You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@beam.apache.org by "Robert Burke (Jira)" <ji...@apache.org> on 2021/10/08 22:01:00 UTC
[jira] [Commented] (BEAM-12999) Improve Reshuffle Transform
[ https://issues.apache.org/jira/browse/BEAM-12999?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17426397#comment-17426397 ]
Robert Burke commented on BEAM-12999:
-------------------------------------
How does one "reshuffle" using the user Key? That is, rather than using a random key as currenty implemented?
Much of the time these are implemented WRT composite implementations with a Group By Key under the hood, so trying to keep keys together defeats the purpose of a reshuffle AFAICT.
> Improve Reshuffle Transform
> ---------------------------
>
> Key: BEAM-12999
> URL: https://issues.apache.org/jira/browse/BEAM-12999
> Project: Beam
> Issue Type: Improvement
> Components: sdk-go, sdk-java-core, sdk-py-core
> Reporter: Ke Wu
> Assignee: Ke Wu
> Priority: P2
> Time Spent: 5h 50m
> Remaining Estimate: 0h
>
> See discussion [https://lists.apache.org/thread.html/r83adaad3a512ad186f2f9dc9dc4bec2a789070677c07cdcaad6fcfa5%40%3Cdev.beam.apache.org%3E]
>
> “beam:transform:reshuffle:v1" Transform represents different semantic transforms in different SDKs. The proposal is to replace "beam:transform:reshuffle:v1" with two new urns, one to represent reshuffle KV PCollection using the K, and the other to reshuffle based on random key.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)