zzwqqq created CALCITE-7885:
-------------------------------
Summary: RelDecorrelator merges distinct correlation fields and
produces extra rows
Key: CALCITE-7885
URL: https://issues.apache.org/jira/browse/CALCITE-7885
Project: Calcite
Issue Type: Bug
Components: core
Reporter: zzwqqq
Assignee: zzwqqq
{{rewriteScalarAggregate}} uses aggregate column positions as keys in
{{{}newProjectMap{}}}. Different outer references can share a position, so one
overwrites another. When their values differ on unmatched rows, the final join
can attach an aggregate result to the wrong outer row.
For example:
{code:sql}
with t(x, y) as (values (1, 1), (2, 1)),
s(z) as (values (1))
select t.x, t.y,
(select count(*) from s where s.z = t.x and s.z = t.y) as c
from t;
{code}
Expected:
{code:sql}
+---+---+---+
| X | Y | C |
+---+---+---+
| 1 | 1 | 1 |
| 2 | 1 | 0 |
+---+---+---+
(2 rows)
{code}
Actual:
{code:sql}
+---+---+---+
| X | Y | C |
+---+---+---+
| 1 | 1 | 1 |
| 2 | 1 | 0 |
| 1 | 1 | 0 |
+---+---+---+
(3 rows)
{code}
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
