carbondata-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ASF GitHub Bot (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CARBONDATA-308) Use CarbonInputFormat in CarbonScanRDD compute
Date Mon, 07 Nov 2016 14:36:58 GMT

    [ https://issues.apache.org/jira/browse/CARBONDATA-308?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15644342#comment-15644342
] 

ASF GitHub Bot commented on CARBONDATA-308:
-------------------------------------------

Github user jackylk commented on a diff in the pull request:

    https://github.com/apache/incubator-carbondata/pull/262#discussion_r86786637
  
    --- Diff: core/src/main/java/org/apache/carbondata/core/carbon/datastore/block/Distributable.java
---
    @@ -16,10 +16,12 @@
      */
     package org.apache.carbondata.core.carbon.datastore.block;
     
    +import java.io.IOException;
    +
     /**
    - * Abstract class which is maintains the locations of node.
    + * interface to get the locations of node. Used for making task distribution based on
locality
      */
    -public abstract class Distributable implements Comparable<Distributable> {
    +public interface Distributable extends Comparable<Distributable> {
     
    -  public abstract String[] getLocations();
    +  String[] getLocations() throws IOException;
    --- End diff --
    
    Because CarbonInputSplit need to implement Distributable, and InputSplit has a getLocation
function that throws IOException


> Use CarbonInputFormat in CarbonScanRDD compute
> ----------------------------------------------
>
>                 Key: CARBONDATA-308
>                 URL: https://issues.apache.org/jira/browse/CARBONDATA-308
>             Project: CarbonData
>          Issue Type: Sub-task
>          Components: spark-integration
>            Reporter: Jacky Li
>            Assignee: Jacky Li
>             Fix For: 0.2.0-incubating
>
>
> Take CarbonScanRDD as the target RDD, modify as following:
> 1. In driver side, only getSplit is required, so only filter condition is required, no
need to create full QueryModel object, so we can move creation of QueryModel from driver side
to executor side.
> 2. use CarbonInputFormat.createRecordReader in CarbonScanRDD.compute instead of use QueryExecutor
directly



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

Mime
View raw message