disallow table creation with out-of-range scale/width

src/pgduckdb_ddl.cpp

@@ -46,6 +46,45 @@ static bool ctas_skip_data = false;
 static bool top_level_ddl = true;
 static ProcessUtility_hook_type prev_process_utility_hook = NULL;
 
+// Get precision and scale from postgres type modifier.
+static int
+GetPgNumericPrecision(int32 typmod) {
+	return ((typmod - VARHDRSZ) >> 16) & 0xffff;
+}
+static int
+GetPgNumericScale(int32 typmod) {
+	return (((typmod - VARHDRSZ) & 0x7ff) ^ 1024) - 1024;
+}
+
+// If column is of type "numeric", check whether it's acceptable for duckdb;
+// If not, exception is thrown via `elog`.
+static void
+ValidateColumnNumericType(Form_pg_attribute &attribute) {
+	auto &type = attribute->atttypid;
+	if (type != NUMERICOID) {
+		return;
+	}
+	auto &typmod = attribute->atttypmod;
+	auto precision = GetPgNumericPrecision(typmod);
+	auto scale = GetPgNumericScale(typmod);
+	// duckdb's "numeric" type's max supported precision is 38.
+	if (typmod == -1 || precision < 0 || scale < 0 || precision > 38) {
+		elog(ERROR,
+		     "Unsupported type when creating column: type precision %d with scale %d is not supported by duckdb, which "
+		     "only allows maximum precision 38",
+		     precision, scale);
+	}
+}
+
+// Validate new relation creation, if invalid throw exception via `elog`.
+static void
+ValidateRelationCreation(TupleDesc desc) {
+	for (int i = 0; i < desc->natts; i++) {
+		Form_pg_attribute attr = &desc->attrs[i];
+		ValidateColumnNumericType(attr);
+	}
+}
+
 static void
 DuckdbHandleDDL(Node *parsetree) {
 	if (!pgduckdb::IsExtensionRegistered()) {
@@ -253,6 +292,10 @@ DECLARE_PG_FUNCTION(duckdb_create_table_trigger) {
 		elog(ERROR, "Expected single table to be created, but found %" PRIu64, static_cast<uint64_t>(SPI_processed));
 	}
 
+	// Check whether new table creation is supported in duckdb.
+	// TODO(hjiang): Same type validation should be applied to other DDL as well.
+	ValidateRelationCreation(SPI_tuptable->tupdesc);
+
 	HeapTuple tuple = SPI_tuptable->vals[0];
 	bool isnull;
 	Datum relid_datum = SPI_getbinval(tuple, SPI_tuptable->tupdesc, 1, &isnull);

src/pgduckdb_types.cpp

@@ -178,8 +178,7 @@ ConvertNumericDatum(const duckdb::Value &value) {
 		return ConvertDoubleDatum(value);
 	}
 	NumericVar numeric_var;
-	D_ASSERT(value_type_id == duckdb::LogicalTypeId::DECIMAL ||
-	         value_type_id == duckdb::LogicalTypeId::HUGEINT ||
+	D_ASSERT(value_type_id == duckdb::LogicalTypeId::DECIMAL || value_type_id == duckdb::LogicalTypeId::HUGEINT ||
 	         value_type_id == duckdb::LogicalTypeId::UBIGINT);
 	auto physical_type = value.type().InternalType();
 	const bool is_decimal = value_type_id == duckdb::LogicalTypeId::DECIMAL;
@@ -766,8 +765,7 @@ ConvertPostgresToDuckColumnType(Form_pg_attribute &attribute) {
 		auto scale = numeric_typmod_scale(typmod);
 		// duckdb's "numeric" type's max supported precision is 38.
 		if (typmod == -1 || precision < 0 || scale < 0 || precision > 38) {
-			const auto user_type_info = duckdb::StringUtil::Format(
-				"type precision %d with scale %d", precision, scale);
+			const auto user_type_info = duckdb::StringUtil::Format("type precision %d with scale %d", precision, scale);
 			base_type = duckdb::LogicalType::USER(user_type_info);
 		} else {
 			base_type = duckdb::LogicalType::DECIMAL(precision, scale);

test/regression/expected/array_type_support.out

@@ -199,7 +199,7 @@ SELECT * FROM float8_array_1d;
 (4 rows)
 
