[
https://issues.apache.org/jira/browse/HBASE-16143?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15354885#comment-15354885
]
ramkrishna.s.vasudevan commented on HBASE-16143:
------------------------------------------------
bq.Why use pos based add? It is ArrayList we know so we will get insert order.
It is fine to just call add(<element>)
Ok. will change. I just used the way it was done in DefaultMemstore.
bq.Where do u use MemstoreScanner now? Previous it was returning one
MemstoreScanner to the StoreScannr layer.Now it is N KVScanners? Intended
change?
I verified. No problem any where. It should be fine. But to ensure that we
don't introduce such a big change with this patch (though it is right it causes
confusion) - I will just wrap the list into a MemstoreScanner and return a
singleton list as it was done before.
> Change MemstoreScanner constructor to accept List<KeyValueScanner>
> ------------------------------------------------------------------
>
> Key: HBASE-16143
> URL: https://issues.apache.org/jira/browse/HBASE-16143
> Project: HBase
> Issue Type: Sub-task
> Affects Versions: 2.0.0
> Reporter: ramkrishna.s.vasudevan
> Assignee: ramkrishna.s.vasudevan
> Priority: Minor
> Fix For: 2.0.0
>
> Attachments: HBASE-16143.patch, HBASE-16143_1.patch
>
>
> A minor change that helps in creating a memstore that avoids the compaction
> process and just allows to creates a pipeline of segments and on flush
> directly reads the segments in the pipeline and flushes it out after creating
> a snapshot of the pipeline. Based on test results and updated patch on
> HBASE-14921 (to be provided) will see how much flattening helps us.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
