[ https://issues.apache.org/jira/browse/CASSANDRA-7445?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14042812#comment-14042812 ]
Alex Liu commented on CASSANDRA-7445: ------------------------------------- [~pauloricardomg] Code is good, can you add some Pig unit test for the thrift tables with fake clustering keys? You can find the pig unit test @https://github.com/apache/cassandra/blob/trunk/test/pig/org/apache/cassandra/pig/ThriftColumnFamilyTest.java > Support Thrift tables clustering columns on CqlPagingInputFormat > ---------------------------------------------------------------- > > Key: CASSANDRA-7445 > URL: https://issues.apache.org/jira/browse/CASSANDRA-7445 > Project: Cassandra > Issue Type: Bug > Components: Hadoop > Reporter: Paulo Motta > Assignee: Paulo Motta > Fix For: 1.2.17 > > Attachments: 1.2-CASSANDRA-7445.txt > > > CASSANDRA-5752 introduced support to thrift tables on CQLPagingInputFormat > via the retrieveKeysForThriftTables() method. > However, this method only retrieves partition keys from CFMetaData, so > clustering columns are ignored. > So, when the RowIterator tries to fetch the next page of a wide-row thrift > CF, it ignores the clustering column altogether, going to the next partition > token. So, only cassandra.input.page.row.size CQL rows are retrieved for each > partition key. > CqlRecordWriter had the same issue and was fixed on > bcfe352ea6ed3786f274b90191f988210360689d (CASSANDRA-5718). -- This message was sent by Atlassian JIRA (v6.2#6252)