In general I think it would be beneficial if a Coprocessor could interfere with the sequence number logic. The problem is that it would constitute an API change.
-- Lars ________________________________ From: Ted Yu <yuzhih...@gmail.com> To: dev@hbase.apache.org Sent: Saturday, June 2, 2012 7:40 AM Subject: [pre,post]ScannerOpen() returning a custom RegionScanner implementation Hi, In reviewing HBASE-5974, Anoop brought up the following consideration ( https://issues.apache.org/jira/browse/HBASE-5974?focusedCommentId=13285365&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-13285365 ): Why I was not doing that because through the CP we have exposed the RegionScanner interface. The pre and post scannerOpen methods in the CP can return an impl for the RegionScanner now. So in that case customer need to worry abt maintain this seqNo? I felt that wont be good. If anyone utilizes [pre,post]ScannerOpen() to return a custom RegionScanner implementation, would you please share your use case ? Thanks
