def test_task_priority(self):
# Automatic random priority
base.RANDOM_TASK_PRIORITY_RANGE = (10, 20)
bot = build_spider(SimpleSpider, priority_mode='random')
bot.setup_queue()
task = Task('baz', url='http://xxx.com')
self.assertEqual(task.priority, None)
bot.add_task(task)
self.assertTrue(10 <= task.priority <= 20)
# Automatic constant priority
base.DEFAULT_TASK_PRIORITY = 33
bot = build_spider(SimpleSpider, priority_mode='const')
bot.setup_queue()
task = Task('baz', url='http://xxx.com')
self.assertEqual(task.priority, None)
bot.add_task(task)
self.assertEqual(33, task.priority)
# Automatic priority does not override explictily setted priority
base.DEFAULT_TASK_PRIORITY = 33
bot = build_spider(SimpleSpider, priority_mode='const')
bot.setup_queue()
task = Task('baz', url='http://xxx.com', priority=1)
self.assertEqual(1, task.priority)
bot.add_task(task)
self.assertEqual(1, task.priority)
self.assertRaises(SpiderMisuseError,
lambda: SimpleSpider(priority_mode='foo'))
def test_spider_custom_proxy_source(self):
proxy_port = self.server.port
class TestSpider(Spider):
def task_page(self, grab, unused_task):
self.stat.collect('ports',
int(grab.doc.headers.get('Listen-Port', 0)))
class CustomProxySource(BaseProxySource):
def load(self):
return [
Proxy(ADDRESS, proxy_port, None, None, 'http'),
]
def load_raw_data(self):
return None
bot = build_spider(TestSpider)
bot.setup_queue()
bot.load_proxylist(CustomProxySource())
bot.add_task(Task('page', url='http://yandex.ru/'))
bot.run()
self.assertEqual(self.server.request['headers']['host'], 'yandex.ru')
self.assertEqual(set(bot.stat.collections['ports']),
set([TEST_SERVER_PORT]))
def test_task_callback(self):
class TestSpider(Spider):
def task_page(self, unused_grab, unused_task):
self.meta['tokens'].append('0_handler')
class FuncWithState(object):
def __init__(self, tokens):
self.tokens = tokens
def __call__(self, grab, task):
self.tokens.append('1_func')
tokens = []
func = FuncWithState(tokens)
bot = build_spider(TestSpider, )
bot.meta['tokens'] = tokens
bot.setup_queue()
# classic handler
bot.add_task(Task('page', url=self.server.get_url()))
# callback option overried classic handler
bot.add_task(Task('page', url=self.server.get_url(), callback=func))
# callback and null task name
bot.add_task(Task(name=None, url=self.server.get_url(), callback=func))
# callback and default task name
bot.add_task(Task(url=self.server.get_url(), callback=func))
bot.run()
self.assertEqual(['0_handler', '1_func', '1_func', '1_func'],
sorted(tokens))
def test_spider_custom_proxy_source(self):
proxy_port = self.server.port
class TestSpider(Spider):
def task_page(self, grab, unused_task):
self.stat.collect(
'ports', int(grab.doc.headers.get('Listen-Port', 0)))
class CustomProxySource(BaseProxySource):
def load(self):
return [
Proxy(ADDRESS, proxy_port, None, None, 'http'),
]
def load_raw_data(self):
return None
bot = build_spider(TestSpider)
bot.setup_queue()
bot.load_proxylist(CustomProxySource())
bot.add_task(Task('page', url='http://yandex.ru/'))
bot.run()
self.assertEqual(self.server.request['headers']['host'], 'yandex.ru')
self.assertEqual(set(bot.stat.collections['ports']),
set([TEST_SERVER_PORT]))
def test_delay_error(self):
bot = build_spider(self.SimpleSpider)
self.setup_queue(bot)
bot.task_queue.clear()
self.assertRaises(SpiderMisuseError,
bot.add_task,
Task('page', url=self.server.get_url(), delay=1))
def test_generator_with_invalid_url(self):
class SomeSpider(Spider):
def task_generator(self):
yield Task('page', url=INVALID_URL)
bot = build_spider(SomeSpider)
bot.run()
def test_task_nohandler_error(self):
class TestSpider(Spider):
pass
bot = build_spider(TestSpider, )
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url()))
self.assertRaises(NoTaskHandler, bot.run)
def test_task_retry(self):
self.server.response['get.data'] = 'xxx'
self.server.response_once['code'] = 403
bot = build_spider(SimpleSpider)
bot.setup_queue()
bot.add_task(Task('baz', self.server.get_url()))
bot.run()
self.assertEqual(b'xxx', bot.stat.collections['SAVED_ITEM'][0])
def test_generator_with_invalid_url(self):
class SomeSpider(Spider):
def task_generator(self):
yield Task('page', url=INVALID_URL)
bot = build_spider(SomeSpider)
bot.run()
def test_task_retry(self):
self.server.response['get.data'] = 'xxx'
self.server.response_once['code'] = 403
bot = build_spider(SimpleSpider)
bot.setup_queue()
bot.add_task(Task('baz', self.server.get_url()))
bot.run()
self.assertEqual(b'xxx', bot.stat.collections['SAVED_ITEM'][0])
def test_fatal_error(self):
class TestSpider(Spider):
def task_page(self, unused_grab, unused_task):
raise FatalError
bot = build_spider(TestSpider)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url()))
self.assertRaises(FatalError, bot.run)
def test_spider(self):
self.server.response['get.data'] = 'Hello spider!'
