[ 
https://issues.apache.org/jira/browse/HBASE-4364?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13435292#comment-13435292
 ] 

Lucas Bernardi commented on HBASE-4364:
---------------------------------------


Hi there, we're facing a similar issue. But I guess the first question would 
be, what is the expected behavior for the treatment of requested columns and 
filtered columns?
One possible use case is when I want to get a specific set of columns, but 
filter the row using another set of columns. For this use case it looks like 
the Filter should always be applied to all columns, regardless of the columns 
requested by the client. This would be SQL-like 'where' and 'projection'.
But, hbase is supposed to be able to handle tons of columns, so performance 
wise, the SQL-like behavior doesn't look like a good idea to me.
I think the behavior should be: filter is only applied on requested columns, a 
more bigdata oriented behavior. Or, may be add the posibility to specify, a set 
of 'projected' columns and another set of filtered 'columns'.

Anyway, it looks like 0.92.1 does something really weird. It will apply the 
filter on the first column and the 'projected' columns. This is closer to the 
bigdata oriented behavior, but with a bug, so, I think it should be fixed 
first, meaning just apply the filter to requested columns vs first column + 
requested columns. Once that works, we can think about filtering based on other 
columns.


                
> Filters applied to columns not in the selected column list are ignored
> ----------------------------------------------------------------------
>
>                 Key: HBASE-4364
>                 URL: https://issues.apache.org/jira/browse/HBASE-4364
>             Project: HBase
>          Issue Type: Bug
>          Components: filters
>    Affects Versions: 0.90.4, 0.92.0, 0.94.0
>            Reporter: Todd Lipcon
>            Priority: Critical
>         Attachments: 
> HBASE-4364-failing-test-with-simplest-custom-filter.patch, 
> hbase-4364_trunk.patch
>
>
> For a scan, if you select some set of columns using addColumns(), and then 
> apply a SingleColumnValueFilter that restricts the results based on some 
> other columns which aren't selected, then those filter conditions are ignored.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: 
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Reply via email to