[ https://issues.apache.org/jira/browse/HIVE-8512?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14211257#comment-14211257 ]
Sergey Shelukhin edited comment on HIVE-8512 at 11/13/14 8:36 PM: ------------------------------------------------------------------ The more fundamental problem is that when * is resolved, RR used it the one from preceding operator, in this case GBY. This is incorrect, since * does not refer to the columns produced by the GROUP BY. For example, this (which is a valid it would seem, although strange; src contains 2 columns) - {noformat}select *, sum(key) from src group by key, value{noformat} duplicates the SUM column, because it derives * columns from operator generated during group by was (Author: sershe): The more fundamental problem is that when * is resolved, RR used it the one from preceding operator, in this case GBY. This is incorrect, since * does not refer to the columns produced by the GROUP BY. For example, this (which is a valid it would seem, although strange; src contains 2 columns) - "select *, sum(key) from src group by key, value" duplicates the SUM column, because it derives * columns from operator generated during group by > queries with star and aggregate should fail > ------------------------------------------- > > Key: HIVE-8512 > URL: https://issues.apache.org/jira/browse/HIVE-8512 > Project: Hive > Issue Type: Bug > Reporter: Sergey Shelukhin > Assignee: Sergey Shelukhin > > ctas_colname test uses these, for example > This errors out: {noformat} select key, sum(key) from src;{noformat} > But this passes: {noformat} select *, sum(key), count(value) from > src;{noformat} > It looks like it returns 2 sums and 2 counts -- This message was sent by Atlassian JIRA (v6.3.4#6332)