[
https://issues.apache.org/jira/browse/CALCITE-3181?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16885545#comment-16885545
]
Julian Hyde commented on CALCITE-3181:
--------------------------------------
The {{enum RelFieldCollation}} has values ASCENDING, STRICTLY_ASCENDING,
DESCENDING, STRICTLY_DESCENDING and CLUSTERED. CLUSTERED exists for precisely
this purpose.
We have no way of saying in a physical property (trait) "for a given deptno key
there are no more than 10 rows". My instinct says we probably shouldn't add a
way; it would make the trait more complicated for everyone else. So we'd do
this by transformation (forward chaining) not by traits (backward chaining).
> Support limit per group in Window
> ---------------------------------
>
> Key: CALCITE-3181
> URL: https://issues.apache.org/jira/browse/CALCITE-3181
> Project: Calcite
> Issue Type: Improvement
> Components: core
> Reporter: Haisheng Yuan
> Priority: Major
>
> We have a lot of queries like the following to retrieve top N tuples per
> group:
> {code:java}
> SELECT x, y FROM
> (SELECT x, y, ROW_NUMBER() OVER (PARTITION BY x ORDER BY y)
> AS rn FROM t1) t2 WHERE rn <= 3;
> {code}
> The performance is not good if each group has a lot more tuples than wanted,
> because we will retrieve and sort all the tuples, instead of just doing a
> top-N heap sort.
> In order to do optimization for this kind of query, we need to extend window
> to support limit, if and only if there is only 1 window function, and it is
> {{row_number()}}. We also need a substitute rule to push the limit into
> window. Of course, we also need to modify executor to support this
> optimization (can be later).
> {code:java}
> Filter (rn <= 3)
> +- Window (window#0={Partition by x order by y ROW_NUMBER()})
> {code}
> to
> {code:java}
> Filter (rn <= 3)
> +- Window (window#0={Partition by x order by y limit 3 ROW_NUMBER()})
> {code}
> Thoughts? Objections?
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
