[
https://issues.apache.org/jira/browse/KYLIN-2841?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16720322#comment-16720322
]
XiaoXiang Yu edited comment on KYLIN-2841 at 12/13/18 4:46 PM:
---------------------------------------------------------------
Same cube, same SQL, only difference is the limit number. It should not cause
the lost of return result rows. When set LIMIT to 10000, result has 8 rows;
when set LIMIT to 10, result has 4 rows with each sum smaller than last result.
After that I using hive to execute same sql query. By comparing result with
hive , it is clear that when set LIMIT to 10, we got a error result.
I upload two log file which show details about two query with "calicite.debug"
is enable. These files contains whole query analyse process.
* {color:#FF0000}*+kylin-2841-1.txt : I set LIMIT to 10, and got wrong
result.+*{color}
* {color:#FF0000}+*kylin-2841-2.txt : I set LIMIT to 10000, and got correct
result.*+{color}
!image-2018-12-14-00-08-27-648.png!
After that, I apply contributor's patch, {color:#FF0000}+*and it does fix that
issue.*+{color} You can see it in the following or check my log file
(kylin-2841-apply-patch.txt).
!image-2018-12-14-00-23-08-869.png!
was (Author: hit_lacus):
Same cube, same SQL, only difference is the limit number. It should not cause
the lost of return result rows. When set LIMIT to 10000, result has 8 rows;
when set LIMIT to 10, result has 4 rows with each sum smaller than last result.
After that I using hive to execute same sql query. By comparing result with
hive , it is clear that when set LIMIT to 10, we got a error result.
I upload two log file which show details about two query with "calicite.debug"
is enable.
!image-2018-12-14-00-08-27-648.png!
After that, I apply contributor's patch, and it does fix that issue. You can
see it in the following or check my log file (kylin-2841-apply-patch.txt).
!image-2018-12-14-00-23-08-869.png!
> LIMIT is buggy with subquery
> ----------------------------
>
> Key: KYLIN-2841
> URL: https://issues.apache.org/jira/browse/KYLIN-2841
> Project: Kylin
> Issue Type: Bug
> Affects Versions: v2.1.0
> Reporter: Mu Kong
> Assignee: zhengdong
> Priority: Major
> Labels: scope
> Fix For: v2.6.0
>
> Attachments: 0001-KYLIN-2841-LIMIT-is-buggy-with-subquery.patch,
> image-2018-12-13-23-41-53-732.png, image-2018-12-13-23-45-06-271.png,
> image-2018-12-14-00-05-19-114.png, image-2018-12-14-00-08-27-648.png,
> image-2018-12-14-00-23-08-869.png, kylin-2841-1.txt, kylin-2841-2.txt,
> kylin-2841-apply-patch.txt
>
>
> Hi, all.
> I found that limit in the web UI seems not behaving as expected.
> When I run a query like the follows:
> {code:sql}
> SELECT
> SUM(col3) AS col4,
> SUM(col5) AS total_col5,
> col1
> FROM
> (
> SELECT
> col1,
> col2,
> MAX(col3) AS col3,
> COUNT(*) AS col5
> FROM db.table
> WHERE col6 = 'somestring'
> GROUP BY col1, col2
> )
> GROUP BY col1
> {code}
> When I specify the limit as 50, the result has 19 records, and when I specify
> the limit as 500000, there are 90+ records in the result and each record has
> higher col4 and total_col5.
> But for query that doesn't have subquery, the result remains the same no
> matter how I change the limit.
> I guess for the query with subquery, limit somehow limits the number of the
> result from the inner query instead of the result of the outer query.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
