[jira] [Commented] (HAWQ-931) HiveORCAccessor with support for Predicate pushdown and Column Projection

["Shivram Mani (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HAWQ-931?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15397813#comment-15397813 ] 

Shivram Mani commented on HAWQ-931:
-----------------------------------

The current approach is to use Hive 1.2.1's org.apache.hadoop.hive.ql.io.sarg.SearchArgument and org.apache.hadoop.hive.ql.io.sarg.SearchArgumentFactory to build the Predicate from the HAWQ predicate. 
These apis have changed with Hive 2.0 and will need to be revisited.
Specifically with functions on SearchArgument.Builder also need the data type.
eg: builder.lessThanEquals(filterColumnName, filterValue) will chagne to builder.lessThanEquals(filterColumnName, type, filterValue);

> HiveORCAccessor with support for Predicate pushdown and Column Projection
> -------------------------------------------------------------------------
>
>                 Key: HAWQ-931
>                 URL: https://issues.apache.org/jira/browse/HAWQ-931
>             Project: Apache HAWQ
>          Issue Type: Sub-task
>          Components: PXF
>            Reporter: Shivram Mani
>            Assignee: Shivram Mani
>             Fix For: backlog
>
>
> HiveORCAccessor will be a subclass of the existing HiveAccessor with support for Predicate Pushdown and Column projection.
> We will be using the job configuration object which is used by the ORC reader.
> We will  map the filter information passed from HAWQ via PXF into SearchArgument object and set in sarg.pushdown configuration property.
> We will populate the column info passed from HAWQ (HAWQ-927) into the following configuration properties
> hive.io.file.readcolumn.ids, hive.io.file.readcolumn.names. hive.io.file.read.all.columns will be set to false.



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