hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jim Kellerman (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HBASE-584) Names in the filter interface are confusing
Date Tue, 22 Apr 2008 18:23:22 GMT

    [ https://issues.apache.org/jira/browse/HBASE-584?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12591394#action_12591394
] 

Jim Kellerman commented on HBASE-584:
-------------------------------------

Clint,

Our policy is that all APIs need to be supported for one release past the point where they
are deprecated. Some users do not live on the bleeding edge and need time to adjust. Deprecating
existing interfaces while providing the new interface gives them warning that they will need
to change their code. One of the reasons I suggested making this a patch for 0.1 branch is
that if we deprecate it in 0.1, we can remove it in 0.2

However, if this is only going to be a 0.2 patch, then users need 1 release cycle to upgrade
to the new API. So if this is a 0.2 patch, the existing APIs need to be supported until 0.3
comes out.

> Names in the filter interface are confusing
> -------------------------------------------
>
>                 Key: HBASE-584
>                 URL: https://issues.apache.org/jira/browse/HBASE-584
>             Project: Hadoop HBase
>          Issue Type: Improvement
>          Components: filters
>            Reporter: Clint Morgan
>            Assignee: Clint Morgan
>            Priority: Minor
>         Attachments: hbase-584.patch
>
>
> I don't like the names of the filter methods in RowFilterInterface. They don't really
tell how the methods are being used in the implementation of scanners.
> I'd like to change:
>  - filter(Text) to filterRow(...)
>  - filter(Text, Text, byte[]) to filterColumn(...)
>   and the worst one is
>  - filterNotNull(SortedMap<Text, byte[]>). This should be filterRow(Text, SortedMap<Text,
byte[]>) (so we add the row key/).
> It may be nice to have timestamps in the methods as well? 
> Also the java doc could be cleaned and improved to tell how the filtering is implemented
(check rows keys first, then check each individual columns, finally check the assembled row)
> Upon positive feedback, and I'll create a patch.

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


Mime
View raw message