[ 
https://issues.apache.org/jira/browse/CALCITE-4702?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17405485#comment-17405485
 ] 

Julian Hyde commented on CALCITE-4702:
--------------------------------------

I think that, for each dialect, we want to generate the simplest query that 
works in that dialect. This should not be the same for all dialects, because 
some, such as MSSQL, require you to generate a very complex query. Other 
dialects allow you group by constants.

Because of the issues raised CALCITE-4723, we should probably try not to 
generate "GROUP BY ()". Some dialects don't accept that syntax, and others 
return non-standard results.

> Error when executing query with GROUP BY constant via JDBC adapter
> ------------------------------------------------------------------
>
>                 Key: CALCITE-4702
>                 URL: https://issues.apache.org/jira/browse/CALCITE-4702
>             Project: Calcite
>          Issue Type: Bug
>          Components: core, jdbc-adapter
>    Affects Versions: 1.27.0
>            Reporter: Stamatis Zampetakis
>            Assignee: Soumyakanti Das
>            Priority: Major
>
> The following functionally equivalent SQL queries are accepted by Calcite and 
> produce a valid plan 
> {noformat}
> select avg(salary) from employee group by true
> select avg(salary) from employee group by 'a'
> {noformat}
> but they may fail if they are executed via the JDBC adapter since not all 
> DBMS allow grouping by constants expressions. Moreover, what works for one 
> may not work for the other. 
> +Examples+
> The {{GROUP BY TRUE}} query works in Postgres, and MySQL but fails in 
> Redshift with the following exception:
> {noformat}
> com.amazon.redshift.util.RedshiftException: ERROR: non-integer constant in 
> GROUP BY
> {noformat}
> The {{GROUP BY 'a'}} query works in MySQL but fails in Postgres with the 
> following exception:
> {noformat}
> ERROR:  non-integer constant in GROUP BY
> {noformat}
> +Edit:+
> The {{GROUP BY}} constant is similar to {{GROUP BY ()}} "nothing" but as 
> shown in the discussion below they are not equivalent. There is a nice [blog 
> post|https://blog.jooq.org/2018/05/25/how-to-group-by-nothing-in-sql/] 
> listing some limitations of various DBMS when it comes to {{GROUP BY ()}}. 



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Reply via email to