[IMPALA-5336] Inconsistent results when comparing string and timestamp fields

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: Impala 2.7.0
• Fix Version/s: Impala 2.10.0
• Component/s: Frontend
• Labels: correctness
• Environment: CDH 5.10.1
• Target Version: Impala 2.10.0
• Epic Color: ghx-label-9
• Assignee: Matthew Jacobs
• Reporter: Nicholas Brenwald
• Created: 18/May/17 13:33
• Updated: 31/Jul/17 21:53
• Resolved: 31/Jul/17 21:53

Description

create table t1 (c1 string, c2 string);
insert into t1 values ('a string', '2017-01-01');
create table t2 (c1 string) partitioned by (c2 string);
insert into t2 partition (c2) values ('a string', '2017-01-01');

This correctly returns 1 row

select * from t1 where c2 = cast('2017-01-01' as timestamp);

This incorrectly returns 0 rows

select * from t2 where c2 = cast('2017-01-01' as timestamp);

These 2 queries correctly return 1 row.

select * from t1 where c2 >= date_sub('2017-01-01', 1) and c2 <= '2017-01-02'; 
select * from t2 where c2 >= date_sub('2017-01-01', 1) and c2 <= '2017-01-02'; 

This queries behave differently. The 2nd query returns 0 rows:

select * from t1 where c2 between date_sub('2017-01-01', 1) and '2017-01-02'; 
select * from t2 where c2 between date_sub('2017-01-01', 1) and '2017-01-02'; 

This bug was not present in version:
Server version: impalad version 2.6.0-cdh5.8.2 RELEASE (build f25aa5b2bcdabf1eb4233747a7b04a067059ee3b)"

But is present in version:
Server version: impalad version 2.7.0-cdh5.10.1 RELEASE (build 876895d2a90346e69f2aea02d5528c2125ae7a32)