[IMPALA-817] Wrong expression may be used in aggregate query if there are multiple similar expressions

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: Impala 1.3
• Fix Version/s: Impala 1.3
• Assignee: Alexander Behm
• Reporter: casey
• Created: 18/Feb/14 21:28
• Updated: 19/Feb/14 19:38
• Resolved: 19/Feb/14 19:38

Description

I only saw this in the master branch, not in 1.2.4 or 1.3.0.

In the plan below you can see it's using the wrong expression. This seems to happen with expressions that use the same columns but only differ by an operator.

[localhost:21000] > explain select int_col + int_col, int_col * int_col FROM alltypesagg GROUP BY int_col + int_col, int_col * int_col having (int_col * int_col) < 0 limit 10;
Query: explain select int_col + int_col, int_col * int_col FROM alltypesagg GROUP BY int_col + int_col, int_col * int_col having (int_col * int_col) < 0 limit 10
+----------------------------------------------------------+
| Explain String                                           |
+----------------------------------------------------------+
| Estimated Per-Host Requirements: Memory=74.00MB VCores=2 |
|                                                          |
| 04:EXCHANGE [PARTITION=UNPARTITIONED]                    |
| |  limit: 10                                             |
| |                                                        |
| 03:AGGREGATE [MERGE FINALIZE]                            |
| |  group by: int_col + int_col                           |
| |  having: int_col + int_col < 0                         |
| |  limit: 10                                             |
| |                                                        |
| 02:EXCHANGE [PARTITION=HASH(int_col + int_col)]          |
| |                                                        |
| 01:AGGREGATE                                             |
| |  group by: int_col + int_col                           |
| |                                                        |
| 00:SCAN HDFS [functional.alltypesagg]                    |
|    partitions=10/10 size=743.67KB                        |
+----------------------------------------------------------+
Returned 17 row(s) in 0.01s