Mailing-List: contact dev-help@flink.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@flink.apache.org
Date: Thu, 20 Oct 2016 16:33:59 +0000 (UTC)
From: "Greg Hogan (JIRA)" <jira@apache.org>
To: dev@flink.apache.org
Message-ID: <JIRA.13013956.1476981220000.41037.1476981239836@Atlassian.JIRA>
In-Reply-To: <JIRA.13013956.1476981220000@Atlassian.JIRA>
References: <JIRA.13013956.1476981220000@Atlassian.JIRA> <JIRA.13013956.1476981220696@arcas>
Subject: [jira] [Created] (FLINK-4869) Store record pointer after record
 keys
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
archived-at: Thu, 20 Oct 2016 16:34:01 -0000

Greg Hogan created FLINK-4869:
---------------------------------

             Summary: Store record pointer after record keys
                 Key: FLINK-4869
                 URL: https://issues.apache.org/jira/browse/FLINK-4869
             Project: Flink
          Issue Type: Sub-task
          Components: Core
    Affects Versions: 1.2.0
            Reporter: Greg Hogan
            Assignee: Greg Hogan
            Priority: Minor


{{NormalizedKeySorter}} serializes records into a {{RandomAccessInputView}} separate from the memory segments used for the sort keys. By storing the pointer after the sort keys the addition of the offset is moved from {{NormalizedKeySorter.compare}} which is O(n log n)) to other methods which are O\(n).

Will run a performance comparison before submitting a PR to how significant a performance improvement this would yield.


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