def test_build_request(self):
r = Request.of("http://test")
self.assertEqual(r.url, "http://test")
r1 = Request.of(r)
self.assertIs(r1, r)
with self.assertRaises(TypeError):
Request.of(1)
def test_recover(self):
spider = RecoverMySpider()
spider.filter = spider.filter + CustomFilter(lambda x: True)
spider.set_session(env.session)
list(
run_and_get_result(
spider.crawl(
Request.of("http://localhost:5000/test_extract"))))
spider.stash(EXE_PATH)
del spider
recovered_spider = RecoverMySpider()
recovered_spider.recover(EXE_PATH)
self.assertTrue(
recovered_spider.crawled_filter.contains(
Request.of("http://localhost:5000/a.html")))
def test_recover_queue(self):
queue = RecoverableRequestQueue()
queue.put(Request.of("1"))
queue.put(Request.of("2"))
queue.put(Request.of("3"))
queue.stash(EXE_PATH)
del queue
recovered_queue = RecoverableRequestQueue()
recovered_queue.recover(EXE_PATH)
self.assertFalse(recovered_queue.empty())
i = 1
while not recovered_queue.empty():
request = recovered_queue.get()
self.assertEqual(request.url, str(i))
i += 1
def test_auto_save(self):
test_data = {"key": 1}
spider = MySpider()
spider.set_session(mock_env.env.session)
spider.start_targets = ["http://localhost:5000/test_extract"]
spider.auto_save_frequency = 10
task = CountDownRecoverableTask(spider)
task.add_actions(lambda: test_data.clear())
for _ in range(5):
mock_env.env.loop.run_until_complete(task.run())
task._request_queue.put(
Request.of("http://localhost:5000/test_extract"))
self.assertEqual(test_data, {"key": 1})
for _ in range(5):
mock_env.env.loop.run_until_complete(task.run())
task._request_queue.put(
Request.of("http://localhost:5000/test_extract"))
self.assertEqual(test_data, {})
def test_stash(self):
spider = MySpider()
spider.start_targets = ["http://localhost:5000/test_extract"]
task = RecoverableTask(spider)
exists(".task_stash") or makedirs(".task_stash")
task.stash(".task_stash")
del task
recovered_task = RecoverableTask(MySpider())
self.assertTrue(recovered_task.can_recover(".task_stash"))
recovered_task.recover(".task_stash")
self.assertTrue(
recovered_task.spider.crawled_filter.contains(
Request.of("http://localhost:5000/test_extract")))
def test_crawled_spider(self):
spider = MySpider()
with self.assertRaises(TypeError):
spider.crawled_filter = spider.filter
start_request = Request.of("http://localhost:5000/test_extract")
self.assertFalse(spider.crawled_filter.accept(start_request))
def test_reg_filter(self):
f = URLRegFilter(r"http://")
self.assertTrue(f.accept(Request.of("http://www.baidu.com")))
def assert_false(self, r, url):
self.assertFalse(r.accept(Request.of(url)))
def assert_true(self, r, url):
self.assertTrue(r.accept(Request.of(url)))
def from_url_or_request(url: str):
return Request.of(url)