hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "ramkrishna.s.vasudevan (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-8849) CellCodec should write and read the memstoreTS/mvccVersion
Date Thu, 04 Jul 2013 02:59:21 GMT

    [ https://issues.apache.org/jira/browse/HBASE-8849?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13699730#comment-13699730
] 

ramkrishna.s.vasudevan commented on HBASE-8849:
-----------------------------------------------

bq.The style should be more MvccCellCodec for the codec that includes mvcc and another NoMvccCellCodec
for the codec that does not. For now I'd say just under the wraps have CellCodec encode the
mvcc.
I agree with you on this.  Infact individual codecs should take care of handling mvcc.  But
how should the read path be in my HFileReader?  Should i have 2 paths where if a codec that
does not handle mvcc with in itself should be handled specifically and the codec that handles
mvcc within itself should be handled specifically? This will make things complex.
With this codec based approach we should stick to a strategy wrt mvcc otherwise the Read path
(infact even write path) will have to handle them.  If  we are ok on this then i can make
a base class saying MvccCellCodec and NoMvccCellCodec  - so any new CellCodec would be inheriting
either one of them and in the read/write path we need to switch based on what codec is used?
Does this make sense?  
                
> CellCodec should write and read the memstoreTS/mvccVersion
> ----------------------------------------------------------
>
>                 Key: HBASE-8849
>                 URL: https://issues.apache.org/jira/browse/HBASE-8849
>             Project: HBase
>          Issue Type: Bug
>    Affects Versions: 0.95.1
>            Reporter: ramkrishna.s.vasudevan
>            Assignee: ramkrishna.s.vasudevan
>              Labels: keyvalue
>             Fix For: 0.98.0
>
>         Attachments: HBASE-8849_1.patch, HBASE-8849_2.patch
>
>
> This JIRA is wrt discussion over in HBASE-8496.
> Cell interface and the corresponding CellCodec provides a new way of serializing the
Keyvalues.  
> Cell interface supports getMvccVersion() and the javadoc says it could be > 0 if it
exists or 0 otherwise.
> But we don't tend to write/read the memstoreTS/mvccVersion in the Cell codec.

--
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