From 7c8a366ba2cf18e7ab8cb273e3716cb94b3c6376 Mon Sep 17 00:00:00 2001
From: paragjain <paragjain@microsoft.com>
Date: Fri, 14 Jun 2024 04:39:17 +0000
Subject: [PATCH] some more

---
 src/test/regress/expected/merge.out | 21 +++++++++++++++++++--
 src/test/regress/multi_schedule     |  3 ++-
 src/test/regress/sql/merge.sql      | 14 ++++++++++++++
 3 files changed, 35 insertions(+), 3 deletions(-)

diff --git a/src/test/regress/expected/merge.out b/src/test/regress/expected/merge.out
index 188ed897e..1a7e58f36 100644
--- a/src/test/regress/expected/merge.out
+++ b/src/test/regress/expected/merge.out
@@ -1965,8 +1965,8 @@ on t.id = s.somekey
 WHEN NOT MATCHED THEN
   INSERT (id)
   VALUES (s.somekey);
-DEBUG:  MERGE INSERT must use the source table distribution column value, try repartitioning
-DEBUG:  MERGE INSERT must use the source table distribution column value
+DEBUG:  MERGE INSERT must use the source table distribution column value for push down to workers. Otherwise, repartitioning will be applied
+DEBUG:  MERGE INSERT must use the source table distribution column value for push down to workers. Otherwise, repartitioning will be applied
 DEBUG:  Creating MERGE repartition plan
 DEBUG:  Using column - index:0 from the source list to redistribute
                                                     QUERY PLAN
@@ -2118,6 +2118,23 @@ SELECT * FROM target_table;
 ---------------------------------------------------------------------
 (0 rows)
 
+--
+DELETE FROM source_withdata;
+DELETE FROM target_table;
+INSERT INTO source VALUES (1,1);
+merge into target_table sda
+using source_withdata sdn
+on sda.id = sdn.id AND sda.id = 1
+when not matched then
+  insert (id)
+  values (10000);
+ERROR:  MERGE INSERT is using unsupported expression type for distribution column
+DETAIL:  Inserting arbitrary values that don't correspond to the joined column values can lead to unpredictable outcomes where rows are incorrectly distributed among different shards
+SELECT * FROM target_table WHERE id = 10000;
+ id | name
+---------------------------------------------------------------------
+(0 rows)
+
 RESET client_min_messages;
 -- This will prune shards with restriction information as NOT MATCHED is void
 BEGIN;
diff --git a/src/test/regress/multi_schedule b/src/test/regress/multi_schedule
index fca36f5ab..7f0c7ca57 100644
--- a/src/test/regress/multi_schedule
+++ b/src/test/regress/multi_schedule
@@ -116,7 +116,8 @@ test: function_with_case_when
 test: clock
 
 # MERGE tests
-test: merge pgmerge merge_repartition2
+test: merge pgmerge
+test: merge_repartition2
 test: merge_repartition1 merge_schema_sharding
 test: merge_partition_tables
 
diff --git a/src/test/regress/sql/merge.sql b/src/test/regress/sql/merge.sql
index cc7f0d9a6..4538d8d81 100644
--- a/src/test/regress/sql/merge.sql
+++ b/src/test/regress/sql/merge.sql
@@ -1308,6 +1308,20 @@ WHEN MATCHED THEN
 -- let's verify if data deleted properly.
 SELECT * FROM target_table;
 
+--
+DELETE FROM source_withdata;
+DELETE FROM target_table;
+INSERT INTO source VALUES (1,1);
+
+merge into target_table sda
+using source_withdata sdn
+on sda.id = sdn.id AND sda.id = 1
+when not matched then
+  insert (id)
+  values (10000);
+
+SELECT * FROM target_table WHERE id = 10000;
+
 RESET client_min_messages;