如何在Pytest类作用域Fixture的Teardown阶段执行断言？

我有一个以类形式组织的参数化测试套件，包含多个测试用例和类作用域fixture。我希望在某个fixture实例对应的所有测试完成后运行一些断言操作，但尝试在fixture的销毁阶段添加assert或pytest.fail语句时，Pytest似乎不支持这种用法。

此外，该fixture执行的操作开销很大，不能为每个测试用例重新创建实例。我可以接受两种失败报告方式：

1. 将失败显示为一个从fixture名称推导而来的独立“虚拟”测试用例
2. 追溯标记所有依赖该fixture的测试用例为失败

示例代码如下：

import pytest

class Car:
    def __init__(self, car_model: str):
        pass

class TestCar:
    @pytest.fixture(scope='class', params=["Ford", "Jeep"])
    def car(self, request):
        car = Car(car_model=request.param)
        yield car

        # 这里希望在依赖该fixture实例的所有测试完成后执行清理检查，但Pytest不支持
        part_test_ratio = car.used_parts / car.total_parts
#        assert part_test_ratio > 0.9, "至少90%的部件必须被测试！"

    def test_ignition(self, car):
        pass

    def test_steering(self, car):
        pass

    def test_brakes(self, car):
        pass

正常测试结果：

PASSED car_test.py::TestCar::test_ignition[Ford]
PASSED car_test.py::TestCar::test_steering[Ford]
PASSED car_test.py::TestCar::test_brakes[Ford]
PASSED car_test.py::TestCar::test_ignition[Jeep]
PASSED car_test.py::TestCar::test_steering[Jeep]
PASSED car_test.py::TestCar::test_brakes[Jeep]

期望的选项1结果：

...
PASSED car_test.py::TestCar::test_brakes[Jeep]
FAILED car_test.py::TestCar::car[Jeep]

期望的选项2结果：

...
FAILED car_test.py::TestCar::test_ignition[Jeep]
FAILED car_test.py::TestCar::test_steering[Jeep]
FAILED car_test.py::TestCar::test_brakes[Jeep]

选项1：生成独立的虚拟测试用例报告 #

通过自定义pytest_runtest_makereport钩子结合fixture，在类测试完成后添加虚拟测试结果，专门用于验证fixture的后置断言。

修改后的代码示例：

import pytest

class Car:
    def __init__(self, car_model: str):
        self.car_model = car_model
        self.used_parts = 8  # 模拟Jeep测试率不足，Ford可设为10
        self.total_parts = 10

class TestCar:
    @pytest.fixture(scope='class', params=["Ford", "Jeep"])
    def car(self, request):
        car = Car(car_model=request.param)
        # 保存断言结果到类属性，供钩子使用
        request.cls.car_assert_result = None
        yield car
        # 执行后置断言并捕获错误
        try:
            part_test_ratio = car.used_parts / car.total_parts
            assert part_test_ratio > 0.9, f"{car.car_model} 仅测试了{part_test_ratio*100:.1f}%的部件，未达到90%要求！"
        except AssertionError as e:
            request.cls.car_assert_result = str(e)

    def test_ignition(self, car):
        pass

    def test_steering(self, car):
        pass

    def test_brakes(self, car):
        pass

@pytest.hookimpl(tryfirst=True, hookwrapper=True)
def pytest_runtest_makereport(item, call):
    outcome = yield
    report = outcome.get_result()
    # 仅处理类测试的销毁阶段，且存在断言错误时
    if report.when == 'teardown' and hasattr(item.cls, 'car_assert_result') and item.cls.car_assert_result:
        from _pytest.runner import TestReport
        fixture_name = 'car'
        param = item.funcargs.get(fixture_name).car_model
        # 构造虚拟测试用例的节点ID
        nodeid = f"{item.nodeid.split('::test_')[0]}::{fixture_name}[{param}]"
        # 生成失败报告
        failure_report = TestReport(
            nodeid=nodeid,
            location=item.location,
            keywords=item.keywords,
            outcome='failed',
            longrepr=item.cls.car_assert_result,
            when='teardown',
            user_properties=[]
        )
        # 将报告添加到测试结果中
        item.session.add_report(failure_report)
        # 清理断言结果，避免影响下一个参数化实例
        del item.cls.car_assert_result

运行后，Jeep对应的fixture断言失败时，会额外生成FAILED car_test.py::TestCar::car[Jeep]的独立结果，符合选项1需求。

选项2：标记所有依赖fixture的测试用例为失败 #

利用request.addfinalizer方法捕获断言失败，在fixture销毁时遍历类内所有测试用例，将对应参数化实例的测试结果标记为失败。

修改后的代码示例：

import pytest

class Car:
    def __init__(self, car_model: str):
        self.car_model = car_model
        self.used_parts = 8  # 模拟Jeep测试率不足
        self.total_parts = 10

class TestCar:
    @pytest.fixture(scope='class', params=["Ford", "Jeep"])
    def car(self, request):
        car = Car(car_model=request.param)
        assert_error = None

        # 定义终结器，执行后置断言
        def finalize():
            nonlocal assert_error
            try:
                part_test_ratio = car.used_parts / car.total_parts
                assert part_test_ratio > 0.9, f"{car.car_model} 仅测试了{part_test_ratio*100:.1f}%的部件，未达到90%要求！"
            except AssertionError as e:
                assert_error = e

        request.addfinalizer(finalize)
        yield car

        # 若存在断言错误，修改对应测试用例的结果
        if assert_error:
            test_prefix = f"{request.node.nodeid}::"
            param_suffix = f"[{car.car_model}]"
            # 遍历会话中的所有测试报告
            for report in request.session.items:
                # 匹配当前参数化实例的测试用例
                if test_prefix in report.nodeid and param_suffix in report.nodeid:
                    report.outcome = 'failed'
                    report.longrepr = str(assert_error)

    def test_ignition(self, car):
        pass

    def test_steering(self, car):
        pass

    def test_brakes(self, car):
        pass

这种方式会将Jeep对应的三个测试用例全部标记为失败，符合选项2需求。

内容的提问来源于stack exchange，提问作者vvv444