Fix: populate PostgresType.oid for array columns to enable binary COPY

src/include/postgres_utils.hpp

@@ -67,6 +67,7 @@ class PostgresUtils {
 	static string TypeToString(const LogicalType &input);
 	static string PostgresOidToName(uint32_t oid);
 	static uint32_t ToPostgresOid(const LogicalType &input);
+	static uint32_t TypeNameToPostgresOid(const string &type_name);
 	static bool SupportedPostgresOid(const LogicalType &input);
 	static LogicalType RemoveAlias(const LogicalType &type);
 	static PostgresType CreateEmptyPostgresType(const LogicalType &type);

src/postgres_utils.cpp

@@ -72,6 +72,45 @@ LogicalType PostgresUtils::RemoveAlias(const LogicalType &type) {
 	}
 }
 
+uint32_t PostgresUtils::TypeNameToPostgresOid(const string &type_name) {
+	if (type_name == "bool") {
+		return BOOLOID;
+	} else if (type_name == "int2") {
+		return INT2OID;
+	} else if (type_name == "int4") {
+		return INT4OID;
+	} else if (type_name == "int8") {
+		return INT8OID;
+	} else if (type_name == "float4") {
+		return FLOAT4OID;
+	} else if (type_name == "float8") {
+		return FLOAT8OID;
+	} else if (type_name == "varchar") {
+		return VARCHAROID;
+	} else if (type_name == "text") {
+		return TEXTOID;
+	} else if (type_name == "bytea") {
+		return BYTEAOID;
+	} else if (type_name == "date") {
+		return DATEOID;
+	} else if (type_name == "time") {
+		return TIMEOID;
+	} else if (type_name == "timestamp") {
+		return TIMESTAMPOID;
+	} else if (type_name == "timestamptz") {
+		return TIMESTAMPTZOID;
+	} else if (type_name == "interval") {
+		return INTERVALOID;
+	} else if (type_name == "timetz") {
+		return TIMETZOID;
+	} else if (type_name == "bit") {
+		return BITOID;
+	} else if (type_name == "uuid") {
+		return UUIDOID;
+	}
+	return 0;
+}
+
 LogicalType PostgresUtils::TypeToLogicalType(optional_ptr<PostgresTransaction> transaction,
                                              optional_ptr<PostgresSchemaEntry> schema,
                                              const PostgresTypeData &type_info, PostgresType &postgres_type) {
@@ -103,6 +142,10 @@ LogicalType PostgresUtils::TypeToLogicalType(optional_ptr<PostgresTransaction> t
 		child_type_info.type_modifier = type_info.type_modifier;
 		PostgresType child_pg_type;
 		auto child_type = PostgresUtils::TypeToLogicalType(transaction, schema, child_type_info, child_pg_type);
+		// populate the child OID from the actual Postgres type name
+		if (child_pg_type.oid == 0) {
+			child_pg_type.oid = TypeNameToPostgresOid(child_type_info.type_name);
+		}
 		// construct the child type based on the number of dimensions
 		for (idx_t i = 1; i < dimensions; i++) {
 			PostgresType new_pg_type;
@@ -275,9 +318,15 @@ PostgresType PostgresUtils::CreateEmptyPostgresType(const LogicalType &type) {
 			result.children.push_back(CreateEmptyPostgresType(child_type.second));
 		}
 		break;
-	case LogicalTypeId::LIST:
-		result.children.push_back(CreateEmptyPostgresType(ListType::GetChildType(type)));
+	case LogicalTypeId::LIST: {
+		auto child_pg_type = CreateEmptyPostgresType(ListType::GetChildType(type));
+		auto &child_type = ListType::GetChildType(type);
+		if (child_type.id() != LogicalTypeId::LIST && SupportedPostgresOid(child_type)) {
+			child_pg_type.oid = ToPostgresOid(child_type);
+		}
+		result.children.push_back(std::move(child_pg_type));
 		break;
+	}
 	default:
 		break;
 	}

