多次自动执行Python代码与Pytest测试的实现方案及pytest.main()多次调用可行性咨询

Great question—let's break this down clearly, since the pytest docs warning can feel a bit vague at first.

你的当前代码：暂时可行，但有潜在隐患 #

First off: your example script works right now, and that's totally expected! The warning in the pytest docs is pointing out a edge case that might bite you later, not a hard "this will fail immediately" rule.

Here's why the warning exists: when you call pytest.main(), pytest imports your test files and any modules they depend on. Python caches imported modules, so if those files change between calls to pytest.main() (e.g., your some_action modifies a module's state, or you edit a test file mid-loop), the second call to pytest.main() will reuse the cached, outdated versions of those modules. Your example doesn't modify test files or their dependencies between runs, so you don't hit this issue yet.

But if you ever expand this workflow—say, adding dynamic updates to your code under test, or modifying test files during the loop—your tests will start giving incorrect results because they're running against stale code.

更优方案：用子进程运行pytest #

The cleanest, most reliable way to avoid the import caching problem is to run pytest in a separate process every time, instead of calling pytest.main() in the same process. This is exactly how you'd run pytest manually in your terminal, and it completely bypasses Python's module caching.

Here's how to modify your script to use subprocess.run() instead:

import subprocess

A = [0]
def some_action(x):
    x[0] += 1

if __name__ == '__main__':
    print('Initial value of A: {}'.format(A))
    for i in range(10):
        test_dir = "./tests/functional_tests/test_mock2.py" if i == 5 else "./tests/functional_tests/test_mock.py"
        
        some_action(A)
        
        # Run pytest as a separate process
        test_result = subprocess.run(
            ["pytest", "-q", "--tb=no", test_dir],
            capture_output=True,
            text=True
        )
        
        if test_result.returncode != 0:
            print(f'Interrupted at i={i} because of test failures')
            # Optional: print test error output for debugging
            print("Test errors:\n", test_result.stderr)
            break
        
        status_msg = 'All tests validated, final value of A: {}' if i > 5 else 'Final value of A: {}'
        print(status_msg.format(A))

为什么这个方法更好： #

• No import caching issues: Each pytest run is a fresh Python process, so it always picks up the latest version of your code and tests.
• More flexible: You can easily add pytest flags, redirect output, or integrate with other tools just like you would in the terminal.
• Easier debugging: Capturing stdout and stderr lets you see exactly why tests failed, which is harder with pytest.main() in some cases.

如果你真的想坚持用pytest.main()（不推荐） #

If for some reason you need to keep pytest in the same process, you'd have to manually reload all modules between runs using importlib.reload(). This is fragile, though—you'd need to track every module imported by your tests and reload them one by one, which gets complicated fast. It's almost never worth the effort compared to using subprocesses.

总结 #

Your current script works for your specific example, but it's not robust for long-term use. Switching to subprocesses is the safest, simplest way to follow pytest's recommendations and avoid subtle bugs down the line.

内容的提问来源于stack exchange，提问作者gdego