zhaodong created HIVE-29783:
-------------------------------
Summary: Iceberg CoW MERGE INTO missing column ambiguity check
Key: HIVE-29783
URL: https://issues.apache.org/jira/browse/HIVE-29783
Project: Hive
Issue Type: Bug
Environment: hadoop:3.3.6
hive:4.2.0
Reporter: zhaodong
Attachments: 111.png, 222.png, 333.png
Description:
When executing a MERGE INTO statement on an Iceberg Copy-on-Write (CoW) table,
if the ON condition contains an unqualified column name that exists in both the
target and source tables, Hive fails to throw a semantic ambiguity exception.
Instead, it executes the statement without any errors but returns incorrect
query results.
In contrast, the same statement executed on an Iceberg Merge-on-Read (MoR)
table correctly throws a SemanticException: Column a Found in more than One
Tables/Subqueries.
Steps to Reproduce:
1.Create an ORC source table and insert test data:
CREATE TABLE testorc (a INT, b INT) STORED AS ORC;
INSERT INTO testorc VALUES (9, 10);
2.Create Iceberg MoR and CoW target tables and insert identical test data:
CREATE TABLE testmor (a INT, b INT)
STORED BY iceberg
STORED AS ORC
TBLPROPERTIES (
'format-version' = '1',
'write.delete.mode' = 'merge-on-read',
'write.update.mode' = 'merge-on-read',
'write.merge.mode' = 'merge-on-read'
);
CREATE TABLE testcow (a INT, b INT)
STORED BY iceberg
STORED AS ORC
TBLPROPERTIES (
'format-version' = '1',
'write.delete.mode' = 'copy-on-write',
'write.update.mode' = 'copy-on-write',
'write.merge.mode' = 'copy-on-write'
);
INSERT INTO testmor VALUES (9, 3), (2, 3);
INSERT INTO testcow VALUES (9, 3), (2, 3);
3.Execute MERGE INTO with an unqualified column a in the ON condition:
-- Test 1: MoR table (Correct behavior)
MERGE INTO testmor
USING testorc
ON a = testorc.a
WHEN MATCHED THEN UPDATE SET b = testorc.b;
-- Expected: SemanticException: Column a Found in more than One
Tables/Subqueries
-- Test 2: CoW table (Buggy behavior)
MERGE INTO testcow
USING testorc
ON a = testorc.a
WHEN MATCHED THEN UPDATE SET b = testorc.b;
-- Actual: Executes successfully without throwing an exception.
4.Verify the data in the CoW table:
SELECT * FROM testcow;
Expected Behavior:
The CoW table should behave consistently with the MoR table and throw a
SemanticException indicating that column a is ambiguous.
Actual Behavior:
The CoW table execution succeeds silently, but the data is corrupted. The query
SELECT * FROM testcow; returns:
(9, 10), (2, 10)
The row (2, 3) was incorrectly updated to (2, 10) .
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
