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-10453) PrefixTree encoding for tags can honor HCD#shouldCompressTags()
Date Tue, 04 Feb 2014 04:18:11 GMT

    [ https://issues.apache.org/jira/browse/HBASE-10453?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13890362#comment-13890362

ramkrishna.s.vasudevan commented on HBASE-10453:

Am just thinking if we really need to address this. What I felt is adding the tags in row,family,qualifer
part (as we are doing now) is a way of making the prefix tree work with tags. 
If we say not to handle that way then we need to handle the way we do handle value part in
the prefix tree.  Anyway the prefix tree meta data has to be changed to still accomodate this
but the only thing is it is not part of the trie structure. 
I was not totally sure as how to proceed here.  Also, in one of the reviews of V3 with tags
it was suggested to handle tags in the prefix tree trie itself and not as the value portion.
As Andy pointed out in the other JIRA HBASE-10454, can we say in doc like the compress tags
applies only in case of other encoding tags other than Prefix Tree.  Because the dictionary
based compression is used in that algo and not here?

> PrefixTree encoding for tags can honor HCD#shouldCompressTags()
> ---------------------------------------------------------------
>                 Key: HBASE-10453
>                 URL: https://issues.apache.org/jira/browse/HBASE-10453
>             Project: HBase
>          Issue Type: Bug
>    Affects Versions: 0.98.0
>            Reporter: Anoop Sam John
>            Assignee: ramkrishna.s.vasudevan
>             Fix For: 0.98.1
> Right now we will encode tags always with prefix tree encoding. With other encoders it
checks HCD#shouldCompressTags().  Suggest we can do the same fro PrefixTree also.
> I can see some places PrefixTree impl passes booleans like isIncludeMvcc isIncludeTags
etc.  We have encapsulated all such info into a HFileContext and the code path uses that now.
 We can do the same with PrefixTree code path also.

This message was sent by Atlassian JIRA

View raw message