class TestWeiboHarvesterVCR(tests.TestCase):
def setUp(self):
self.working_path = tempfile.mkdtemp()
self.harvester = WeiboHarvester(self.working_path)
self.harvester.state_store = DictHarvestStateStore()
self.harvester.result = HarvestResult()
self.harvester.stop_harvest_seeds_event = threading.Event()
self.harvester.message = {
"id": "test:2",
"type": "weibo_timeline",
"path": "/collections/test_collection_set",
"credentials": {"access_token": tests.WEIBO_ACCESS_TOKEN},
"collection_set": {"id": "test_collection_set"},
"collection": {"id": "test_collection"},
"options": {},
}
def tearDown(self):
if os.path.exists(self.working_path):
shutil.rmtree(self.working_path)
@vcr.use_cassette(filter_query_parameters=["access_token"])
def test_search_vcr(self):
self.harvester.harvest_seeds()
# check the total number, for new users don't how to check
self.assertEqual(self.harvester.result.harvest_counter["weibos"], 181)
# check the harvester status
self.assertTrue(self.harvester.result.success)
@vcr.use_cassette(filter_query_parameters=["access_token"])
def test_incremental_search_vcr(self):
self.harvester.message["options"]["incremental"] = True
collection_set_id = self.harvester.message["collection_set"]["id"]
self.harvester.state_store.set_state(
"weibo_harvester", u"{}.since_id".format(collection_set_id), 3935747172100551
)
self.harvester.harvest_seeds()
# Check harvest result
self.assertTrue(self.harvester.result.success)
# for check the number of get
self.assertEqual(self.harvester.result.harvest_counter["weibos"], 5)
class TestWeiboHarvester(tests.TestCase):
def setUp(self):
self.working_path = tempfile.mkdtemp()
self.harvester = WeiboHarvester(self.working_path)
self.harvester.state_store = DictHarvestStateStore()
self.harvester.result = HarvestResult()
self.harvester.stop_harvest_seeds_event = threading.Event()
self.harvester.message = {
"id": "test:1",
"type": "weibo_timeline",
"path": "/collections/test_collection_set",
"credentials": {"access_token": tests.WEIBO_ACCESS_TOKEN},
"collection_set": {"id": "test_collection_set"},
"collection": {"id": "test_collection"},
"options": {"web_resources": True, "image_sizes": ["Large"]},
}
def tearDown(self):
if os.path.exists(self.working_path):
shutil.rmtree(self.working_path)
@patch("weibo_harvester.Weiboarc", autospec=True)
def test_search_timeline(self, mock_weiboarc_class):
mock_weiboarc = MagicMock(spec=Weiboarc)
# Expecting 2 results. First returns 1tweets. Second returns none.
mock_weiboarc.search_friendships.side_effect = [(weibo1, weibo2), ()]
# Return mock_weiboarc when instantiating a weiboarc.
mock_weiboarc_class.side_effect = [mock_weiboarc]
self.harvester.harvest_seeds()
self.assertDictEqual({"weibos": 2}, self.harvester.result.harvest_counter)
mock_weiboarc_class.assert_called_once_with(tests.WEIBO_ACCESS_TOKEN)
self.assertEqual([call(since_id=None)], mock_weiboarc.search_friendships.mock_calls)
@patch("weibo_harvester.Weiboarc", autospec=True)
def test_incremental_search(self, mock_weiboarc_class):
mock_weiboarc = MagicMock(spec=Weiboarc)
# Expecting 2 searches. First returns 2 weibos,one is none. Second returns none.
mock_weiboarc.search_friendships.side_effect = [(weibo2,), ()]
# Return mock_weiboarc when instantiating a weiboarc.
mock_weiboarc_class.side_effect = [mock_weiboarc]