test/sql/storage/attach_array_binary_copy.test

@@ -0,0 +1,117 @@
+# name: test/sql/storage/attach_array_binary_copy.test
+# description: Test that binary copy works for tables with array columns (issue #431)
+# group: [storage]
+
+require postgres_scanner
+
+require-env POSTGRES_TEST_DATABASE_AVAILABLE
+
+statement ok
+PRAGMA enable_verification
+
+statement ok
+ATTACH 'dbname=postgresscanner' AS s (TYPE POSTGRES)
+
+statement ok
+USE s;
+
+# Scenario 1: DDL-created table with varchar[] (CreateEmptyPostgresType path)
+# Before the fix, this silently fell back to TEXT copy because PostgresType.oid was never populated
+statement ok
+CREATE OR REPLACE TABLE array_binary_ddl(id INTEGER, name VARCHAR, tags VARCHAR[], scores INT[]);
+
+statement ok
+INSERT INTO array_binary_ddl VALUES (1, 'alice', ['rock', 'jazz'], [10, 20, 30]);
+
+statement ok
+INSERT INTO array_binary_ddl VALUES (2, 'bob', ['pop'], []);
+
+statement ok
+INSERT INTO array_binary_ddl VALUES (3, 'charlie', [], [42]);
+
+statement ok
+INSERT INTO array_binary_ddl VALUES (NULL, NULL, NULL, NULL);
+
+query IIII
+SELECT * FROM array_binary_ddl ORDER BY id
+----
+1	alice	[rock, jazz]	[10, 20, 30]
+2	bob	[pop]	[]
+3	charlie	[]	[42]
+NULL	NULL	NULL	NULL
+
+# Scenario 2: DDL-created table with other array types
+statement ok
+CREATE OR REPLACE TABLE array_binary_types(bools BOOLEAN[], dates DATE[], floats DOUBLE[], uuids UUID[]);
+
+statement ok
+INSERT INTO array_binary_types VALUES ([true, false, NULL], ['2024-01-01', '2024-12-31'], [1.5, 2.5], ['6d3d2541-710b-4bde-b3af-4711738636bf']);
+
+statement ok
+INSERT INTO array_binary_types VALUES ([], [], [], []);
+
+statement ok
+INSERT INTO array_binary_types VALUES (NULL, NULL, NULL, NULL);
+
+query IIII
+SELECT * FROM array_binary_types ORDER BY bools::VARCHAR
+----
+[]	[]	[]	[]
+[true, false, NULL]	[2024-01-01, 2024-12-31]	[1.5, 2.5]	[6d3d2541-710b-4bde-b3af-4711738636bf]
+NULL	NULL	NULL	NULL
+
+# Scenario 3: Existing table with varchar[] created via postgres_execute (TypeToLogicalType path)
+statement ok
+DROP TABLE IF EXISTS array_binary_catalog_varchar;
+
+statement ok
+CALL postgres_execute('s', 'CREATE TABLE array_binary_catalog_varchar(id INTEGER, tags VARCHAR[])');
+
+statement ok
+INSERT INTO array_binary_catalog_varchar VALUES (1, ['a', 'b', 'c']);
+
+statement ok
+INSERT INTO array_binary_catalog_varchar VALUES (2, []);
+
+statement ok
+INSERT INTO array_binary_catalog_varchar VALUES (3, [NULL, 'x']);
+
+query II
+SELECT * FROM array_binary_catalog_varchar ORDER BY id
+----
+1	[a, b, c]
+2	[]
+3	[NULL, x]
+
+# Scenario 4: Existing table with text[] created via postgres_execute
+# text[] should still work (falls back to TEXT copy because binary writer emits VARCHAROID but Postgres expects TEXTOID)
+statement ok
+DROP TABLE IF EXISTS array_binary_catalog_text;
+
+statement ok
+CALL postgres_execute('s', 'CREATE TABLE array_binary_catalog_text(id INTEGER, notes TEXT[])');
+
+statement ok
+INSERT INTO array_binary_catalog_text VALUES (1, ['hello', 'world']);
+
+statement ok
+INSERT INTO array_binary_catalog_text VALUES (2, []);
+
+query II
+SELECT * FROM array_binary_catalog_text ORDER BY id
+----
+1	[hello, world]
+2	[]
+
+# Cleanup
+statement ok
+DROP TABLE IF EXISTS array_binary_ddl;
+
+statement ok
+DROP TABLE IF EXISTS array_binary_types;
+
+statement ok
+DROP TABLE IF EXISTS array_binary_catalog_varchar;
+
+statement ok
+DROP TABLE IF EXISTS array_binary_catalog_text;