[
https://issues.apache.org/jira/browse/FLINK-9868?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16546079#comment-16546079
]
Kostas Kloudas commented on FLINK-9868:
---------------------------------------
Hi [~yanghua]! From the discussion you posted, I do not see any need for
exposing the channel id.
The only thing I get is that the users want to key by an id (they name it
channel id but it is their own channel concept, not Flink), they want to have
the index as "timestamp" and they want to process the incoming events per key
on an ascending timestamp order.
If my understanding is correct, then we should close this issue.
Also, in a higher level, exposing the channelId to the user seems to break the
"separation of concerns" between user APIs and system implementation.
> Expose channel id to ProcessFunction
> ------------------------------------
>
> Key: FLINK-9868
> URL: https://issues.apache.org/jira/browse/FLINK-9868
> Project: Flink
> Issue Type: New Feature
> Components: Local Runtime
> Reporter: Hequn Cheng
> Assignee: vinoyang
> Priority: Major
>
> Currently, channel id has not been exposed from {{StreamInputProcessor}} to
> the {{ProcessOperator}} and {{ProcessFunction}}. There are some cases that
> users want the channel id, as discovered [here(mailing
> list)|http://apache-flink-user-mailing-list-archive.2336050.n4.nabble.com/Parallelism-and-keyed-streams-td21501.html].
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
