[HIVE-3824] bug if different serdes are used for different partitions

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 0.11.0
• Component/s: Query Processor
• Hadoop Flags: Reviewed
• Assignee: Namit Jain
• Reporter: Namit Jain
• Created: 20/Dec/12 09:12
• Updated: 16/May/13 21:10
• Resolved: 14/Jan/13 10:35

Description

Consider the following testcase:

create table tst5 (key string, value string) partitioned by (ds string) stored as rcfile;
insert overwrite table tst5 partition (ds='1') select * from src;
insert overwrite table tst5 partition (ds='2') select * from src;
insert overwrite table tst5 partition (ds='3') select * from src;

alter table tst5 stored as sequencefile;

insert overwrite table tst5 partition (ds='4') select * from src;
insert overwrite table tst5 partition (ds='5') select * from src;
insert overwrite table tst5 partition (ds='6') select * from src;

alter table tst5 set serde 'org.apache.hadoop.hive.serde2.lazy.LazySimpleSerDe';

insert overwrite table tst5 partition (ds='7') select * from src;
insert overwrite table tst5 partition (ds='8') select * from src;
insert overwrite table tst5 partition (ds='9') select * from src;

The following query works fine:

select key + key, value from tst5 where ((ds = '4') or (ds = '1'));

since both the partitions use ColumnarSerDe

But the following query fails:

select key + key, value from tst5 where ((ds = '4') or (ds = '1') or (ds='7'));

since different serdes are used.

Issue Links

• relates to: HIVE-3833 object inspectors should be initialized based on partition metadata