[IMPALA-1353] Reuse of a column in JOIN predicate may lead to incorrect results

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: Impala 2.0
• Fix Version/s: Impala 2.0
• Labels: correctness
• Assignee: Alexander Behm
• Reporter: casey
• Created: 03/Oct/14 22:40
• Updated: 14/May/15 22:48
• Resolved: 07/Oct/14 18:22

Description

In the query below, the "WHERE t1.int_col = tt1.year" condition is dropped. This is very similar to other issues but Alex thinks fixing this case may be simpler.

Query: explain 
select straight_join 1
FROM alltypestiny t1
WHERE t1.int_col IN
  (SELECT tt1.year - tt1.year  AS int_col_1
   FROM alltypesagg tt1
   WHERE t1.int_col = tt1.year)
+----------------------------------------------------------+
| Explain String                                           |
+----------------------------------------------------------+
| Estimated Per-Host Requirements: Memory=80.01MB VCores=2 |
|                                                          |
| 05:EXCHANGE [UNPARTITIONED]                              |
| |                                                        |
| 02:HASH JOIN [LEFT SEMI JOIN, PARTITIONED]               |
| |  hash predicates: t1.int_col = tt1.year - tt1.year     |
| |                                                        |
| |--04:EXCHANGE [HASH(tt1.year - tt1.year)]               |
| |  |                                                     |
| |  01:SCAN HDFS [functional.alltypesagg tt1]             |
| |     partitions=11/11 size=814.73KB                     |
| |                                                        |
| 03:EXCHANGE [HASH(t1.int_col)]                           |
| |                                                        |
| 00:SCAN HDFS [functional.alltypestiny t1]                |
|    partitions=4/4 size=460B                              |
+----------------------------------------------------------+
Fetched 16 row(s) in 0.02s