[HIVE-14349] Vectorization: LIKE should anchor the regexes

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.2.1, 2.2.0
• Fix Version/s: 1.3.0, 2.1.1, 2.2.0
• Component/s: UDF, Vectorization
• Release Note: Vectorization: LIKE should anchor the regexes
• Assignee: Gopal Vijayaraghavan
• Reporter: Gopal Vijayaraghavan
• Created: 26/Jul/16 23:53
• Updated: 08/Dec/16 14:15
• Resolved: 28/Jul/16 21:09

Description

RLIKE works like contains() and LIKE works like matches().

The UDFLike LIKE -> Regex conversion returns unanchored regexes making the vectorized LIKE behave like RLIKE.

create temporary table x (a string) stored as orc;
insert into x values('XYZa'), ('badXYZa');

select * from x where a LIKE 'XYZ%a%' order by 1;
OK
XYZa
badXYZa
Time taken: 4.029 seconds, Fetched: 2 row(s)

Issue Links

• is related to: HIVE-14318 Vectorization: LIKE should use matches() instead of find(0)
• requires: HIVE-13196 UDFLike: reduce Regex NFA sizes