hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Anoop Sam John (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-10771) Primitive type put/get APIs in ByteRange
Date Wed, 26 Mar 2014 04:42:16 GMT

    [ https://issues.apache.org/jira/browse/HBASE-10771?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13947544#comment-13947544
] 

Anoop Sam John commented on HBASE-10771:
----------------------------------------

bq. Now we have to have two code paths... one for those with backing byte arrays and then
another for those offheap?
That need not be.  Now we have put/get APIs in BR.  so the impl can know whether it is on
heap or offheap and do the best way of put/get.
Another thing can be the comprator. Here also we can come up with a comparator and it can
have one code path only which can work with offheap also.

Adding of this kind of an API will allow some one to have 2 code paths *if needed*..
As u can see in the path, the put(BR) itself uses this new API.  When it is onheap (the src
BR) System.arrayCopy can be the best way of copy rather than byte by byte assign (?)  When
on heap we have no such choice..

So we wont have 2 code paths based on this.. But may be at some use places we *can have also*
like above case. (Now or later)

How that sounds Stack?


> Primitive type put/get APIs in ByteRange 
> -----------------------------------------
>
>                 Key: HBASE-10771
>                 URL: https://issues.apache.org/jira/browse/HBASE-10771
>             Project: HBase
>          Issue Type: Improvement
>            Reporter: Anoop Sam John
>            Assignee: Anoop Sam John
>             Fix For: 0.99.0
>
>         Attachments: HBASE-10771.patch, HBASE-10771_V2.patch
>
>
> While doing HBASE-10713 I came across the need to write int/long (and read also) from
a ByteRange.  CellBlocks are backed by ByteRange. So we can add such APIs.
> Also as per HBASE-10750  we return a ByteRange from MSLAB and also discussion under HBASE-10191
suggest we can have BR backed HFileBlocks etc.  



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message