Dayakar M created HIVE-27760:
--------------------------------

             Summary: Filter on date type partitioning column producing 0 
results
                 Key: HIVE-27760
                 URL: https://issues.apache.org/jira/browse/HIVE-27760
             Project: Hive
          Issue Type: Bug
          Components: HiveServer2
            Reporter: Dayakar M
            Assignee: Dayakar M


Filter on date type partitioning columns producing 0 results.

{*}Reproduction steps{*}:

1. test.q
{noformat}
CREATE EXTERNAL TABLE test(a string,b String) PARTITIONED BY(PartitionDate 
DATE) STORED AS ORC;
INSERT into test(PartitionDate, a,b) 
VALUES('2023-01-01','2023-01-01','2023-01-01');
INSERT into test(PartitionDate, a,b) 
VALUES('2023-01-02','2023-01-02','2023-01-02');

select count(*) from test where PartitionDate = '2023-01-01';{noformat}
2. Command to execute (pass different timezone than server)
{noformat}
mvn test -Dtest=TestMiniTezCliDriver -Dqfile=test.q 
-Dtest.output.overwrite=true -Duser.timezone=Asia/Hong_Kong{noformat}
 

*RootCause:* As a part of [HIVE-27373|http://example.com/] issue fix to parse 
the string to java.sql.Date object, java.text.SimpleDateFormat is replaced with 
java.time.format.DateTimeFormatter using java.time.LocalDate which represents a 
Date without TimeZone.  Here this input is passed [here 
|https://github.com/apache/hive/blob/master/standalone-metastore/metastore-server/src/main/java/org/apache/hadoop/hive/metastore/MetaStoreDirectSql.java#L1370]
 which uses SimpleDateFormat(parsing dates in a locale-sensitive manner) and 
java.sql.Date. Here user timezone is passed different so actual value is 
getting changed to a different value (for example 2023-01-01 is changed to 
2022-12-31) which is not matching with any partition so nothing gets returned.

*Solution:* In MetaStoreDirectSql.java, we should use 
java.time.format.DateTimeFormatter with java.time.LocalDate so that it will 
return proper date string.

 



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to