[jira] [Commented] (CALCITE-370) Support GROUPING SETS in SQL and algebra

["Vladimir Sitnikov (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/CALCITE-370?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14204883#comment-14204883 ] 

Vladimir Sitnikov commented on CALCITE-370:
-------------------------------------------

What is the purpose of {{GroupingType}}?

How are you going to support {{GROUPING SETS ((X, Y), (Z), (Y), ())}}?
What would be the value of {{groupingType}}?

What is the preferred implementation approach?
"implementations that cannot implement grouping sets throw HaveNoIdeaHowToImplement" and "some Calcite rule rewrites grouping sets to a set of unions and retries"? (e.g. something similar to {{avg(..)}})

> Support GROUPING SETS in SQL and algebra
> ----------------------------------------
>
>                 Key: CALCITE-370
>                 URL: https://issues.apache.org/jira/browse/CALCITE-370
>             Project: Calcite
>          Issue Type: Bug
>            Reporter: Julian Hyde
>            Assignee: Julian Hyde
>
> Support grouping sets in the relational algebra. This feature would add a new field, groupKeys, to aggregate rel that has a set of bit maps, each of which has a grouping key.
> For example, if 'group by x, y, z' is represented by
> {code}
> AggregateRel(groupSet={0, 1, 2})
> {code}
> then 'group by grouping sets (x, y, z), (y, z), ()' would be represented by
> {code}
> AggregateRel(groupSet={0, 1, 2}, groupSets=[ {0, 1, 2}, {1, 2}, {} ])
> {code}
> A regular group by, or a group by with a singleton grouping set, would have groupSets = null.
> If groupSets is present, the returned row type would include an indicator column for each grouping key.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)