[
https://issues.apache.org/jira/browse/HBASE-10771?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13955728#comment-13955728
]
stack commented on HBASE-10771:
-------------------------------
Thanks [~ndimiduk]
And I like how you put it. My comments are of the same class as yours (to be
clear).
We have ByteBuffers coming in the front door currently. We give the socket a
nioBB to read into. Keeping position in another object, BR for instance seems
fine (so we avoid #1 and #2 in @apurtell list above).
Reading in from HDFS, we allocate a BB and read into it.
Looking at ByteBuf, it has range checking (checkIndex) but the getBytes and the
checkIndex are subclasseable so perhaps it could be 'turned off' if we wanted
it to be. ByteBuf has nice features like allocations from pools but if we are
talking tight loops (no range check) and reuse, then neither nioBB nor nettyBB
will do.
> 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)
