[
https://issues.apache.org/jira/browse/CALCITE-1016?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15051614#comment-15051614
]
Vladimir Sitnikov commented on CALCITE-1016:
--------------------------------------------
{quote}
7.9 <group by clause>
2) Case:
a) If there are no grouping columns, then the result of the <group by clause>
is the grouped
table consisting of T as its only group
{quote}
{quote}7.11 <query specification>
b) If T is a grouped table, then
Case:
i) If T has 0 (zero) groups, then the result of the <query specification> is
*an empty table*
{quote}
"I'm not a lawyer, but" it sounds like "as soon as you have a group_by_clause,
empty input results in empty output".
> What should "GROUP BY 1" return on an empty table?
> --------------------------------------------------
>
> Key: CALCITE-1016
> URL: https://issues.apache.org/jira/browse/CALCITE-1016
> Project: Calcite
> Issue Type: Bug
> Reporter: Julian Hyde
> Assignee: Julian Hyde
>
> What should "GROUP BY 1" return on an empty table? Calcite currently returns
> 0 rows. I am hearing claims that according to standard SQL it should return 1
> row.
> Here is what Oracle 11.2.0.2.0 does.
> {noformat}
> SQL> select count(*) from emp where 1 = 0;
> COUNT(*)
> ----------
> 0
> SQL> select count(*) from emp where 1 = 0 group by ();
> no rows selected
> SQL> select count(*) from emp where 1 = 0 group by 1;
> no rows selected
> SQL> select count(*) from emp;
> COUNT(*)
> ----------
> 14
> SQL> select count(*) from emp group by ();
> COUNT(*)
> ----------
> 14
> SQL> select count(*) from emp group by 1;
> COUNT(*)
> ----------
> 14
> {noformat}
> I had expected {code}select count(*) from emp where 1 = 0 group by (){code}
> would return 1 row, but it returns 0, like {code}group by 1{code}.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
