[
https://issues.apache.org/jira/browse/HIVE-19387?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16539306#comment-16539306
]
Eugene Koifman commented on HIVE-19387:
---------------------------------------
truncate table/partition is handled by
{{HiveMetaStore.truncate_table(String,String,List)}}
It works by deleting all files under an X lock.
Since it's not writing anything it doesn't create a writeId and so RS Cache
invalidation doesn't know about this since it relies on ValidWriteIdList.
I suspect Incremental MV maintenance would have the same issue.
> Truncate table for Acid tables conflicts with ResultSet cache
> -------------------------------------------------------------
>
> Key: HIVE-19387
> URL: https://issues.apache.org/jira/browse/HIVE-19387
> Project: Hive
> Issue Type: New Feature
> Components: Transactions
> Reporter: Eugene Koifman
> Assignee: Eugene Koifman
> Priority: Major
>
> How should this work? Should it work like Insert Overwrite T select * from T
> where 1=2?
> This should create a new empty base_x/ and thus operate w/o violating
> Snapshot Isolation semantics.
> This makes sense for specific partition or unpartitioned table. What about
> "Truncate T" where T is partitioned? Is the expectation to wipe out all
> partition info or to make each partition empty?
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
