drill-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Chris Westin (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (DRILL-2261) Replace HashMap<Integer, DataCollector> in IncomingBuffers with an array based data structure
Date Tue, 26 May 2015 20:44:36 GMT

    [ https://issues.apache.org/jira/browse/DRILL-2261?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14559833#comment-14559833
] 

Chris Westin commented on DRILL-2261:
-------------------------------------

We don't have any uses of Trove yet, so it might be better to stick with the carrotsearch
IntObjectOpenHashMap, which we already use.

> Replace HashMap<Integer, DataCollector> in IncomingBuffers with an array based
data structure
> ---------------------------------------------------------------------------------------------
>
>                 Key: DRILL-2261
>                 URL: https://issues.apache.org/jira/browse/DRILL-2261
>             Project: Apache Drill
>          Issue Type: Improvement
>          Components: Execution - Flow
>    Affects Versions: 0.7.0
>            Reporter: Venki Korukanti
>            Assignee: Venki Korukanti
>            Priority: Minor
>             Fix For: 1.2.0
>
>
> Currently we have Map<Integer, DataCollector> which stores the mapping of major
fragment id and DataCollector instance pairs. For every record batch arrived, we need to autobox
primitive integer and hash. This is unnecessary cost, instead we could have a simpler array
based structure like TIntArrayList (trove collection) or custom array based structure. Key
value (in this case major fragment id) is between 0 and numStages - 1. numStages is not going
to be a huge number. Utilizing this range we could either use TIntArrayList or custom array
based structure.   



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

Mime
View raw message