[
https://issues.apache.org/jira/browse/CALCITE-3181?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16880945#comment-16880945
]
Julian Hyde commented on CALCITE-3181:
--------------------------------------
It seems to make sense. Can you devise an example on a "real" data set, and
involving not just ROW_NUMBER but an aggregate function such as AVG or SUM.
I know it is not standard SQL, but can you sketch out how the query would look
if we allowed "OVER (PARTITION BY ... ORDER BY ... FETCH ...)"?
A while ago I was thinking about "partitioned sort-limit". We support
"Sort(Scan(EMP), key=[salary DESC], fetch=10)" (the 10 top-earning employees)
but we don't support "Sort(Scan(EMP), partition=[deptno], key=[salary DESC],
fetch=10)" (the top 10 top-earning employees in each department).
This came up in CALCITE-1317, the "WinMagic" optimization. Regular Sort with
limit can handle non-correlated sub-queries, but for correlated sub-queries we
needed partitioned Sort with limit. Thus it seems a natural and useful
extension to Sort. I haven't thought about whether the extension of Window that
you propose is as natural.
> 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
(v7.6.3#76005)
