使用 @pytest.fixture(scope="module") 和 @pytest.mark.asyncio

Using @pytest.fixture(scope="module") with @pytest.mark.asyncio

我认为下面的示例是一个非常常见的用例:

  1. 创建到数据库的连接一次
  2. 传递这个连接来测试哪个插入数据
  3. 将连接传递给验证数据的测试。

更改 @pytest.fixture(scope="module") 的范围会导致 ScopeMismatch: You tried to access the 'function' scoped fixture 'event_loop' with a 'module' scoped request object, involved factories

此外,test_inserttest_find 协程不需要 event_loop 参数,因为通过传递连接已经可以访问循环。

有解决这两个问题的想法吗?

import pytest

@pytest.fixture(scope="function")  # <-- want this to be scope="module"; run once!
@pytest.mark.asyncio
async def connection(event_loop):
    """ Expensive function; want to do in the module scope. Only this function needs `event_loop`!
    """
    conn await = make_connection(event_loop)
    return conn


@pytest.mark.dependency()
@pytest.mark.asyncio
async def test_insert(connection, event_loop):  # <-- does not need event_loop arg
    """ Test insert into database.

        NB does not need event_loop argument; just the connection.
    """
    _id = 0
    success = await connection.insert(_id, "data")
    assert success == True


@pytest.mark.dependency(depends=['test_insert'])
@pytest.mark.asyncio
async def test_find(connection, event_loop):  # <-- does not need event_loop arg
    """ Test database find.

        NB does not need event_loop argument; just the connection.
    """
    _id = 0
    data = await connection.find(_id)
    assert data == "data"

解决方案是用模块作用域重新定义 event_loop fixture。将其包含在测试文件中。

@pytest.fixture(scope="module")
def event_loop():
    loop = asyncio.get_event_loop()
    yield loop
    loop.close()

在 github 中针对 pytest-asyncio (link) 提出了类似的 ScopeMismatch 问题。解决方案(如下)适用于我:

@pytest.yield_fixture(scope='class')
def event_loop(request):
    loop = asyncio.get_event_loop_policy().new_event_loop()
    yield loop
    loop.close()