You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@phoenix.apache.org by "Samarth Jain (JIRA)" <ji...@apache.org> on 2017/01/13 21:49:26 UTC

[jira] [Updated] (PHOENIX-3599) Take advantage of EncodedQualifiersList when there is no order by/group by

     [ https://issues.apache.org/jira/browse/PHOENIX-3599?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Samarth Jain updated PHOENIX-3599:
----------------------------------
    Issue Type: Sub-task  (was: Task)
        Parent: PHOENIX-1598

> Take advantage of EncodedQualifiersList when there is no order by/group by 
> ---------------------------------------------------------------------------
>
>                 Key: PHOENIX-3599
>                 URL: https://issues.apache.org/jira/browse/PHOENIX-3599
>             Project: Phoenix
>          Issue Type: Sub-task
>            Reporter: Samarth Jain
>            Assignee: Samarth Jain
>
> The way we do work in our co-processors today allows us to take advantage of the EncodedColumnQualifiersCellsList only for cases when there is an order by or group by on the query (with our without a filter). We should be able to take advantage of this list in the other cases too. The way our co-processor is written, it is too late to pass in our custom list to the regionScanner.nextRaw() call. The list is already created in the HBase code and passed as a param to the scanner.nextRaw() call. We don't hit this scenario when we do order by or aggregation because in those cases phoenix does the work of aggregation or ordering itself and then passes to HBase a materialized result scanner. To be able to use the custom list implementation, we would need to enhance our EncodedColumnQualifiersCellsList to accept an additional list. We will also have to write or modify a co-processor that delegates the calls to the HRegionScanner by passing it our enhanced EncodedColumnQualifiersCellsList. 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)