self.harvester.message["options"] = {
# Incremental means that will only retrieve new results.
"incremental": True
}
collection_set_id = self.harvester.message["collection_set"]["id"]
self.harvester.state_store.set_state(
"weibo_harvester", u"{}.since_id".format(collection_set_id), 3927348724716740
)
self.harvester.harvest_seeds()
self.assertDictEqual({"weibos": 1}, self.harvester.result.harvest_counter)
mock_weiboarc_class.assert_called_once_with(tests.WEIBO_ACCESS_TOKEN)
# since_id must be in the mock calls
self.assertEqual([call(since_id=3927348724716740)], mock_weiboarc.search_friendships.mock_calls)
self.assertNotEqual([call(since_id=None)], mock_weiboarc.search_friendships.mock_calls)
@staticmethod
def _iter_items(items):
# This is useful for mocking out a warc iter
iter_items = []
for item in items:
iter_items.append(IterItem(None, None, None, None, item))
return iter_items
@patch("weibo_harvester.WeiboWarcIter", autospec=True)
def test_process(self, iter_class):
mock_iter = MagicMock(spec=WeiboWarcIter)
mock_iter.__iter__.side_effect = [self._iter_items([weibo3, weibo4, weibo5]).__iter__()]
iter_class.side_effect = [mock_iter]
# These are default harvest options
self.harvester.extract_web_resources = False
self.harvester.extract_images_sizes = []
self.harvester.incremental = False
self.harvester.process_warc("test.warc.gz")
# The default will not sending web harvest
self.assertSetEqual(set(), self.harvester.result.urls_as_set())
iter_class.assert_called_once_with("test.warc.gz")
self.assertEqual(3, self.harvester.result.stats_summary()["weibos"])
# State not set
self.assertIsNone(self.harvester.state_store.get_state("weibo_harvester", "test_collection_set.since_id"))
@patch("weibo_harvester.WeiboWarcIter", autospec=True)
def test_process_incremental(self, iter_class):
mock_iter = MagicMock(spec=WeiboWarcIter)
mock_iter.__iter__.side_effect = [self._iter_items([weibo3, weibo4, weibo5]).__iter__()]
iter_class.side_effect = [mock_iter]
# These are default harvest options
self.harvester.extract_web_resources = False
self.harvester.extract_images_sizes = []
self.harvester.incremental = True
self.harvester.state_store.set_state("weibo_harvester", "test_collection_set.since_id", 3927348724716740)
self.harvester.process_warc("test.warc.gz")
# The default will not sending web harvest
self.assertSetEqual(set(), self.harvester.result.urls_as_set())
iter_class.assert_called_once_with("test.warc.gz")
self.assertEqual(3, self.harvester.result.stats_summary()["weibos"])
# State updated
self.assertEqual(
3973784090711192, self.harvester.state_store.get_state("weibo_harvester", "test_collection_set.since_id")
)
@patch("weibo_harvester.WeiboWarcIter", autospec=True)
def test_process_harvest_options_web(self, iter_class):
mock_iter = MagicMock(spec=WeiboWarcIter)
mock_iter.__iter__.side_effect = [self._iter_items([weibo3, weibo4, weibo5]).__iter__()]
iter_class.side_effect = [mock_iter]
# These are default harvest options
self.harvester.extract_web_resources = True
self.harvester.extract_images_sizes = []
self.harvester.incremental = False
self.harvester.process_warc("test.warc.gz")
# Testing URL1&URL2
self.assertSetEqual(
{"http://t.cn/RqmQ3ko", "http://m.weibo.cn/1618051664/3973767505640890"},
self.harvester.result.urls_as_set(),
)
iter_class.assert_called_once_with("test.warc.gz")
@patch("weibo_harvester.WeiboWarcIter", autospec=True)
def test_process_harvest_options_media(self, iter_class):
mock_iter = MagicMock(spec=WeiboWarcIter)
mock_iter.__iter__.side_effect = [self._iter_items([weibo3, weibo4, weibo5]).__iter__()]
iter_class.side_effect = [mock_iter]
# These are default harvest options
self.harvester.extract_web_resources = False
self.harvester.extract_images_sizes = ["Large", "Medium", "Thumbnail"]
self.harvester.incremental = False
self.harvester.process_warc("test.warc.gz")
