[jira] [Resolved] (HIVE-16407) Compilation of a query with lots of "OR" conditions (like 2200) that check for "like" condition take a lot of time.

["Naveen Gangam (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/HIVE-16407?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Naveen Gangam resolved HIVE-16407.
----------------------------------
    Resolution: Fixed

This is not reproducible on {{master}} because it uses CBO. Rule-based optimizer is not invoked. 
This testcase performs very poorly in Hive 1.1. Since then HIVE-11141, HIVE-11671 and HIVE-11842 has been committed that seem to improve performance and provide stability. WIthout these fixes, the query compilation would not complete for over 90+ minus that lead to HS2 restart (query backlog pileup). With these fixes, the query takes about 4 minutes to compile with 2800 {{%like%}} conditions.
Closing this as a fix is no longer required.

> Compilation of a query with lots of "OR" conditions (like 2200) that check for "like" condition take a lot of time.
> -------------------------------------------------------------------------------------------------------------------
>
>                 Key: HIVE-16407
>                 URL: https://issues.apache.org/jira/browse/HIVE-16407
>             Project: Hive
>          Issue Type: Improvement
>          Components: HiveServer2
>            Reporter: Sudarshan
>            Assignee: Naveen Gangam
>         Attachments: issue_reproduce.sql
>
>
> Following is the table structure.
> ----
> create table test_hello (
> a string);
> ----
> A query that involves lots of "OR" conditions can in a "where" clause, can take a really long time to compile. It doesn't finishes compilation in even 30 to 40 minutes. 
> ++++
> select * from test_hello
> where
> (
> a like '%1%'  or 
> <Repeat above line 2000 times>
> a like '%1%' 
> )
> +++++
> (Attaching the entire script to the Jira)



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)