hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "James Taylor (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-7692) Add utility class to generate ordered byte[] serialization
Date Sun, 03 Mar 2013 00:47:13 GMT

    [ https://issues.apache.org/jira/browse/HBASE-7692?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13591580#comment-13591580
] 

James Taylor commented on HBASE-7692:
-------------------------------------

This is definitely not just client side. Phoenix would use this in the innermost loops of
coprocessor region scans and custom filter evaluation. We need a bytes pointer mechanism,
otherwise you're forced to copy into a byte array. Reuse of the byte arrays is often not possible
because you don't know the size in advance. It certainly doesn't have to be the Hadoop ImmutableBytesWritable,
though.

Any opinion, [~lhofhansl]?

To get folks to use the Orderly stuff is a different, bigger question. If Hive and Phoenix
adopt it, you'd get those users. If we deprecate the Bytes utils for serializaing/deserializaing
and point to the new stuff (explaining why it's better), that might help. You could have a
simple cover over the Orderly classes for the common case to make it easier.
                
> Add utility class to generate ordered byte[] serialization
> ----------------------------------------------------------
>
>                 Key: HBASE-7692
>                 URL: https://issues.apache.org/jira/browse/HBASE-7692
>             Project: HBase
>          Issue Type: Improvement
>          Components: util
>            Reporter: Nick Dimiduk
>            Assignee: Nick Dimiduk
>             Fix For: 0.95.0
>
>         Attachments: HBASE-7692.v1.patch, HBASE-7692.v2.patch, HBASE-7692.v3.patch, HBASE-7692.v4.patch,
HBASE-7692.v5.patch
>
>
> The current Bytes utility class works, but produces output that does not maintain the
native sort ordering of the input value. This results in, for example, a negative value that
does not necessarily sort before a positive value. HBase should provide a canonical implementation
of such a serialization format so that third-parties can reliably build on top of HBase. This
will allow an implementation for HIVE-3634, HIVE-2599, or HIVE-2903 that is compatible with
similar features in Pig.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message