[teiid-issues] [JBoss JIRA] (TEIID-2249) Enable the use of temporary tables for those data sources that support them instead of IN criteria for EDS

[Steven Hawkins (JIRA)]

    [ https://issues.jboss.org/browse/TEIID-2249?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12752650#comment-12752650 ] 

Steven Hawkins commented on TEIID-2249:
---------------------------------------

There is current logic that unconditionally allows for the pushdown of dependent joins.  In essence the tuplebuffer(s) backing the independent side are made available at the translator level to be dealt with via customization - there is no built-in handling at the translator level.

To provide a solution around temp tables we have to consider:
* when should the dependent join pushdown occur (based upon the raw number of input keys, the potential number of source queries to spawn, etc.)
* what handling is needed at the translator level, assuming the conditional pushdown logic is sufficient then the only handling we would expect is then need to create a temporary table.  If the conditional logic is not sufficient then it seems like the translator level would need most of the engine strategy / logic such as splitting across predicates/queries.
* additional concerns not yet specified, such as what schema should be used, the direction of the join (if indexes would need to be added to the temporary table), etc.

                
> Enable the use of temporary tables for those data sources that support them instead of IN criteria for EDS
> ----------------------------------------------------------------------------------------------------------
>
>                 Key: TEIID-2249
>                 URL: https://issues.jboss.org/browse/TEIID-2249
>             Project: Teiid
>          Issue Type: Feature Request
>          Components: Query Engine
>            Reporter: Debbie Steigner
>            Assignee: Steven Hawkins
>             Fix For: 8.3
>
>
> Our proposal is to allow for the more efficient use of large ad-hoc result-sets by rather than creating a long 'IN' list, inserting them in to a temporary table - for example a # table in Sybase and SQL Server - and then generating an SQL join to that instead.
> One of the difference to materialized views (or at least my understanding), is that this work happens at a data-source rather than within the Teiid server.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira