[ https://issues.apache.org/jira/browse/CASSANDRA-7100?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13982412#comment-13982412 ]
Rohit Rai commented on CASSANDRA-7100: -------------------------------------- I was not able to reopen #CASSANDRA-5718 so created this clone. This is a regression in 2.0.x as the code to support reversed type was overwritten in a merge. The culprit diff here - https://git-wip-us.apache.org/repos/asf?p=cassandra.git;a=blobdiff;f=src/java/org/apache/cassandra/hadoop/cql3/CqlPagingRecordReader.java;h=d9b9a39cebe0ec50e5e80c6f8057d4cce0646b6a;hp=b6e793c7e46adf3694c71e365833960c650c8152;hb=3a4d6beb;hpb=8e7d7285cdeac4f2527c933280d595bbddd26935 > Regression in 2.0.x - Cql3 reader returns duplicate rows if the cluster > column is reversed > ------------------------------------------------------------------------------------------ > > Key: CASSANDRA-7100 > URL: https://issues.apache.org/jira/browse/CASSANDRA-7100 > Project: Cassandra > Issue Type: Bug > Affects Versions: 1.2.6 > Reporter: Rohit Rai > Assignee: Alex Liu > Fix For: 1.2.9 > > > To reproduce it, > cqlsh:test> select * from wordfreq; > title | occurances | word > ---------+------------+------- > alex123 | 4 | liu3 > alex1 | 23456 | liu2 > alex10 | 10 | liu10 > alex12 | 34 | liu3 > alex | 123456 | liu1 > alex | 1000 | liu > CREATE TABLE wordfreq ( title text, word text, occurances int, PRIMARY KEY > (title,occurances)) WITH CLUSTERING ORDER by (occurances DESC); > The hadoop job returns 7 rows instead of 6 rows. > I will post a patch soon. -- This message was sent by Atlassian JIRA (v6.2#6252)