[
https://issues.apache.org/jira/browse/HBASE-12374?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14633597#comment-14633597
]
Anoop Sam John commented on HBASE-12374:
----------------------------------------
Ya I can do the rename.
bq.Will it be better to copy the tags part incase of tagCompressionContext ==
null && includeTags case - using a seperate asSubByteBuffer call rather than
copying them in one call?
My thinking here was, mostly there wont be copy need for this asSubBuffer call.
Even if copy is needed, we will do it is one call and make one object
including both tags and value part (tags here is not compressed at all). The
only case where this is an overhead (one call rather than 2 calls for value and
tags) the value part ends exactly at one item BB and tags part starts in next
BB. Then one call for value and tags will need a copy where as the other model
dont need at all. Very rare chance.
2 calls to asSubBuffer having overhead of calculations in MBB like finding the
position and item BB. IMO current way is ok.
> Change DBEs to work with new BB based cell
> ------------------------------------------
>
> Key: HBASE-12374
> URL: https://issues.apache.org/jira/browse/HBASE-12374
> Project: HBase
> Issue Type: Sub-task
> Components: regionserver, Scanners
> Reporter: ramkrishna.s.vasudevan
> Assignee: Anoop Sam John
> Attachments: HBASE-12374_v1.patch, HBASE-12374_v2.patch,
> HBASE-12374_v3.patch
>
>
> Once we are changing the read path to use BB based cell then the DBEs should
> also return BB based cells. Currently they are byte[] array backed.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
