How to mock nested / multiple layers of return objects in python -

i'm struggling find way of mocking multiple layers / nested return values. in other words, want return magic mock in turn returns magic mock it's own set return values. i'm finding relatively cumbersome , looking more elegant , maintainable solution.

i'm trying test following code efficiently. url returns json string needs further processing:

import json urllib.request import url open  def load_json():     # first return value     response = urlopen("http://someurl.com/api/getjson")     # in turn, contains 2 nested return values read , decode         response_dict = json.loads(response.read().decode('utf-8')) 

this how i've mocked far, extremely inelegant , makes maintenance complicated:

class mytestcase(testcase):     @patch('load_json_path.urlopen')     def test_load_json(self, mock_urlopen):         ### trying simplify of         # third nested return         mock_decode = magicmock(return_value='["myjsondata"]')         # second nested return value         mock_response = magicmock()         mock_response.read.return_value=mock_decode         # first nested return value         mock_urlopen.return_value = mock_response         ### trying simplify of                      load_json() 

in end, i'm trying mock returned data decode function, originates url open function. should possible in 1 line or in simpler way, using perhaps enter methods. ideally mock in test_load_json function:

mock_urlopen.__enter__.loads.__enter__.decode.return_value = '["myjsondata"]' 

unfortunately, can't seem find useful in mock documentation. appreciated.

turns out possible , documented. however, naming not straightforward , needed know 1 looking for. referred mocking chained calls, in fact documented in mock library.

in example, mock_urlopen should this:

    mock_urlopen.return_value.read.return_value.decode.return_value = '["myjsondata"]' 

this works beautifully. more details check out python doc: https://docs.python.org/3/library/unittest.mock-examples.html#mocking-chained-calls