Mailing-List: contact commits-help@beam.incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@beam.incubator.apache.org
Date: Fri, 8 Jul 2016 05:22:11 +0000 (UTC)
From: "Daniel Halperin (JIRA)" <jira@apache.org>
To: commits@beam.incubator.apache.org
Message-ID: <JIRA.12984192.1467078286000.54933.1467955331073@Atlassian.JIRA>
In-Reply-To: <JIRA.12984192.1467078286000@Atlassian.JIRA>
References: <JIRA.12984192.1467078286000@Atlassian.JIRA> <JIRA.12984192.1467078286843@arcas>
Subject: [jira] [Resolved] (BEAM-381) OffsetBasedReader should construct
 sources before updating the range tracker
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
archived-at: Fri, 08 Jul 2016 05:22:15 -0000


     [ https://issues.apache.org/jira/browse/BEAM-381?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Daniel Halperin resolved BEAM-381.
----------------------------------
    Resolution: Fixed

> OffsetBasedReader should construct sources before updating the range tracker
> ----------------------------------------------------------------------------
>
>                 Key: BEAM-381
>                 URL: https://issues.apache.org/jira/browse/BEAM-381
>             Project: Beam
>          Issue Type: Bug
>          Components: sdk-java-core
>    Affects Versions: 0.1.0-incubating, 0.2.0-incubating
>            Reporter: Daniel Halperin
>            Assignee: Daniel Halperin
>             Fix For: 0.2.0-incubating
>
>
> OffsetBasedReader has the following code:
> {code}
>       if (!rangeTracker.trySplitAtPosition(splitOffset)) {
>         return null;
>       }
>       long start = source.getStartOffset();
>       long end = source.getEndOffset();
>       OffsetBasedSource<T> primary = source.createSourceForSubrange(start, splitOffset);
>       OffsetBasedSource<T> residual = source.createSourceForSubrange(splitOffset, end);
>       this.source = primary;
>       return residual;
> {code}
> The first line is the line that updates the range of this source. However, subsequent lines might throw (specifically, in source.createSourceForSubrange). We should construct the sources first, and then catch exceptions and return null if they fail. This way, the splitAtFraction call will not throw (so work is not wasted) and the range tracker will not be updated if either the primary or (more likely) the residual could not be created.


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