[ https://issues.apache.org/jira/browse/HIVE-23598?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17124966#comment-17124966 ]
Zoltan Haindrich commented on HIVE-23598: ----------------------------------------- I think this could be rewritten to be use 'cume_dist' and let the other rewrite do the heavy lifting... {code} select id,ntile(4) over (order by id),CEIL(cume_dist() over (order by id)*4) from sketch_input order by id; {code} the NTILE udf has an easy advantage: it may return different values for the same input value. {code} select id,ntile(2) over (order by id) from TABLE(VALUES(1),(1),(1),(1)) AS t(id); +-----+-----------------+ | id | ntile_window_0 | +-----+-----------------+ | 1 | 1 | | 1 | 1 | | 1 | 2 | | 1 | 2 | +-----+-----------------+ {code} > Add option to rewrite NTILE to sketch functions > ----------------------------------------------- > > Key: HIVE-23598 > URL: https://issues.apache.org/jira/browse/HIVE-23598 > Project: Hive > Issue Type: Sub-task > Reporter: Zoltan Haindrich > Assignee: Zoltan Haindrich > Priority: Major > -- This message was sent by Atlassian Jira (v8.3.4#803005)