diff --git a/backend/tests/test_eval_gold.py b/backend/tests/test_eval_gold.py index 02a736f7..7366587b 100644 --- a/backend/tests/test_eval_gold.py +++ b/backend/tests/test_eval_gold.py @@ -133,17 +133,16 @@ async def test_the_read_only_guard_refuses_writes_before_anything_runs(source): assert db.execute("SELECT COUNT(*) FROM orders").fetchone()[0] == len(shop.ROWS) -@pytest.mark.parametrize("sql, refused", [ - ("SELECT status, COUNT(*) AS n FROM orders GROUP BY status", True), - ("SELECT status AS 状态 FROM orders", True), # 改了名字也认得出 - ("SELECT * FROM orders", True), - ("SELECT COUNT(*) AS n FROM orders WHERE status = 'paid'", False), # 结果里没有取自遮罩列的格 +@pytest.mark.parametrize("sql, expectation, refused", [ + ("SELECT status, COUNT(*) AS n FROM orders GROUP BY status", {"kind": "set", "column": "status"}, True), + ("SELECT status AS 状态 FROM orders", {"kind": "set", "column": "状态"}, True), # 改了名字也认得出 + ("SELECT * FROM orders", {"kind": "table", "keys": ["id"]}, True), + # 结果里没有取自遮罩列的格:条件里用到遮罩列不算 + ("SELECT COUNT(*) AS n FROM orders WHERE status = 'paid'", {"kind": "scalar", "column": "n"}, False), ]) -async def test_gold_that_uses_a_masked_column_is_refused(tmp_path, sql, refused): +async def test_gold_that_uses_a_masked_column_is_refused(tmp_path, sql, expectation, refused): name, sid = await shop.add_source(shop.build(tmp_path / "masked.db"), options={"mask_columns": ["status"]}) try: - expectation = {"kind": "scalar", "column": "n"} if not refused else {"kind": "table", "keys": ["id"]} \ - if sql.startswith("SELECT *") else {"kind": "set", "column": "状态" if "状态" in sql else "status"} outcome = await gold.compute_gold(sql_case(sid, sql, expectation), ctx()) if refused: assert outcome.status == "refused" and outcome.code == gold.GOLD_MASKED, outcome.reason