Astroquery Testing

Testing in astroquery is a bit more complicated than in other modules since we depend on remote servers to supply data. In order to keep the tests green and fast, we use monkeypatching to test most functions on local copies of the data.

In order to set up testing for any given module, you therefore need to have local copies of the data.

The testing directory structure should look like:


This file should contain only tests that do not require an internet connection. It also contains the tricky monkeypatching components. At a minimum, monkeypatching requires a few methods that are defined locally in the test file for each module.


At a minimum, monkeypatching will require these changes:

class MockResponse(object):

    def __init__(self, content):
        self.content = content

MockResponse is an object intended to have any of the attributes that a normal requests.Response object would have. However, it only needs to implement the methods that are actually used within the tests.

The tricky bits are in the pytest.fixture.

The first little magical function is the patch_x function, where x should either be post or get.

def patch_get(request):
    mp = request.getfixturevalue("monkeypatch")

    mp.setattr(requests.Session, 'request', get_mockreturn)
    return mp

This function, when called, changes the requests.Session’s request method to call the get_mockreturn function, defined below. @pytest.fixture means that, if any function in this file accepts patch_get as an argument, patch_get will be called prior to running that function.

get_mockreturn is simple but important: this is where you define a function to return the appropriate data stored in the data/ directory as a readable object within the MockResponse class:

def get_mockreturn(url, params=None, timeout=10):
    filename = data_path(DATA_FILES['votable'])
    content = open(filename, 'r').read()
    return MockResponse(content)

data_path is a simple function that looks for the data directory local to the file.

def data_path(filename):
    data_dir = os.path.join(os.path.dirname(__file__), 'data')
    return os.path.join(data_dir, filename)

The remote tests are much easier. Just decorate the test class or test functions with @pytest.mark.remote_data.

This file only needs the get_package_data() function, which will tell to include the relevant files when installing.

import os

def get_package_data():
    paths_test = [os.path.join('data', '*.xml')]

    return {'astroquery.module.tests': paths_test}


Narrative documentation should also be tested, the doctest-remote-data directive provides a way to mark code snippets that relies on remote data access.

If any of the examples include saving data files locally, use the testcleanup directive and the cleanup_saved_downloads function at the end of the narrative documentation.