[
https://issues.apache.org/jira/browse/HIVE-2711?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13238857#comment-13238857
]
Owen O'Malley commented on HIVE-2711:
-------------------------------------
The compatibility is very fragile in that SequenceFile can only read an RCFile
if they have the RCFile class since the key and value classes are contained in
RCFile.java. Additionally, SequenceFile would present the RCFile blocks instead
of the RCFile rows.
Furthermore, in the future if either RCFile or SequenceFile has a new format
revision, it breaks badly.
> Make the header of RCFile unique
> --------------------------------
>
> Key: HIVE-2711
> URL: https://issues.apache.org/jira/browse/HIVE-2711
> Project: Hive
> Issue Type: Bug
> Components: Serializers/Deserializers
> Reporter: Owen O'Malley
> Assignee: Owen O'Malley
> Attachments: HIVE-2711.D2115.1.patch
>
>
> The RCFile implementation was copied from Hadoop's SequenceFile and copied
> the 'magic' string in the header. This means that you can't use the header to
> distinguish between RCFiles and SequenceFiles.
> I'd propose that we create a new header for RCFiles (RCF?) to replace the
> current SEQ. To maintain compatibility, we'll need to continue to accept the
> current 'SEQ\06' and just make new files contain the new header.
--
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
