Murshid Chalaev created HIVE-21950:
--------------------------------------
Summary: Optimizer complicates execution plan of queries with
SUBSTR function in EXISTS clause
Key: HIVE-21950
URL: https://issues.apache.org/jira/browse/HIVE-21950
Project: Hive
Issue Type: Bug
Affects Versions: 3.1.1, 2.3.0
Reporter: Murshid Chalaev
Queries with SUBSTR function in EXISTS clause have much more complicated
execution plan in Hive-2.3 then it was in Hive-1.2. A query below has 8 stages
which submit 4 MR jobs in Hive-2.3, while in Hive-1.2 it has 4 stages and
submits 1 MR job. Without SUBSTR function or with disabled CBO in Hive-2.3
execution plan is the same as in Hive-1.2 with enabled CBO.
*STEPS TO REPRODUCE:*
{code:java}
CREATE TABLE i1122 (id STRING);
INSERT INTO i1122 VALUES (1),(1001);
EXPLAIN
SELECT *
FROM i1122 AS t1
WHERE EXISTS (
SELECT 1
FROM i1122 AS t2
WHERE t2.id = substr(t1.id,4)
);{code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
