project: Flink
> Issue Type: Improvement
> Components: DataStream API
> Reporter: Chesnay Schepler
> Assignee: Chesnay Schepler
> Priority: Minor
> Fix For: 1.4.0
>
>
> When using side outputs it is currently possible to call
> {Context#collect(OutputTag<X>, X record)} with null for the OutputTag.
> Effectively this causes the data to be swallowed. It is unlikely that a user
> actually intends this, so i propose to either log a warning or throw an
> exception.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
