You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hive.apache.org by "Peter Vary (Jira)" <ji...@apache.org> on 2022/05/04 08:51:00 UTC
[jira] [Resolved] (HIVE-26183) Create delete writer for the UPDATE statemens
[ https://issues.apache.org/jira/browse/HIVE-26183?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Peter Vary resolved HIVE-26183.
-------------------------------
Fix Version/s: 4.0.0
Resolution: Fixed
Pushed to master.
Thanks for the review [~szita] and [~Marton Bod]!
> Create delete writer for the UPDATE statemens
> ---------------------------------------------
>
> Key: HIVE-26183
> URL: https://issues.apache.org/jira/browse/HIVE-26183
> Project: Hive
> Issue Type: Sub-task
> Reporter: Peter Vary
> Priority: Major
> Labels: pull-request-available
> Fix For: 4.0.0
>
> Time Spent: 6h
> Remaining Estimate: 0h
>
> During the investigation of the updates of partitioned table we had the following issue:
> - Iceberg inserts are needed to be sorted by the new partition keys
> - Iceberg deletes are needed to be sorted by the old partition keys and filenames
> This could contradict each other. OTOH Hive updates create a single query and writes out the insert/delete record for ever row. This would mean plenty of open writers.
> We might want to create something like a https://github.com/apache/iceberg/blob/master/core/src/main/java/org/apache/iceberg/io/SortedPosDeleteWriter.java, but we do not want to keep the whole rows in memory.
--
This message was sent by Atlassian Jira
(v8.20.7#820007)