Tests should use VCR or a fixture server and avoid connecting to third-party services

Some tests connect to third-party services, which we should avoid because they could cause intermittent and confusing failures in CI (like this one for @vbanos: #233 (comment)). This also prevents tests from being run locally without a network connection, which is also no fun.

Our web_monitoring.db tests use VCR for this:

web-monitoring-processing/web_monitoring/tests/test_db.py

Lines 14 to 59 in 70292c8

    
           # This stashes web-monitoring-dbserver responses in JSON files (one per test) 
        
           # so that an actual server does not have to be running. 
        
           cassette_library_dir = str(Path(__file__).parent / Path('cassettes')) 
        
           db_vcr = vcr.VCR( 
        
                    serializer='json', 
        
                    cassette_library_dir=cassette_library_dir, 
        
                    record_mode='once', 
        
                    match_on=['uri', 'method'], 
        
           ) 
        
           global_stash = {}  # used to pass info between tests 
        
           # Refers to real data that is part of the 'seed' dataset in web-monitoring-db 
        
           URL = 'https://www3.epa.gov/climatechange/impacts/society.html' 
        
           SITE = 'site:EPA - www3.epa.gov' 
        
           AGENCY = 'EPA' 
        
           PAGE_ID = '3d068c64-967a-4ec7-af49-f8fa0f19e6f1' 
        
           TO_VERSION_ID = '795e6ff4-fcc0-444c-9f31-2b156f7dd4d4' 
        
           VERSIONISTA_ID = '13708349' 
        
           # This is used in new Versions that we add. 
        
           TIME = datetime(2017, 11, 15, tzinfo=timezone.utc) 
        
           NEW_VERSION_ID = '06620776-d347-4abd-a423-a871620299a9' 
        
           # The only matters when re-recording the tests for vcr. 
        
           AUTH = {'url': "http://localhost:3000", 
        
                   'email': "seed-admin@example.com", 
        
                   'password': "PASSWORD"} 
        
           def test_missing_creds(): 
        
               try: 
        
                   env = os.environ.copy() 
        
                   os.environ.clear() 
        
                   with pytest.raises(MissingCredentials): 
        
                       Client.from_env() 
        
                   os.environ.update({'WEB_MONITORING_DB_URL': AUTH['url'], 
        
                                      'WEB_MONITORING_DB_EMAIL': AUTH['email'], 
        
                                      'WEB_MONITORING_DB_PASSWORD': AUTH['password']}) 
        
                   Client.from_env()  # should work 
        
               finally: 
        
                   os.environ.update(env) 
        
           @db_vcr.use_cassette() 
        
           def test_list_pages():

And our utils tests use mock requests for simpler cases:

web-monitoring-processing/web_monitoring/tests/test_utils.py

Lines 48 to 53 in edc47ba

    
           def test_retryable_request_retries(): 
        
               with requests_mock.Mocker() as mock: 
        
                   mock.get('http://test.com', [{'text': 'bad', 'status_code': 503}, 
        
                                                {'text': 'good', 'status_code': 200}]) 
        
                   response = retryable_request('GET', 'http://test.com', backoff=0) 
        
                   assert response.ok

First, we need to see what tests this includes. So far:

Lots of these in test_diffing_server_exc_handling.py

We need to be better about making sure future PRs that add new tests do this, too.

More case in point: https://circleci.com/gh/edgi-govdata-archiving/web-monitoring-processing/585

And again, this prevented @jsnshrmn from doing a release today: https://circleci.com/gh/edgi-govdata-archiving/web-monitoring-processing/772

Update: need to re-check this issue for whether it still applies to anything in this repo after #638.

This is indeed no longer relevant.

	# This stashes web-monitoring-dbserver responses in JSON files (one per test)
	# so that an actual server does not have to be running.
	cassette_library_dir = str(Path(__file__).parent / Path('cassettes'))
	db_vcr = vcr.VCR(
	serializer='json',
	cassette_library_dir=cassette_library_dir,
	record_mode='once',
	match_on=['uri', 'method'],
	)
	global_stash = {} # used to pass info between tests


	# Refers to real data that is part of the 'seed' dataset in web-monitoring-db
	URL = 'https://www3.epa.gov/climatechange/impacts/society.html'
	SITE = 'site:EPA - www3.epa.gov'
	AGENCY = 'EPA'
	PAGE_ID = '3d068c64-967a-4ec7-af49-f8fa0f19e6f1'
	TO_VERSION_ID = '795e6ff4-fcc0-444c-9f31-2b156f7dd4d4'
	VERSIONISTA_ID = '13708349'

	# This is used in new Versions that we add.
	TIME = datetime(2017, 11, 15, tzinfo=timezone.utc)
	NEW_VERSION_ID = '06620776-d347-4abd-a423-a871620299a9'

	# The only matters when re-recording the tests for vcr.
	AUTH = {'url': "http://localhost:3000",
	'email': "seed-admin@example.com",
	'password': "PASSWORD"}


	def test_missing_creds():
	try:
	env = os.environ.copy()
	os.environ.clear()
	with pytest.raises(MissingCredentials):
	Client.from_env()
	os.environ.update({'WEB_MONITORING_DB_URL': AUTH['url'],
	'WEB_MONITORING_DB_EMAIL': AUTH['email'],
	'WEB_MONITORING_DB_PASSWORD': AUTH['password']})
	Client.from_env() # should work
	finally:
	os.environ.update(env)


	@db_vcr.use_cassette()
	def test_list_pages():

	def test_retryable_request_retries():
	with requests_mock.Mocker() as mock:
	mock.get('http://test.com', [{'text': 'bad', 'status_code': 503},
	{'text': 'good', 'status_code': 200}])
	response = retryable_request('GET', 'http://test.com', backoff=0)
	assert response.ok