# Testing URL3 photos URLs
self.assertSetEqual(
{
"http://ww2.sinaimg.cn/large/6b23a52bgw1f3pjhhyofnj208p06c3yq.jpg",
"http://ww4.sinaimg.cn/large/60718250jw1f3qtzyhai3j20de0vin32.jpg",
"http://ww2.sinaimg.cn/bmiddle/6b23a52bgw1f3pjhhyofnj208p06c3yq.jpg",
"http://ww4.sinaimg.cn/bmiddle/60718250jw1f3qtzyhai3j20de0vin32.jpg",
"http://ww2.sinaimg.cn/thumbnail/6b23a52bgw1f3pjhhyofnj208p06c3yq.jpg",
"http://ww4.sinaimg.cn/thumbnail/60718250jw1f3qtzyhai3j20de0vin32.jpg",
},
self.harvester.result.urls_as_set(),
)
iter_class.assert_called_once_with("test.warc.gz")
class TestWeiboHarvesterVCR(tests.TestCase):
def setUp(self):
self.working_path = tempfile.mkdtemp()
self.harvester = WeiboHarvester(self.working_path)
self.harvester.state_store = DictHarvestStateStore()
self.harvester.result = HarvestResult()
self.harvester.stop_harvest_seeds_event = threading.Event()
self.harvester.message = base_timeline_message
def tearDown(self):
if os.path.exists(self.working_path):
shutil.rmtree(self.working_path)
@vcr.use_cassette(filter_query_parameters=['access_token'])
def test_timeline_vcr(self):
self.harvester.message = base_timeline_message
self.harvester.harvest_seeds()
# check the total number, for new users don't how to check
self.assertEqual(self.harvester.result.harvest_counter["weibos"], 181)
# check the harvester status
self.assertTrue(self.harvester.result.success)
@vcr.use_cassette(filter_query_parameters=['access_token'])
def test_incremental_timeline_vcr(self):
message = copy.deepcopy(base_timeline_message)
message["options"]["incremental"] = True
self.harvester.message = message
collection_set_id = self.harvester.message["collection_set"]["id"]
self.harvester.state_store.set_state(
"weibo_harvester", u"{}.since_id".format(collection_set_id),
3935747172100551)
self.harvester.harvest_seeds()
# Check harvest result
self.assertTrue(self.harvester.result.success)
# for check the number of get
self.assertEqual(self.harvester.result.harvest_counter["weibos"], 5)
@vcr.use_cassette(filter_query_parameters=['access_token'])
def test_search_topic_vcr(self):
self.harvester.message = base_search_message
self.harvester.harvest_seeds()
# check the total number, one search return 200
self.assertEqual(self.harvester.result.harvest_counter["weibos"], 200)
self.assertTrue(self.harvester.result.success)
@vcr.use_cassette(filter_query_parameters=['access_token'])
def test_search_topic_empty_vcr(self):
self.harvester.message = base_search_message
self.harvester.harvest_seeds()
# check the total number, one search return 0 with empty list
self.assertEqual(self.harvester.result.harvest_counter["weibos"], 0)
self.assertTrue(self.harvester.result.success)
@vcr.use_cassette(filter_query_parameters=['access_token'])
def test_incremental_search_topic_vcr(self):
message = copy.deepcopy(base_search_message)
message["options"]["incremental"] = True
self.harvester.message = message
query = self.harvester.message["seeds"][0]["token"]
self.harvester.state_store.set_state("weibo_harvester",
u"{}.since_id".format(query),
4061065610091375)
self.harvester.harvest_seeds()
# Check harvest result
self.assertTrue(self.harvester.result.success)
# for check the number, it count as 6
self.assertEqual(self.harvester.result.harvest_counter["weibos"], 6)
class TestWeiboHarvester(tests.TestCase):
def setUp(self):
self.working_path = tempfile.mkdtemp()
self.harvester = WeiboHarvester(self.working_path)
self.harvester.state_store = DictHarvestStateStore()
self.harvester.result = HarvestResult()
self.harvester.stop_harvest_seeds_event = threading.Event()
self.harvester.message = base_timeline_message
def tearDown(self):
if os.path.exists(self.working_path):
shutil.rmtree(self.working_path)
@patch("weibo_harvester.Weiboarc", autospec=True)
def test_search_timeline(self, mock_weiboarc_class):
mock_weiboarc = MagicMock(spec=Weiboarc)
