[
https://issues.apache.org/jira/browse/HBASE-16310?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15408947#comment-15408947
]
ramkrishna.s.vasudevan commented on HBASE-16310:
------------------------------------------------
bq. FilterList revist logic is another topic then. That is untouched here. Or
done?
Revisiting the filterList filterRowKey() logic shows that all the other filters
should also go with what filterRowKey() in filterList does. Ok in that way may
be changing the title makes sense. Because reviewing and revisiting filterList
shows that the individual filters should be changed to respect
filterallRemaining().
bq. If filterAllRemaining() return true, filterRowKey(Cell) must return true.
It would be better to add this contract very clear in Filter javadoc. (Around
these methods) - This is some thing like a Object equal() hashCode() contract.
I won't say it is a like a contract. Rather if filterallRemaining is true then
when HRegion tries to do filterRowKey for the next cell that the Hregion
scanner is going to process it makes sense to just do that based on
FilterallRemaining().
bq.ColumnCountGetFilter - Why not changed filterRowKey(byte[], int, int)?
Those filters extends FilterBase and there filterAllRemining is not implemented
and also Filterbase has filterAllRemaining false.
I have selected all filters which has filterrowKey and filterAllRemaining impls
and only those were modified.
> Revisit the logic of filterRowKey for FilterList
> ------------------------------------------------
>
> Key: HBASE-16310
> URL: https://issues.apache.org/jira/browse/HBASE-16310
> Project: HBase
> Issue Type: Bug
> Affects Versions: 1.0.3, 1.1.5, 1.2.2
> Reporter: ramkrishna.s.vasudevan
> Assignee: ramkrishna.s.vasudevan
> Fix For: 2.0.0, 1.3.0
>
> Attachments: HBASE-16310.patch
>
>
> Based on HBASE-16926 it was observed that filterRowKey() API behaviour for
> FilterList and a standalone filter should be made consistent. This JIRA is to
> track down the changes and ensure 1.3+ on wards we have consistent behaviour.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
