[
https://issues.apache.org/jira/browse/PHOENIX-5065?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17049388#comment-17049388
]
Richard Antal commented on PHOENIX-5065:
----------------------------------------
I Added some extra line to the InListExpression class and removed every null
from the IN list. Because in my understanding if you compare something and null
by the "=" operator it should return false, so we do not have to deal with that.
{code:java}
jdbc:phoenix:> explain SELECT COUNT(*) FROM SYSTEM.CATALOG WHERE TENANT_ID in
('', 'FOO');
+-----------------------------------------------------------------------+-----------------+----------------+--------------+
| PLAN |
EST_BYTES_READ | EST_ROWS_READ | EST_INFO_TS |
+-----------------------------------------------------------------------+-----------------+----------------+--------------+
| CLIENT 1-CHUNK PARALLEL 1-WAY RANGE SCAN OVER SYSTEM:CATALOG ['FOO'] | null
| null | null |
| SERVER FILTER BY FIRST KEY ONLY | null
| null | null |
| SERVER AGGREGATE INTO SINGLE ROW | null
| null | null |
+-----------------------------------------------------------------------+-----------------+----------------+--------------+{code}
> Inconsistent treatment of NULL and empty string
> -----------------------------------------------
>
> Key: PHOENIX-5065
> URL: https://issues.apache.org/jira/browse/PHOENIX-5065
> Project: Phoenix
> Issue Type: Bug
> Affects Versions: 4.14.1
> Reporter: Geoffrey Jacoby
> Priority: Major
> Attachments: PHOENIX-5065.master.v1.patch
>
>
> Phoenix doesn't handle NULLs consistently with other SQL dialects, and it
> doesn't handle them consistently internally either.
> In PHOENIX-2422, [~jamestaylor] mentioned that Phoenix's intended behavior is
> for empty string and NULL to be equivalent. That's inconsistent with other
> SQL dialects (in which NULL is never equal to anything, including itself),
> but if that's our documented behavior, then that's fine unless PHOENIX-2422
> to change it is ever worked.
> But consider the following queries:
> {code:java}
> SELECT COUNT(*) FROM SYSTEM.CATALOG WHERE TENANT_ID = '';
> -- Returns 0 rows
> SELECT COUNT(*) FROM SYSTEM.CATALOG WHERE TENANT_ID IS NULL;
> -- Returns some number of rows. Call it N
> SELECT COUNT(*) FROM SYSTEM.CATALOG WHERE TENANT_ID IN ('');
> -- Returns 0 rows
> SELECT COUNT(*) FROM SYSTEM.CATALOG WHERE TENANT_ID IN ('', 'FOO');
> -- Returns N rows. Note that FOO does not exist, and is just a nonsense string
> SELECT COUNT(*) FROM SYSTEM.CATALOG WHERE TENANT_ID = '' OR TENANT_ID = 'FOO'
> --Returns 0 rows, but slowly
> {code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
