[
https://issues.apache.org/jira/browse/FLINK-14886?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Leonard Xu updated FLINK-14886:
-------------------------------
Description:
In TPC-DS query6.sql， I found the query result is incorrect, I analyze the root
cause and simplify the issue as following:
{code:java}
create table t(
a varchar(100),
b float
);
insert into t(a,b) values('boy', 100);
insert into t(a,b) values(null, 50);
select t1.a, t1.b, (select avg(t2.b)
from t t2
where t2.a = t1.a)
from t t1;
{code}
the result when use blink planner is:
{code:java}
| a | b | EXP |
|-----|-----|-----|
| boy| 100| 100|
| null| 50| null|
{code}
but correct result should be :
{code:java}
| a | b | EXP |
|-----|-----|-----|
| boy| 100| null|
| null| 50| null|
{code}
Both flink legacy planner and other DB system can produce the correct result.
was:
In TPC-DS query6.sql， I found the query result is incorrect, I analyze the root
cause and simplify the issue as following:
{code:java}
create table t(
a varchar(100),
b float
);
insert into t(a,b) values('boy', 100);
insert into t(a,b) values(null, 50);
select t1.a, t1.b, (select avg(t2.b)
from t t2
where t2.a = t1.a)
from t t1;
{code}
the result when use blink planner is:
{code:java}
| a | b | EXP |
|-----|-----|-----|
| boy| 100| 100|
| null| 50| null|
{code}
but correct result should be :
{code:java}
| a | b | EXP |
|-----|-----|-----|
| boy| 100| null|
| null| 50| null|
{code}
Both flink legacy planner and other DB system can produce the correct result.
> Wrong result in scalar query using blink planner
> ------------------------------------------------
>
> Key: FLINK-14886
> URL: https://issues.apache.org/jira/browse/FLINK-14886
> Project: Flink
> Issue Type: Bug
> Components: Table SQL / Planner
> Affects Versions: 1.9.1
> Reporter: Leonard Xu
> Priority: Blocker
> Fix For: 1.10.0
>
>
>
> In TPC-DS query6.sql， I found the query result is incorrect, I analyze the
> root cause and simplify the issue as following:
> {code:java}
> create table t(
> a varchar(100),
> b float
> );
> insert into t(a,b) values('boy', 100);
> insert into t(a,b) values(null, 50);
> select t1.a, t1.b, (select avg(t2.b)
> from t t2
> where t2.a = t1.a)
> from t t1;
> {code}
> the result when use blink planner is:
> {code:java}
> | a | b | EXP |
> |-----|-----|-----|
> | boy| 100| 100|
> | null| 50| null|
> {code}
> but correct result should be :
> {code:java}
> | a | b | EXP |
> |-----|-----|-----|
> | boy| 100| null|
> | null| 50| null|
> {code}
> Both flink legacy planner and other DB system can produce the correct
> result.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
