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-15879) Introduce Key interface
Date Fri, 17 Jun 2016 12:13:05 GMT

    [ https://issues.apache.org/jira/browse/HBASE-15879?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15335938#comment-15335938

ramkrishna.s.vasudevan commented on HBASE-15879:

bq.Having a concept for "cell having backed by byte-array" should not have a name "Key" IMO.
Thanks for the comment here. It is not only for byte-array. It is to indicate that the key
part is contiguous and we could write the key instead of trying to use them in parts like
in case of flushing to the disk.
But one thing is that every cell can have a key - either contiguous or non-contiguous. So
a Cell can have getKeyXXX with offset and length APIs and the CEll's impl can decide how to
make out the key.
The initial intention was to have this Key as an optimization for cells with contiguous key
but may be we should see if we can conceptualize this inside a Cell itself.

> Introduce Key interface
> -----------------------
>                 Key: HBASE-15879
>                 URL: https://issues.apache.org/jira/browse/HBASE-15879
>             Project: HBase
>          Issue Type: Improvement
>            Reporter: ramkrishna.s.vasudevan
>            Assignee: ramkrishna.s.vasudevan
>         Attachments: HBASE-15879.patch, HBASE-15879_1.patch
> Introduce an interface called Key and allow Cell implementations to implement this Key
interface for cases like KeyValue, OffheapKeyValue and DBE cells (Except prefix tree) so that
we can avoid copies when we deal with only Cells in case of block index creations (like ROOT,
Bloom etc). Helps in reduction of garbage also. 

This message was sent by Atlassian JIRA

View raw message