[HIVE-5655] Hive incorrecly handles divide-by-zero case

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.10.0, 0.11.0, 0.12.0
• Fix Version/s: 0.13.0
• Component/s: Types
• Release Note: The behaviour needs to be documented.
• Assignee: Xuefu Zhang
• Reporter: Xuefu Zhang
• Created: 25/Oct/13 23:48
• Updated: 07/Nov/13 06:14
• Resolved: 07/Nov/13 06:14

Description

Unlike other databases, Hive currently has only one mode (default mode) regarding error handling, in which NULL value is returned. However, in case of divide-by-zero, Hive demonstrated a different behavior.

hive> select 5/0 from tmp2 limit 1;
Total MapReduce jobs = 1
...
Total MapReduce CPU Time Spent: 860 msec
OK
Infinity

The correct behaviour should be Hive returning NULL instead in order to be consistent w.r.t error handling. (BTW, the same situation is handled corrected for decimal type.)

MySQL has server modes control the behaviour. By default, NULL is returned. For instance,

mysql> select 3/0 from dual;
+------+
| 3/0  |
+------+
| NULL |
+------+
1 row in set (0.00 sec)

Issue Links

• is related to: HIVE-5438 Introduce different server modes for error handling in Hive server2
• relates to: HIVE-5660 Hive incorrecly handles data overflow/underflow
• relates to: HIVE-5656 Hive produces unclear, confusing SemanticException when dealing with mod or pmod by zero
• links to: https://reviews.apache.org/r/14973/