Details
-
Bug
-
Resolution: Fixed
-
Major
-
7.1.4, 7.2.0
-
Untriaged
-
0
-
Yes
Description
| Issue | Resolution |
| A Prepared statement might have resulted in an incorrect result in a multi-node environment. For example, a database with two query nodes. | Correlated subqueries from an encoded plan are now detected and marked. This ensures correct results are provided. |
Cluster with two query nodes.
Create bucket shellTest.
INSERT INTO shellTest (KEY,VALUE) VALUES("test11_withs", {"c11": 1, "c12": 10, "a11": [ 1, 2, 3, 4 ], "type": "left", "test_id": "withs"}), VALUES("test12_withs", {"c11": 2, "c12": 20, "a11": [ 3, 3, 5, 10 ], "type": "left", "test_id": "withs"}), VALUES("test13_withs", {"c11": 3, "c12": 30, "a11": [ 3, 4, 20, 40 ], "type": "left", "test_id": "withs"}), VALUES("test14_withs", {"c11": 4, "c12": 40, "a11": [ 30, 30, 30 ], "type": "left", "test_id": "withs"});
|
|
|
CREATE INDEX st_idx1 on shellTest(c11, c12) WHERE type = "left";
|
Then on the first query node:
prepare p1 as SELECT (WITH a AS (s.a11) SELECT RAW c1 FROM a AS c1) AS arr, s.test_id FROM shellTest s WHERE s.c11 IN [1, 2] AND type = "left";
|
|
|
execute p1;
|
This should produce:
"results": [
|
{
|
"arr": [
|
1,
|
2,
|
3,
|
4
|
],
|
"test_id": "withs"
|
},
|
{
|
"arr": [
|
3,
|
3,
|
5,
|
10
|
],
|
"test_id": "withs"
|
}
|
],
|
which is the correct result.
Now from the second query node:
execute p1;
|
This should produce the same result, however, it produces instead:
"results": [
|
{
|
"arr": [
|
1,
|
2,
|
3,
|
4
|
],
|
"test_id": "withs"
|
},
|
{
|
"arr": [
|
1,
|
2,
|
3,
|
4
|
],
|
"test_id": "withs"
|
}
|
],
|
Attachments
Issue Links
- is triggered by
-
MB-29710 Support expression term and subquery term as right-hand side of ANSI JOIN
-
- Closed
-