[HIVE-12197] Parquet timestamps differ between Hive and Impala

• Type: Sub-task
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 1.1.1
• Component/s: Hive
• Assignee: Unassigned
• Reporter: Ryan Blue
• Created: 15/Oct/15 21:23
• Updated: 15/Oct/15 21:23

Description

Timestamps several hundred years in the past differ between Hive and Impala when stored in Parquet:

Query: create table ts_parquet (id bigint, ts timestamp) stored as parquet
Query: insert into ts_parquet values (1, '1400-01-01 00:01:01.101'), (2, '1581-11-14 07:08:01.101'), (3, '1582-02-11 07:08:01.101'), (4, '1999-01-02 14:15:34.101')
Query: select * from ts_parquet
+----+-------------------------------+
| id | ts                            |
+----+-------------------------------+
| 1  | 1400-01-01 00:01:01.101000000 |
| 2  | 1581-11-14 07:08:01.101000000 |
| 3  | 1582-02-11 07:08:01.101000000 |
| 4  | 1999-01-02 14:15:34.101000000 |
+----+-------------------------------+

hive> select * from ts_parquet;
1       1399-12-24 00:01:01.101
2       1581-11-04 07:08:01.101
3       1582-02-01 07:08:01.101
4       1999-01-02 14:15:34.101