hive-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Thejas M Nair (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HIVE-3168) LazyBinaryObjectInspector.getPrimitiveJavaObject copies beyond length of underlying BytesWritable
Date Fri, 06 Jul 2012 16:31:35 GMT

    [ https://issues.apache.org/jira/browse/HIVE-3168?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13408135#comment-13408135
] 

Thejas M Nair commented on HIVE-3168:
-------------------------------------

Neha,
Try 'ant clean' on the hive dir before building it. 
To verify that the patched hive code is getting used, try adding a debug message in LazyBinaryObjectInspector
around line 43 (after applying patch) (note the debug message would come out of map task).
Or change the data setting the first byte, and see if changes your result -
{code}
    System.arraycopy(bWritable.getBytes(), 0, data, 0, bWritable.getLength());
    ba.setData(data);
{code}
to 
{code}
    System.arraycopy(bWritable.getBytes(), 0, data, 0, bWritable.getLength());
    data[0]='Z';
    ba.setData(data);
{code}
I am not sure if I tested with hcat 0.4 branch or trunk. But I don't think there has been
any change that should affect this behavior, it should work with hcat 0.4 release as well.

                
> LazyBinaryObjectInspector.getPrimitiveJavaObject copies beyond length of underlying BytesWritable
> -------------------------------------------------------------------------------------------------
>
>                 Key: HIVE-3168
>                 URL: https://issues.apache.org/jira/browse/HIVE-3168
>             Project: Hive
>          Issue Type: Bug
>          Components: Serializers/Deserializers
>    Affects Versions: 0.9.0
>            Reporter: Thejas M Nair
>            Assignee: Thejas M Nair
>             Fix For: 0.10.0, 0.9.1
>
>         Attachments: HIVE-3168.1.patch, HIVE-3168.2.patch
>
>
> LazyBinaryObjectInspector.getPrimitiveJavaObject copies the full capacity of the LazyBinary's
underlying BytesWritable object, which can be greater than the size of the actual contents.

> This leads to additional characters at the end of the ByteArrayRef returned. When the
LazyBinary object gets re-used, there can be remnants of the later portion of previous entry.

> This was not seen while reading through hive queries, which I think is because a copy
elsewhere seems to create LazyBinary with length == capacity. (probably LazyBinary copy constructor).
This was seen when MR or pig used Hcatalog to read the data.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message