[
https://issues.apache.org/jira/browse/HIVE-29176?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Thomas Rebele updated HIVE-29176:
---------------------------------
Summary: AntiJoin may not be introduced when the IS NULL filter is applied
to a nullable column (was: Wrong result for HiveAntiJoin when combining
columns from the LHS and RHS in the SELECT clause)
> AntiJoin may not be introduced when the IS NULL filter is applied to a
> nullable column
> --------------------------------------------------------------------------------------
>
> Key: HIVE-29176
> URL: https://issues.apache.org/jira/browse/HIVE-29176
> Project: Hive
> Issue Type: Bug
> Reporter: Thomas Rebele
> Assignee: Thomas Rebele
> Priority: Major
>
> Using tables as defined in `ql/src/test/queries/clientpositive/antijoin.q`,
> except changing the definition of t2_n33 to:
> {code:java}
> create table t2_n33 as select cast(2*key as int) key, value from t1_n55 union
> select 10 key, null value;
> {code}
> There's a difference when executing the query
> {code:java}
> select a.key, a.value, ''+a.value+b.value from t1_n55 a left join t2_n33 b on
> a.key=b.key where b.value is null;
> {code}
> With HiveJoin, the result contains a row {{{}10 val_10 NULL{}}}, with
> HiveAntiJoin the row is missing.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