 -- NUMERIC (single dimension)
-CREATE TABLE numeric_array_1d(a NUMERIC[]);
+CREATE TABLE numeric_array_1d(a NUMERIC(2, 1)[]);
 INSERT INTO numeric_array_1d SELECT CAST(a as NUMERIC[]) FROM (VALUES
     ('{1.1, 2.2, 3.3}'),
     (NULL),
@@ -373,7 +373,7 @@ SELECT * FROM float8_array_2d;
 (5 rows)
 
 -- NUMERIC (two dimensions)
-CREATE TABLE numeric_array_2d(a NUMERIC[][]);
+CREATE TABLE numeric_array_2d(a NUMERIC(3, 1)[][]);
 INSERT INTO numeric_array_2d VALUES
     ('{{1.1,2.2},{3.3,4.4}}'),
     ('{{5.5,6.6,7.7},{8.8,9.9,10.1}}'),

test/regression/expected/type_support.out

@@ -201,21 +201,6 @@ SELECT * FROM float8_tbl;
  4.582345020342342e+20
 (3 rows)
 
--- NUMERIC as DOUBLE
-CREATE TABLE numeric_as_double(a NUMERIC);
-INSERT INTO numeric_as_double SELECT a FROM (VALUES
-    (0.234234234),
-    (NULL),
-    (458234502034234234234.000012)
-) t(a);
-SELECT * FROM numeric_as_double;
-           a           
------------------------
-           0.234234234
-
- 4.582345020342342e+20
-(3 rows)
-
 -- NUMERIC with a physical type of SMALLINT
 CREATE TABLE smallint_numeric(a NUMERIC(4, 2));
 INSERT INTO smallint_numeric SELECT a FROM (VALUES
@@ -330,7 +315,6 @@ DROP TABLE timestamp_tbl;
 DROP TABLE timestamptz_tbl;
 DROP TABLE float4_tbl;
 DROP TABLE float8_tbl;
-DROP TABLE numeric_as_double;
 DROP TABLE smallint_numeric;
 DROP TABLE integer_numeric;
 DROP TABLE bigint_numeric;

test/regression/sql/array_type_support.sql

@@ -125,7 +125,7 @@ INSERT INTO float8_array_1d SELECT CAST(a as FLOAT8[]) FROM (VALUES
 SELECT * FROM float8_array_1d;
 
 -- NUMERIC (single dimension)
-CREATE TABLE numeric_array_1d(a NUMERIC[]);
+CREATE TABLE numeric_array_1d(a NUMERIC(2, 1)[]);
 INSERT INTO numeric_array_1d SELECT CAST(a as NUMERIC[]) FROM (VALUES
     ('{1.1, 2.2, 3.3}'),
     (NULL),
@@ -226,7 +226,7 @@ INSERT INTO float8_array_2d VALUES
 SELECT * FROM float8_array_2d;
 
 -- NUMERIC (two dimensions)
-CREATE TABLE numeric_array_2d(a NUMERIC[][]);
+CREATE TABLE numeric_array_2d(a NUMERIC(3, 1)[][]);
 INSERT INTO numeric_array_2d VALUES
     ('{{1.1,2.2},{3.3,4.4}}'),
     ('{{5.5,6.6,7.7},{8.8,9.9,10.1}}'),

test/regression/sql/type_support.sql

@@ -90,15 +90,6 @@ INSERT INTO float8_tbl SELECT CAST(a AS FLOAT8) FROM (VALUES
 ) t(a);
 SELECT * FROM float8_tbl;
 
--- NUMERIC as DOUBLE
-CREATE TABLE numeric_as_double(a NUMERIC);
-INSERT INTO numeric_as_double SELECT a FROM (VALUES
-    (0.234234234),
-    (NULL),
-    (458234502034234234234.000012)
-) t(a);
-SELECT * FROM numeric_as_double;
-
 -- NUMERIC with a physical type of SMALLINT
 CREATE TABLE smallint_numeric(a NUMERIC(4, 2));
 INSERT INTO smallint_numeric SELECT a FROM (VALUES
@@ -171,7 +162,6 @@ DROP TABLE timestamp_tbl;
 DROP TABLE timestamptz_tbl;
 DROP TABLE float4_tbl;
 DROP TABLE float8_tbl;
-DROP TABLE numeric_as_double;
 DROP TABLE smallint_numeric;
 DROP TABLE integer_numeric;
 DROP TABLE bigint_numeric;