fix: return 422 on invalid SQL (#34303)

tests/integration_tests/datasets/api_tests.py

@@ -974,6 +974,36 @@ class TestDatasetApi(SupersetTestCase):
         assert rv.status_code == 422
         assert data == {"message": "Dataset could not be created."}
 
+    @patch("superset.commands.dataset.create.security_manager.raise_for_access")
+    def test_create_dataset_with_invalid_sql_validation(self, mock_raise_for_access):
+        """
+        Dataset API: Test create dataset with invalid SQL during validation returns 422
+        """
+        from superset.exceptions import SupersetParseError
+
+        # Mock raise_for_access to throw SupersetParseError during validation
+        mock_raise_for_access.side_effect = SupersetParseError(
+            sql="SELECT FROM WHERE AND",
+            engine="postgresql",
+            message="Invalid SQL syntax",
+        )
+
+        self.login(ADMIN_USERNAME)
+        examples_db = get_example_database()
+        dataset_data = {
+            "database": examples_db.id,
+            "schema": "",
+            "table_name": "invalid_sql_table",
+            "sql": "SELECT FROM WHERE AND",
+        }
+        uri = "api/v1/dataset/"
+        rv = self.client.post(uri, json=dataset_data)
+        data = json.loads(rv.data.decode("utf-8"))
+        # The error is caught during validation and returns 422
+        assert rv.status_code == 422
+        assert "sql" in data["message"]
+        assert "Invalid SQL:" in data["message"]["sql"][0]
+
     def test_update_dataset_preserve_ownership(self):
         """
         Dataset API: Test update dataset preserves owner list (if un-changed)