[
https://issues.apache.org/jira/browse/CALCITE-695?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14510176#comment-14510176
]
Julian Hyde edited comment on CALCITE-695 at 4/24/15 12:55 AM:
---------------------------------------------------------------
I'd say it's a bug, because we are generating a more complex, and perhaps less
efficient, expression than we need.
It could be fixed (a) by recognizing the pattern at sql-to-rel time, of (b)
using a planner rule. For (b) it is not safe to rely on estimated row count,
but you could rely on RelMetadataQuery.areColumnsUnique or
RelMetadataQuery.getUniqueKeys for an empty list of columns.
was (Author: julianhyde):
I'd say it's a bug, because we are generating a more complex, and perhaps less
efficient, expression than we need.
It could be fixed (a) by recognizing the pattern at sql-to-rel time, of (b)
using a planner rule. For (b) it is not safe to rely on estimated row count,
but you could rely on RelMdDistinctRowCount or RelMdUniqueKeys.
> SqlSingleValueAggFunction is created when it may not be needed
> --------------------------------------------------------------
>
> Key: CALCITE-695
> URL: https://issues.apache.org/jira/browse/CALCITE-695
> Project: Calcite
> Issue Type: Bug
> Affects Versions: 1.3.0-incubating
> Reporter: Aman Sinha
> Assignee: Julian Hyde
>
> In the following query, the subquery is producing a scalar value but the
> logical plan still creates a SqlSingleValueAggFunction.
> {code}
> select r_regionkey from region
> where r_regionkey > (select min(n_regionkey) * 2 from nation);
> {code}
> If the aggregate is just min(n_regionkey) instead of the expression
> min(n_regionkey) * 2 then no SqlSingleValueAggFunction is created. Ideally,
> both should behave the same.
> This is not necessarily a bug but it does create a burden on the underlying
> execution engine to support this function even in cases where it may not be
> needed.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
