[
https://issues.apache.org/jira/browse/HIVE-28045?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
László Bodor updated HIVE-28045:
--------------------------------
Description:
{code}
create external table test_orc (a string, b string) partitioned by (pt string)
STORED AS ORC;
insert into test_orc PARTITION(pt='1111') values ("aaaa","bbbb");
alter table test_orc add columns (c string); -- no CASCADE
insert overwrite table test_orc PARTITION(pt='1111') values ("aaa", "bbb",
"ccc");
select * from test_orc;
{code}
returns
{code}
aaa bbb ccc 1111
{code}
which is unexpected without cascading the alter table, so the existing
partition is supposed to be untouched, like below:
{code}
aaa bbb NULL 1111
{code}
the same applies to Avro, if there is no external schema defined (which is
table level one)
repro:
https://github.com/abstractdog/hive/commit/e0de2c52de288cafbba3bcec3533bc78890e22f9#diff-1ca7746de2d6586dc2d9c1edbe5c18ca35c37fd39382c030142c782a5326faa5
was:
{code}
create external table test_orc (a string, b string) partitioned by (pt string)
STORED AS ORC;
insert into test_orc PARTITION(pt='1111') values ("aaaa","bbbb");
alter table test_orc add columns (c string); -- no CASCADE
insert overwrite table test_orc PARTITION(pt='1111') values ("aaa", "bbb",
"ccc");
select * from test_orc;
{code}
returns
{code}
aaa bbb ccc 1111
{code}
which is unexpected without cascading the alter table, so the existing
partition is supposed to be untouched, like below:
{code}
aaa bbb NULL 1111
{code}
the same applies to Avro, if there is no external schema defined (which is
table level one)
> Alter table without cascade seems to change metadata on existing partitions
> (Avro, ORC)
> ---------------------------------------------------------------------------------------
>
> Key: HIVE-28045
> URL: https://issues.apache.org/jira/browse/HIVE-28045
> Project: Hive
> Issue Type: Bug
> Reporter: László Bodor
> Assignee: László Bodor
> Priority: Major
>
> {code}
> create external table test_orc (a string, b string) partitioned by (pt
> string) STORED AS ORC;
> insert into test_orc PARTITION(pt='1111') values ("aaaa","bbbb");
> alter table test_orc add columns (c string); -- no CASCADE
> insert overwrite table test_orc PARTITION(pt='1111') values ("aaa", "bbb",
> "ccc");
> select * from test_orc;
> {code}
> returns
> {code}
> aaa bbb ccc 1111
> {code}
> which is unexpected without cascading the alter table, so the existing
> partition is supposed to be untouched, like below:
> {code}
> aaa bbb NULL 1111
> {code}
> the same applies to Avro, if there is no external schema defined (which is
> table level one)
> repro:
> https://github.com/abstractdog/hive/commit/e0de2c52de288cafbba3bcec3533bc78890e22f9#diff-1ca7746de2d6586dc2d9c1edbe5c18ca35c37fd39382c030142c782a5326faa5
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
