[ 
https://issues.apache.org/jira/browse/TAJO-2109?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15234654#comment-15234654
 ] 

ASF GitHub Bot commented on TAJO-2109:
--------------------------------------

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

    https://github.com/apache/tajo/pull/992#discussion_r59166686
  
    --- Diff: 
tajo-core/src/main/java/org/apache/tajo/engine/planner/physical/ExternalSortExec.java
 ---
    @@ -133,6 +140,22 @@ private ExternalSortExec(final TaskAttemptContext 
context, final SortNode plan)
         this.localFS = new RawLocalFileSystem();
         this.intermediateMeta = CatalogUtil.newTableMeta(BuiltinStorages.DRAW);
         this.inputStats = new TableStats();
    +    this.sortAlgorithm = getSortAlgorithm(context.getQueryContext(), 
sortSpecs);
    +  }
    +
    +  private static SortAlgorithm getSortAlgorithm(QueryContext context, 
SortSpec[] sortSpecs) {
    +    if (Arrays.stream(sortSpecs)
    +        .filter(sortSpec -> !RadixSort.isApplicableType(sortSpec)).count() 
> 0) {
    +      return SortAlgorithm.TIM_SORT;
    +    }
    +    String sortAlgorithm = context.get(SessionVars.SORT_ALGORITHM, "TIM");
    +    if (sortAlgorithm.equalsIgnoreCase("TIM")) {
    --- End diff --
    
    remove hardcode constants


> Implement Radix sort
> --------------------
>
>                 Key: TAJO-2109
>                 URL: https://issues.apache.org/jira/browse/TAJO-2109
>             Project: Tajo
>          Issue Type: New Feature
>          Components: Sort algorithm
>            Reporter: Jihoon Son
>            Assignee: Jihoon Son
>             Fix For: 0.12.0
>
>
> Radix sort is known for very fast sort algorithm when the length of the sort 
> key is not long. We can benefit from Radix sort if it is used when it is 
> faster than Tim sort.
> In this issue, I will implement Radix sort for Tajo, and conduct some 
> benchmark tests to compare its performance with Tim sort.



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

Reply via email to