You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mahout.apache.org by "Sean Owen (JIRA)" <ji...@apache.org> on 2011/04/04 20:59:06 UTC

[jira] [Commented] (MAHOUT-651) Pass hadoop configuration to methods that use FileSystem operations, even if they don't invoke map/reduce jobs

    [ https://issues.apache.org/jira/browse/MAHOUT-651?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13015548#comment-13015548 ] 

Sean Owen commented on MAHOUT-651:
----------------------------------

Completely agree. It's likely a source of subtle bugs. I adjusted a lot of methods recently along these lines. I'll try to commit your patch shortly.

> Pass hadoop configuration to methods that use FileSystem operations, even if they don't invoke map/reduce jobs
> --------------------------------------------------------------------------------------------------------------
>
>                 Key: MAHOUT-651
>                 URL: https://issues.apache.org/jira/browse/MAHOUT-651
>             Project: Mahout
>          Issue Type: Improvement
>          Components: Clustering
>    Affects Versions: 0.4
>            Reporter: Robert Mahfoud
>             Fix For: 0.5
>
>         Attachments: patch-mahout-651.txt
>
>
> Some classes in the Classification component internally use the hadoop's FileSystem class, however, they instantiate the hadoop configuration locally in the method using {{new Configuration()}}. This limits the ability to integrate these tools within applications that manage and enrich their own configuration rather than rely on the default hadoop resources that get loaded when calling {{new Configuration()}}.
> The fix is simply to make these methods take a {{Configuration}} parameter rather than creating a new instance when needed. An example for an that creates a new {{Configuration}} instances is: {{org.apache.mahout.clustering.kmeans.KMeansUtil.configureWithClusterInfo(Path, List<Cluster>)}}
> This problem could also exists beyond the Clustering module, but this issue only addresses the Clustering code.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira