[IMPALA-1441] Wrong results with outer joins in inline views

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: Impala 2.0
• Fix Version/s: Impala 2.1
• Labels: correctness impala
• Assignee: Dimitris Tsirogiannis
• Reporter: Dimitris Tsirogiannis
• Created: 31/Oct/14 19:23
• Updated: 20/Dec/15 00:05
• Resolved: 07/Nov/14 17:49

Description

A wrong plan is generated for the following query that contains an inline view with an outer join:

create table table_a (lac int, ci int);
create table table_b (lac int, ci int);

insert into table_a values (100, 1);
insert into table_a values (200, 2);
insert into table_a values (300, NULL);

insert into table_b values (200, 2);

Query: explain select t.* from
(
select a.lac a_lac, a.ci a_ci, b.lac b_lac, b.ci b_ci from
table_a a
left outer join
table_b b
on a.lac = b.lac and a.ci = b.ci
) t
+------------------------------------------------------------------------------------+
| Explain String                                                                     |
+------------------------------------------------------------------------------------+
| Estimated Per-Host Requirements: Memory=2.03GB VCores=2                            |
| WARNING: The following tables are missing relevant table and/or column statistics. |
| default.table_a, default.table_b                                                   |
|                                                                                    |
| 04:EXCHANGE [UNPARTITIONED]                                                        |
| |                                                                                  |
| 02:HASH JOIN [LEFT OUTER JOIN, BROADCAST]                                          |
| |  hash predicates: a.lac = b.lac, a.ci = b.ci                                     |
| |  other predicates: a.lac = b.lac, a.ci = b.ci                                    |
| |                                                                                  |
| |--03:EXCHANGE [BROADCAST]                                                         |
| |  |                                                                               |
| |  01:SCAN HDFS [default.table_b b]                                                |
| |     partitions=1/1 size=6B                                                       |
| |                                                                                  |
| 00:SCAN HDFS [default.table_a a]                                                   |
|    partitions=1/1 size=19B                                                         |
+------------------------------------------------------------------------------------+