[HIVE-22945] Hive ACID Data Corruption: Update command mess the other column data and produces incorrect result

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: 3.2.0
• Component/s: Hive, Transactions
• Assignee: Denys Kuzmenko
• Reporter: Rajkumar Singh
• Created: 28/Feb/20 20:08
• Updated: 11/Mar/20 11:47
• Resolved: 11/Mar/20 11:47

Description

Hive Update Operation update the other column incorrectly and produces incorrect results:

Steps to reproduce:

CREATE TABLE `test`(
  `start_dt` timestamp, 
  `stop_dt` timestamp
  );
  
INSERT INTO test (start_dt, stop_dt) SELECT  CURRENT_TIMESTAMP, CAST(NULL AS TIMESTAMP);

select * from test; 
+--------------------------+---------------+
|      test.start_dt       | test.stop_dt  |
+--------------------------+---------------+
| 2020-02-28 20:06:29.116  | NULL          |
+--------------------------+---------------+

UPDATE test SET STOP_DT = CURRENT_TIMESTAMP WHERE CAST(START_DT AS DATE) = CURRENT_DATE;

+------------------------+--------------------------+
|     test.start_dt      |       test.stop_dt       |
+------------------------+--------------------------+
| 2020-02-28 00:00:00.0  | 2020-02-28 20:07:12.248  |
+------------------------+--------------------------+

Issue Links

• is a child of: HIVE-22999 typeCast issue while constant propagation when cbo is disabled