hadoop-pig-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ankur (JIRA)" <j...@apache.org>
Subject [jira] Updated: (PIG-958) Splitting output data on key field
Date Mon, 26 Oct 2009 14:30:59 GMT

     [ https://issues.apache.org/jira/browse/PIG-958?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Ankur updated PIG-958:

    Attachment: 958.v4.patch

1. When run in cluster mode, static variable PigMapReduce.sJobConf is null when checked in
the UDF constructor but NOT null when UDF is actually invoked. This  causes incorrect initialization
of FileSystem object 'fs' to local filesystem, causing the test to fail. Moved to 'fs' initialization
to intijobSpecificParams() method.

2. Deleting the temporary directory manually in finish(), causes the job to fail. Removed
the manual deletion. As a side effect, user specified PARENT output directory in the UDF will
have empty part-* files. These should be deleted manually by the user.

Verfied that UDF works correctly and that unit test pass

> Splitting output data on key field
> ----------------------------------
>                 Key: PIG-958
>                 URL: https://issues.apache.org/jira/browse/PIG-958
>             Project: Pig
>          Issue Type: Bug
>    Affects Versions: 0.4.0
>            Reporter: Ankur
>         Attachments: 958.v3.patch, 958.v4.patch
> Pig users often face the need to split the output records into a bunch of files and directories
depending on the type of record. Pig's SPLIT operator is useful when record types are few
and known in advance. In cases where type is not directly known but is derived dynamically
from values of a key field in the output tuple, a custom store function is a better solution.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message