hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "stack (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HBASE-1189) Changing the map type used internally for HbaseMapWritable.
Date Sat, 07 Feb 2009 20:06:59 GMT

    [ https://issues.apache.org/jira/browse/HBASE-1189?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12671518#action_12671518

stack commented on HBASE-1189:

Hey Erik.  Which map are we talking of?  Is it CODE_TO_CLASS (and its inverse) or is it
  private SortedMap<byte [], V> instance =
    new TreeMap<byte [], V>(Bytes.BYTES_COMPARATOR);

If the former, thats once for all instances of HBW.  If latter, you want to just something
other than TreeMap?

If you subclass, you think there much overhead?  8 bytes or so?  You trying to be parsimonious?
 Good stuff.

> Changing the map type used internally  for HbaseMapWritable. 
> -------------------------------------------------------------
>                 Key: HBASE-1189
>                 URL: https://issues.apache.org/jira/browse/HBASE-1189
>             Project: Hadoop HBase
>          Issue Type: New Feature
>          Components: io
>            Reporter: Erik Holstad
>            Assignee: Erik Holstad
>            Priority: Minor
> For the CellCache the need for a HbaseMapWritable that has a different kind of map used
> has showed up. The problem is that the instantiation of the map is done static so you
have no control 
> over it and if extending HMW you get both the memory usage for the parent as well as
the child.
> After trying out different ideas on how to solve this, it seems like the easiest way
would be to have a setMapType
> method and just set it to null in the code. If the old code would be kept you would need
to instantiate 2
> different maps. The problem with setting it to null is that all the old code needs to
be changed to fit the new model,
> and also be used in the future.

This message is automatically generated by JIRA.
You can reply to this email to add a comment to the issue online.

View raw message