[HIVE-5996] Query for sum of a long column of a table with only two rows produces wrong result

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Won't Fix
• Affects Version/s: 0.12.0
• Component/s: UDF
• Hadoop Flags: Incompatible change
• Assignee: Xuefu Zhang
• Reporter: Xuefu Zhang
• Created: 10/Dec/13 03:46
• Updated: 13/Dec/13 19:40
• Resolved: 13/Dec/13 15:37

Description

hive> desc test2;
OK
l                   	bigint              	None                
hive> select * from test2;                                 
OK
6666666666666666666
5555555555555555555
hive> select sum(l) from test2;
OK
-6224521851487329395

It's believed that a wrap-around error occurred. It's surprising that it happens only with two rows. Same query in MySQL returns:

mysql> select sum(l) from test;
+----------------------+
| sum(l)               |
+----------------------+
| 12222222222222222221 |
+----------------------+
1 row in set (0.00 sec)

Hive should accommodate large number of rows. Overflowing with only two rows is very unusable.

Issue Links

• relates to: HIVE-5878 Hive standard avg UDAF returns double as the return type for some exact input types
• relates to: HIVE-5356 Move arithmatic UDFs to generic UDF implementations