[ https://issues.apache.org/jira/browse/CASSANDRA-8946?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14363272#comment-14363272 ]
Sylvain Lebresne commented on CASSANDRA-8946: --------------------------------------------- Patch seems to cover it all and thanks for the thorough testing. Just 2 minor feedbacks: * the naming of the {{Boundary}} related functions ({{nonEmpty}}, {{interesectLeft}} and {{interesectRight}}) don't speak to me intuitively. For {{nonEmpty}}, a static {{isEmpty(left, right)}} would feel more natural. For the others, I would have added a {{leftBoundary()/rightBoundary()}} to {{AbstractBounds}} and just call the methods {{min}} and {{max}} (on 2 boundaries). * Could invert the cases of the {{isWrapAround}} case to generate bounds in disk order (note that it matters much, just feels more natural to me). > Make SSTableScanner always respect its bound > -------------------------------------------- > > Key: CASSANDRA-8946 > URL: https://issues.apache.org/jira/browse/CASSANDRA-8946 > Project: Cassandra > Issue Type: Improvement > Components: Core > Reporter: Sylvain Lebresne > Assignee: Benedict > Fix For: 2.1.4 > > > When {{SSTableScanner}} takes a {{DataRange}}, it doesn't fully respect the > bounds provided as it always generate a {{Bounds}} object, thus potentially > ending up including a key it should have excluded. It's currently compensated > by in {{ColumnFamilyStore.getSequentialIterator}} but that is still an > unexpected behavior and is such error prone. We should fix that and remove > the compensation in {{ColumnFamilyStore.getSequentialIterator}}. -- This message was sent by Atlassian JIRA (v6.3.4#6332)