jean-claude created DRILL-4661:
----------------------------------
Summary: Optimze the LIKE function
Key: DRILL-4661
URL: https://issues.apache.org/jira/browse/DRILL-4661
Project: Apache Drill
Issue Type: Improvement
Reporter: jean-claude
Priority: Minor
The LIKE function is implemented using the java regex. This is different from
the STRPOS function which compares byte by byte.
When doing to testing I've noticed that for cases where the user wants to do a
restriction like WHERE COL1 LIKE '%xyz%' it would be faster to actually do
WHERE STRPOS(COL1, '%xyz%') > 0
I propose to change the LIKE operator's algorithm so that it detects if a byte
by byte compare could be applied and if so use the same technique as the STRPOS
function. Cases of the form '%xyz%', 'xyz%' and '%xyz' could be handled this
way.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
