[
https://issues.apache.org/jira/browse/HIVE-25948?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18005137#comment-18005137
]
Denys Kuzmenko commented on HIVE-25948:
---------------------------------------
no, it's not complete. atm we do not evaluate partition stats and auto mode
doesn't work
> Optimize Iceberg writes by directing records either Clustered- or Fanoutwriter
> ------------------------------------------------------------------------------
>
> Key: HIVE-25948
> URL: https://issues.apache.org/jira/browse/HIVE-25948
> Project: Hive
> Issue Type: Improvement
> Reporter: Ádám Szita
> Assignee: Denys Kuzmenko
> Priority: Major
> Labels: pull-request-available
>
> Currently Hive writes Iceberg tables with ClusteredWriter. This has less
> memory footprint as it only keeps one writer open at a time, but requires the
> records to be sorted.
> However if data cardinality is low Fanoutwriter is a better choice for
> performance.
> We should add support so that either can be used, and the decision could be
> based similarly how currently SortedDynPartitonOptimizer has it.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
