使用 pytest 在测试烧瓶应用程序中使用模拟服务器

Question

tl;dr I am looking for the best practice for testing a flask app which uses external APIs under its routes, and probably this is where I have to simulate the external APIs using mock servers in the testing stage.

---

Ok, all I have got searching pytest-mock and similar packages is that I can mock a function and have the results prepared for test assertions. But unfortunately I cannot find a best-practice explanation for my situation:

I have a flask app, under some of the routes of which I use external APIs and I want to use mock servers replacing those real external API URLs. I have come to two apparently fair solutions:

Using test URLs when testing

What I have got so far is that I can set different URLs dynamically by app.config so that in my conftest.py I can invoke using test URLs, and in reality, my app will be launched by the main URLs. Now my question here is how I can run a mock server automatically in the pytest fixture setup stage to occupy the test URLs localhost ports. I said automatically because I can naively run a manual mock app in background myself, which is silly and useless for automated testing purposes

Using a mock function instead of sending the request to the main URL

I have realized that I can setup a mocker which is a function returning my desired value, and use it alongside calling my to-be-tested app. but the point is I want to use the mock inside my app, not comparing the returning value with a mock generated value

So where am I and what should I do?!

Here is a sample:

My flask app:

app = Flask(__name__)

@app.route('/test')
def test():
    value = request.args.get('dummy_val')

    # using external APIs
    # This is where I need mock server
    output = requests.get(
         'http://external.api/some/url',
         params={'val':value}
    )
    return output*2

my test file:

def test_can_double_value(test_client):
    result = test_client('/test', query_string={'dummy_val':'foo'})
    # test if status code == 200
    # test if string equals to something
    # blah blah blah

Answer 1

Bit late to answer, but I have stumbled across the same problem just now. Previous answers seems to be solving the problem in practice, but if you think about your problem in theory, I think you have 2 options.

Either:

• mock the interface on 'requests' level, which works fine, but exposes your test code to implementation detail, as suggested before me
• create adapters for these services in your code, and replace them with your mocks/stubs in your test code.
• using a test fixture, you can start your own webserver for the process using the code from this gist: https://gist.github.com/eruvanos/f6f62edb368a20aaa880e12976620db8

The benefit of the last approach is quite straightforward. You can leave your code as is when testing, no stubbing/mocking, and create the mocks outside your system, where you have free control over mocks, and don't have to worry about actual implementation (like before, you have to KNOW that the service is called via requests).

Answer 2

Thanks to jonrsharpe I now am aware that this can be done using responses library. Now if we want to mock an API in the fixture, it can be done as follows:

in conftest.py

import response
from pytest import fixture

@fixture
def client():
    # add response
    responses.add(responses.GET,
        'http://mock.api/v',
        json={'key': 'value'},
        status=200
    )
    # do fixture stuff -> here it is yielding app test client
    app.config['TESTING'] = True
    client = app.test_client() 

    yield client

in test_module.py

import response

@response.activate
def test_can_connect(client):
    result = client.get('/test')
    assert result.status_code == 200

and in app.py

@app.route('/test')
def test():
    res = requests.get('http://mock.api/v')
    # blah blah

Comment in responses documentation it is suggested that we add a response in the test function. But as I wanted to add it in a fixture, this is how I did it.