Python模块单元测试的最佳文件结构组织？

Question

很遗憾，我发现Python中的单元测试有太多种写法，而且通常没有很好地记录下来。

我在寻找一种“终极”的结构，希望能满足以下大部分要求：

• 能够被测试框架发现，包括：
  • pytest
  • nosetests
  • tox
• 测试代码应该放在模块文件之外，并且放在与模块不同的目录中（方便维护），最好是在包的根目录下有一个tests/目录。
• 应该可以直接执行一个测试文件（测试代码需要知道它要测试的模块在哪里）。

请提供一个做假测试的示例测试文件，说明文件名和目录。

Answer 1

这是我一直在使用的方法：

目录结构

# All __init__.py files are empty in this example.
app
    package_a
        __init__.py
        module_a.py
    package_b
        __init__.py
        module_b.py
    test
        __init__.py
        test_app.py
    __init__.py
main.py

main.py

# This is the application's front-end.
#
# The import will succeed if Python can find the `app` package, which
# will occur if the parent directory of app/ is in sys.path, either 
# because the user is running the script from within that parect directory
# or because the user has included the parent directory in the PYTHONPATH
# environment variable.

from app.package_a.module_a import aaa
print aaa(123, 456)

module_a.py

# We can import a sibling module like this.
from app.package_b.module_b import bbb
def aaa(s, t):
    return '{0} {1}'.format(s, bbb(t))

# We can also run module_a.py directly, using Python's -m option, which
# allows you to run a module like a script.
#
#    python -m app.package_a.module_a
if __name__ == '__main__':
    print aaa(111, 222)
    print bbb(333)

module_b.py

def bbb(s):
    return s + 1

test_app.py

import unittest

# From the point of view of testing code, our working modules
# are siblings. Imports work accordingly, as seen in module_a.
from app.package_a.module_a import aaa
from app.package_a.module_a import bbb

class TestApp(unittest.TestCase):

    def test_aaa(self):
        self.assertEqual(aaa(77, 88), '77 89')

    def test_bbb(self):
        self.assertEqual(bbb(99), 100)

# Simiarly, we can run our test modules directly as scripts using the -m option,
# or using nose.
#
#    python -m app.test.test_app
#    nosetests app/test/test_app.py

if __name__ == '__main__':
    unittest.main()