[teiid-issues] [JBoss JIRA] (TEIID-5900) query with local and remote pushes too much to remote

[Ken Geis (Jira)]

    [ https://issues.redhat.com/browse/TEIID-5900?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13970464#comment-13970464 ] 

Ken Geis commented on TEIID-5900:
---------------------------------

I have not overridden any settings.

I am using the [19.3.0.0 driver|https://search.maven.org/artifact/com.oracle.ojdbc/ojdbc8/19.3.0.0/jar].

You're right. It's creating 37 OR branches with 1000 values in each. I get this error:

{noformat}
2020-02-12 20:11:21.451  WARN 1 --- [ssorQueue766010] com.zaxxer.hikari.pool.ProxyConnection   : HikariPool-1 - Connection oracl
e.jdbc.driver.T4CConnection at 7f87b585 marked as broken because of SQLSTATE(08000), ErrorCode(17410)

java.sql.SQLRecoverableException: No more data to read from socket
{noformat}

I'd like to prevent it from pushing down these criteria. They are unnecessary because they will match all rows of the target table.

> query with local and remote pushes too much to remote
> -----------------------------------------------------
>
>                 Key: TEIID-5900
>                 URL: https://issues.redhat.com/browse/TEIID-5900
>             Project: Teiid
>          Issue Type: Bug
>          Components: Query Engine
>    Affects Versions: 12.3.1
>            Reporter: Ken Geis
>            Assignee: Steven Hawkins
>            Priority: Major
>
> I have a table "A" in an Oracle database. I have a date dimension "B" in Teiid, which is a materialized view of a recursive query. The date dimension has all dates from 1950 to 2050, 36890 rows in total.
> I run the following query:
> {code:sql}
> SELECT *
>   FROM a
>        JOIN b
>          ON a.date_column = b.date_key
> {code}
> Teiid sends a query to Oracle, which I will paraphrase as:
> {code:sql}
> SELECT <columns>
>   FROM a
>  WHERE date_column IN (to_date('1950-01-01 00:00:00', 'YYYY-MM-DD HH24:MI:SS'), 
>                        <36899 other date values!>)
> {code}
> Oracle breaks either with the length of the query or the number of options in the IN clause.
> I thought that MAKEDEP/MAKENOTDEP/MAKEIND could help, but they didn't seem to change anything.



--
This message was sent by Atlassian Jira
(v7.13.8#713008)