# Expecting 2 results. First returns 1tweets. Second returns none.
mock_weiboarc.search_friendships.side_effect = [(weibo1, weibo2), ()]
# Return mock_weiboarc when instantiating a weiboarc.
mock_weiboarc_class.side_effect = [mock_weiboarc]
self.harvester.message = base_timeline_message
self.harvester.harvest_seeds()
self.assertDictEqual({"weibos": 2},
self.harvester.result.harvest_counter)
mock_weiboarc_class.assert_called_once_with(tests.WEIBO_ACCESS_TOKEN)
self.assertEqual([call(since_id=None)],
mock_weiboarc.search_friendships.mock_calls)
@patch("weibo_harvester.Weiboarc", autospec=True)
def test_incremental_search_timeline(self, mock_weiboarc_class):
mock_weiboarc = MagicMock(spec=Weiboarc)
# Expecting 2 searches. First returns 2 weibos,one is none. Second returns none.
mock_weiboarc.search_friendships.side_effect = [(weibo2, ), ()]
# Return mock_weiboarc when instantiating a weiboarc.
mock_weiboarc_class.side_effect = [mock_weiboarc]
message = copy.deepcopy(base_timeline_message)
message["options"]["incremental"] = True
self.harvester.message = message
collection_set_id = self.harvester.message["collection_set"]["id"]
self.harvester.state_store.set_state(
"weibo_harvester", u"{}.since_id".format(collection_set_id),
3927348724716740)
self.harvester.harvest_seeds()
self.assertDictEqual({"weibos": 1},
self.harvester.result.harvest_counter)
mock_weiboarc_class.assert_called_once_with(tests.WEIBO_ACCESS_TOKEN)
# since_id must be in the mock calls
self.assertEqual([call(since_id=3927348724716740)],
mock_weiboarc.search_friendships.mock_calls)
self.assertNotEqual([call(since_id=None)],
mock_weiboarc.search_friendships.mock_calls)
@patch("weibo_harvester.Weiboarc", autospec=True)
def test_search_topic(self, mock_weiboarc_class):
mock_weiboarc = MagicMock(spec=Weiboarc)
# search_topic Expecting 2 results. First returns 1tweets. Second returns none.
mock_weiboarc.search_topic.side_effect = [(weibo6, weibo7), ()]
# Return mock_weiboarc when instantiating a weiboarc.
mock_weiboarc_class.side_effect = [mock_weiboarc]
self.harvester.message = base_search_message
self.harvester.harvest_seeds()
query = self.harvester.message["seeds"][0]["token"]
self.assertDictEqual({"weibos": 2},
self.harvester.result.harvest_counter)
mock_weiboarc_class.assert_called_once_with(tests.WEIBO_ACCESS_TOKEN)
self.assertEqual([call(query, since_id=None)],
mock_weiboarc.search_topic.mock_calls)
@patch("weibo_harvester.Weiboarc", autospec=True)
def test_incremental_search_topic(self, mock_weiboarc_class):
mock_weiboarc = MagicMock(spec=Weiboarc)
# search_topic Expecting 2 searches. First returns 1 weibos
mock_weiboarc.search_topic.side_effect = [(weibo7, ), ()]
