[
https://issues.apache.org/jira/browse/HBASE-28633?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Chandra Sekhar K reassigned HBASE-28633:
----------------------------------------
Assignee: Chandra Sekhar K
> Make sure that filterAllRemaining() is not called from filterRowKey()
> ---------------------------------------------------------------------
>
> Key: HBASE-28633
> URL: https://issues.apache.org/jira/browse/HBASE-28633
> Project: HBase
> Issue Type: Improvement
> Reporter: Istvan Toth
> Assignee: Chandra Sekhar K
> Priority: Minor
> Labels: performance
>
> Many of the current filters call filterAllRemaining() from filterRowKey().
> This should not be necessary, as in the normal (RS) code path, filterRowKey()
> is only called AFTER filterAllReamaining() has returned false.
> Well-written filters do cache their filterAllReamaining() status, so this is
> not very expensive, but we could still save a few cycles for each cell.
> * Change the filter API definition to explicitly state this
> * Fix the code where this is not true. At first glance,
> org.apache.hadoop.hbase.mapreduce.Import seems to be one place that does not
> confirm to this behaviour.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
