[ https://issues.apache.org/jira/browse/HIVE-12301?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14991535#comment-14991535 ]
Jesus Camacho Rodriguez commented on HIVE-12301: ------------------------------------------------ It seems like a copy-paste error. [~pxiong], with the patch, the name is taken from rsColInfoLst. If we have a DISTINCT aggregate function, the name is replaced by a hard-coded-autogenerated column name. It seems to me that the complete solution would be to rewrite that part of genReduceSideGB1NoMapGB so we rely on the {{udafParamsIndxInGBInfoDistExprs}} data structure, as the genReduceSideGB1 does. Did you consider this solution? Is it feasible? It certainly would be a better practice than hard-coding the name of a column as it is done now. Further, could we add a test for UDAF referencing a column with DISTINCT to check that hits this path, to verify that both cases are working as expected? Thanks > CBO: Calcite Operator To Hive Operator (Calcite Return Path): fix test > failure for udf_percentile.q > --------------------------------------------------------------------------------------------------- > > Key: HIVE-12301 > URL: https://issues.apache.org/jira/browse/HIVE-12301 > Project: Hive > Issue Type: Sub-task > Components: CBO > Reporter: Pengcheng Xiong > Assignee: Pengcheng Xiong > Attachments: HIVE-12301.01.patch > > > The position in argList is mapped to a wrong column from RS operator -- This message was sent by Atlassian JIRA (v6.3.4#6332)