# Return mock_weiboarc when instantiating a weiboarc.
mock_weiboarc_class.side_effect = [mock_weiboarc]
message = copy.deepcopy(base_search_message)
message["options"]["incremental"] = True
self.harvester.message = message
query = self.harvester.message["seeds"][0]["token"]
self.harvester.state_store.set_state("weibo_harvester",
u"{}.since_id".format(query),
4060927646547531)
self.harvester.harvest_seeds()
mock_weiboarc_class.assert_called_once_with(tests.WEIBO_ACCESS_TOKEN)
self.assertEqual([call(query, since_id=4060927646547531)],
mock_weiboarc.search_topic.mock_calls)
self.assertDictEqual({"weibos": 1},
self.harvester.result.harvest_counter)
@staticmethod
def _iter_items(items):
# This is useful for mocking out a warc iter
iter_items = []
for item in items:
iter_items.append(IterItem(None, None, None, None, item))
return iter_items
@patch("weibo_harvester.WeiboWarcIter", autospec=True)
def test_process_timeline(self, iter_class):
mock_iter = MagicMock(spec=WeiboWarcIter)
mock_iter.__iter__.side_effect = [
self._iter_items([weibo3, weibo4, weibo5]).__iter__()
]
iter_class.side_effect = [mock_iter]
# These are default harvest options
self.harvester.incremental = False
self.harvester.message = base_timeline_message
self.harvester.process_warc("test.warc.gz")
iter_class.assert_called_once_with("test.warc.gz")
self.assertEqual(3, self.harvester.result.stats_summary()["weibos"])
# State not set
self.assertIsNone(
self.harvester.state_store.get_state(
"weibo_harvester", "test_collection_set.since_id"))
@patch("weibo_harvester.WeiboWarcIter", autospec=True)
def test_process_timeline_incremental(self, iter_class):
mock_iter = MagicMock(spec=WeiboWarcIter)
mock_iter.__iter__.side_effect = [
self._iter_items([weibo3, weibo4, weibo5]).__iter__()
]
iter_class.side_effect = [mock_iter]
# These are default harvest options
self.harvester.incremental = True
self.harvester.state_store.set_state("weibo_harvester",
"test_collection_set.since_id",
3927348724716740)
self.harvester.message = base_timeline_message
self.harvester.process_warc("test.warc.gz")
iter_class.assert_called_once_with("test.warc.gz")
self.assertEqual(3, self.harvester.result.stats_summary()["weibos"])
# State updated
self.assertEqual(
3973784090711192,
self.harvester.state_store.get_state(
"weibo_harvester", "test_collection_set.since_id"))
@patch("weibo_harvester.WeiboWarcIter", autospec=True)
def test_process_search_topic(self, iter_class):
mock_iter = MagicMock(spec=WeiboWarcIter)
mock_iter.__iter__.side_effect = [
self._iter_items([weibo6, weibo7]).__iter__()
]
iter_class.side_effect = [mock_iter]
self.harvester.message = base_search_message
self.harvester.process_warc("test.warc.gz")
self.assertDictEqual({"weibos": 2},
self.harvester.result.stats_summary())
iter_class.assert_called_once_with("test.warc.gz")
# State updated
query = self.harvester.message["seeds"][0]["token"]
self.assertEqual(
None,
self.harvester.state_store.get_state("weibo_harvester",
u"{}.since_id".format(query)))
@patch("weibo_harvester.WeiboWarcIter", autospec=True)
def test_process_search_topic_incremental(self, iter_class):
mock_iter = MagicMock(spec=WeiboWarcIter)
mock_iter.__iter__.side_effect = [
self._iter_items([weibo6, weibo7]).__iter__()
]
iter_class.side_effect = [mock_iter]
self.harvester.message = base_search_message
self.harvester.incremental = True
# check the result
query = self.harvester.message["seeds"][0]["token"]
self.harvester.state_store.set_state("weibo_harvester",
u"{}.since_id".format(query),
4060927646547530)
self.harvester.process_warc("test.warc.gz")
self.assertDictEqual({"weibos": 2},
self.harvester.result.stats_summary())
iter_class.assert_called_once_with("test.warc.gz")
# State updated
self.assertEqual(
4060928330955796,
self.harvester.state_store.get_state("weibo_harvester",
u"{}.since_id".format(query)))