[
https://issues.apache.org/jira/browse/HIVE-29724?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18105862#comment-18105862
]
Zhihua Deng commented on HIVE-29724:
------------------------------------
No, HIVE-28972 tries to resolve upon the table schema evolution, every
partition of this table will use a different column id, as the example in the
jira:
alter table test_tbl add columns (col801 string) cascade;
if the test_tbl has 1000 partitions, then 1001 column ids will be created, even
just only one column id is shared among the table and its partitions before the
alter.
In a normal environment, for adding new partitions, e.g, dynamic partitions,
usually this partitions to be added have the same columns with the current
table. In this case, the new partition will reuse the same column id of the
table.
> Provide a Metastore tool for de-duplicating the columns
> -------------------------------------------------------
>
> Key: HIVE-29724
> URL: https://issues.apache.org/jira/browse/HIVE-29724
> Project: Hive
> Issue Type: Improvement
> Reporter: Zhihua Deng
> Priority: Major
>
> Following the discussion on HIVE-29694, we can provide a tool to de-duplicate
> the columns to avoid column metadata bloat, which could happen during
> replication.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
