[
https://issues.apache.org/jira/browse/CALCITE-6214?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17809646#comment-17809646
]
Julian Hyde commented on CALCITE-6214:
--------------------------------------
I'm not sure what you mean by 'merge aggregates'. If you try to create your
first plan
{noformat}
LogicalAggregate(group=[{0}], SDS=[SUM(DISTINCT $1)], SS=[SUM($1)])
LogicalAggregate(group=[{0, 1}]) -- We can optimize it.
LogicalProject(DEPTNO=[$7], SAL=[$5])
LogicalTableScan(table=[[CATALOG, SALES, EMP]])
{noformat}
using {{RelBuilder}}, it should remove the {{DISTINCT}} keyword and share the
aggregate functions, so you get the following, which I believe is optimal:
{noformat}
LogicalProject(DEPTNO=[$0], SDS=[$1], SS=[$1])
LogicalAggregate(group=[{0}], SS=[SUM($1)])
LogicalAggregate(group=[{0, 1}])
LogicalProject(DEPTNO=[$7], SAL=[$5])
LogicalTableScan(table=[[CATALOG, SALES, EMP]])
{noformat}
> Remove DISTINCT in aggregate function if field is unique
> --------------------------------------------------------
>
> Key: CALCITE-6214
> URL: https://issues.apache.org/jira/browse/CALCITE-6214
> Project: Calcite
> Issue Type: Improvement
> Components: core
> Reporter: Jiajun Xie
> Assignee: Jiajun Xie
> Priority: Minor
> Labels: pull-request-available
>
> For the sql
> {code:java}
> select count(distinct x) cnt
> from(
> select distinct sal x from emp
> ) t {code}
> The distinct that in count can be removed.
> {code:java}
> LogicalAggregate(group=[{}], CNT=[COUNT($0)])
> LogicalAggregate(group=[{0}])
> LogicalProject(X=[$5])
> LogicalTableScan(table=[[CATALOG, SALES, EMP]]) {code}
> But `CoreRules#AGGREGATE_REMOVE` not support it, so there are two DISTINCT.
> {code:java}
> LogicalAggregate(group=[{}], CNT=[COUNT(DISTINCT $0)])
> LogicalAggregate(group=[{0}])
> LogicalProject(X=[$5])
> LogicalTableScan(table=[[CATALOG, SALES, EMP]]) {code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
