You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@drill.apache.org by "Vitalii Diravka (JIRA)" <ji...@apache.org> on 2018/07/11 15:14:00 UTC
[jira] [Commented] (DRILL-6417) Project is not pushed into scan if
use subquery with UNION operator
[ https://issues.apache.org/jira/browse/DRILL-6417?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16540226#comment-16540226 ]
Vitalii Diravka commented on DRILL-6417:
----------------------------------------
This is not an issue. See more in CALCITE-2410.
It is not allowed to transpose Project with a filtering SetOp operators, since it can lead to incorrect result.
USE CASE:
{code:java}
Table1
id | Name |
-------------
1 | John |
2 | Steaven |{code}
{code:java}
Table2
id | Name |
-------------
3 | John |
4 | Aman |{code}
{code:sql}
SELECT `Name` FROM (SELECT * FROM `Table1` UNION SELECT * FROM `Table2`)
{code}
The correct output is:
{code:SQL}
Name |
--------
John |
Steaven |
John |
Aman |
{code}
If the top Project will be applied to Scan operators it will lead to the wrong output:
{code:SQL}
Name |
--------
John |
Steaven |
Aman |
{code}
> Project is not pushed into scan if use subquery with UNION operator
> -------------------------------------------------------------------
>
> Key: DRILL-6417
> URL: https://issues.apache.org/jira/browse/DRILL-6417
> Project: Apache Drill
> Issue Type: Bug
> Affects Versions: 1.14.0
> Reporter: Anton Gozhiy
> Assignee: Vitalii Diravka
> Priority: Major
> Attachments: Test_data.tar.gz
>
>
> *Data:*
> Use attached dataset
> *Query:*
> {code:sql}
> explain plan for select id
> from
> (select id, part_col, int_col, bool_col, date_col, float_col, time_col, ts_col
> from dfs.tmp.`DRILL_3855_test_data`
> where part_col = 'Partition_one' or part_col = 'Partition_two'
> union
> select id, part_col, int_col, bool_col, date_col, float_col, time_col, ts_col
> from dfs.tmp.`DRILL_3855_test_data`
> where part_col = 'Partition_two' or part_col = 'Partition_three')
> where int_col = 0
> {code}
> *Expected plan:*
> {noformat}
> Scan ... columns=[`part_col`, `id`, `int_col`]
> {noformat}
> *Actual plan:*
> {noformat}
> Scan ... columns=[`part_col`, `id`, `int_col`, `bool_col`, `date_col`, `float_col`, `time_col`, `ts_col`]
> {noformat}
> *Notes:*
> Works as expected if change "union" to "union all"
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)