Siddharth Dave created ARROW-4016:
-------------------------------------
Summary: Wrong filtering using LIKE operator
Key: ARROW-4016
URL: https://issues.apache.org/jira/browse/ARROW-4016
Project: Apache Arrow
Issue Type: Bug
Components: Gandiva
Reporter: Siddharth Dave
[This issue was discovered during internal testing in Dremio in the java layer]
I noticed that the LIKE operator is not behaving as it should per SQL standard.
Notably it seems to consider \{{.}} as a wildcard in some occasions:
{code:sql}
SELECT A FROM (VALUES('abcde'), ('abc.e'), ('abcd'), ('abc.'), ('abc')) TBL(A)
WHERE A LIKE 'abc_%'
abcde
abc.e
abcd
abc.
abc
{code}
{code:sql}
SELECT A FROM (VALUES('abcde'), ('abc.e'), ('abcd'), ('abc.'), ('abc')) TBL(A)
WHERE A LIKE 'abc.%'
abcde
abc.e
abcd
abc.
abc
{code}
{code:sql}
SELECT A FROM (VALUES('abcde'), ('abc.e'), ('abcd'), ('abc.'), ('abc')) TBL(A)
WHERE A LIKE 'abc_'
(no result)
{code}
{code:sql}
SELECT A FROM (VALUES('abcde'), ('abc.e'), ('abcd'), ('abc.'), ('abc')) TBL(A)
WHERE A LIKE 'abc.'
(no result)
{code}
Based on code review it seemed that same issue may be present in Gandiva. This
bug is to verify & fix it in Gandiva as well.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