self.server.response['sleep'] = 0
bot = build_spider(SimpleSpider)
bot.setup_queue()
bot.add_task(Task('baz', self.server.get_url()))
bot.run()
self.assertEqual(b'Hello spider!',
bot.stat.collections['SAVED_ITEM'][0])
def test_data_nohandler_error(self):
class TestSpider(Spider):
def task_page(self, unused_grab, unused_task):
yield Data('foo', num=1)
bot = build_spider(TestSpider)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url()))
self.assertRaises(NoDataHandler, bot.run)
def test_spider_nonmp_changes(self):
"""This test tests that in non-multiprocess-mode changes made
inside handler applied to main spider instance."""
bot = build_spider(self.SimpleSpider)
bot.setup_queue()
bot.meta['url'] = self.server.get_url()
bot.add_task(Task('page', self.server.get_url()))
bot.run()
self.assertEqual(4, bot.foo_count)
def test_queue_length(self):
bot = build_spider(self.SimpleSpider)
self.setup_queue(bot)
bot.task_queue.clear()
for _ in six.moves.range(5):
bot.add_task(Task('page', url=self.server.get_url()))
self.assertEqual(5, bot.task_queue.size())
bot.run()
self.assertEqual(0, bot.task_queue.size())
def test_spider(self):
self.server.response['get.data'] = 'Hello spider!'
self.server.response['sleep'] = 0
bot = build_spider(SimpleSpider)
bot.setup_queue()
bot.add_task(Task('baz', self.server.get_url()))
bot.run()
self.assertEqual(b'Hello spider!',
bot.stat.collections['SAVED_ITEM'][0])
def test_fatal_error(self):
class TestSpider(Spider):
def task_page(self, unused_grab, unused_task):
raise FatalError
bot = build_spider(TestSpider)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url()))
self.assertRaises(FatalError, bot.run)
def test_task_raw(self):
class TestSpider(Spider):
def task_page(self, grab, unused_task):
self.stat.collect('codes', grab.doc.code)
self.server.response['code'] = 502
bot = build_spider(TestSpider, network_try_limit=1)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url()))
bot.add_task(Task('page', url=self.server.get_url()))
bot.run()
self.assertEqual(0, len(bot.stat.collections['codes']))
bot = build_spider(TestSpider, network_try_limit=1)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url(), raw=True))
bot.add_task(Task('page', url=self.server.get_url(), raw=True))
bot.run()
self.assertEqual(2, len(bot.stat.collections['codes']))
def get_configured_spider(self, pause=None, spider_options=None):
bot = build_spider(
SimpleSpider,
meta={'server': self.server, 'pause': (pause or [])},
parser_pool_size=1,
**(spider_options or {})
)
self.setup_cache(bot)
bot.cache_reader_service.backend.clear()
bot.setup_queue()
return bot
def test_task_limit(self):
class CustomSimpleSpider(SimpleSpider):
def create_grab_instance(self, **kwargs):
return Grab(connect_timeout=1, timeout=1)
self.server.response['get.data'] = 'Hello spider!'
self.server.response['sleep'] = 1.1
bot = build_spider(CustomSimpleSpider, network_try_limit=1)
#bot.setup_grab(connect_timeout=1, timeout=1)
bot.setup_queue()
bot.add_task(Task('baz', self.server.get_url()))
bot.run()
self.assertEqual(bot.stat.counters['spider:task-baz'], 1)
bot = build_spider(SimpleSpider, task_try_limit=2)
bot.setup_queue()
bot.add_task(Task('baz', self.server.get_url(), task_try_count=3))
bot.run()
self.assertEqual(bot.stat.counters['spider:request-network'], 0)
def test_task_limit(self):
class CustomSimpleSpider(SimpleSpider):
def create_grab_instance(self, **kwargs):
return Grab(connect_timeout=1, timeout=1)
self.server.response['get.data'] = 'Hello spider!'
self.server.response['sleep'] = 1.1
bot = build_spider(CustomSimpleSpider, network_try_limit=1)
#bot.setup_grab(connect_timeout=1, timeout=1)
bot.setup_queue()
bot.add_task(Task('baz', self.server.get_url()))
bot.run()
self.assertEqual(bot.stat.counters['spider:task-baz'], 1)
bot = build_spider(SimpleSpider, task_try_limit=2)
bot.setup_queue()
bot.add_task(Task('baz', self.server.get_url(), task_try_count=3))
bot.run()
self.assertEqual(bot.stat.counters['spider:request-network'], 0)
def test_response_not_valid(self):
class SomeSimpleSpider(Spider):
def task_page(self, unused_grab, unused_task):
self.stat.inc('xxx')
raise ResponseNotValid
bot = build_spider(SomeSimpleSpider)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url()))
bot.run()
self.assertEqual(bot.task_try_limit, bot.stat.counters['xxx'])
def test_add_task_invalid_url_raise_error(self):
class TestSpider(Spider):
pass
bot = build_spider(TestSpider, )
bot.setup_queue()
self.assertRaises(SpiderError, bot.add_task,
Task('page', url='zz://zz'), raise_error=True)
self.assertEqual(0, bot.task_queue.size())
bot.add_task(Task('page', url='http://example.com/'))
self.assertEqual(1, bot.task_queue.size())
def test_schedule_list_clear(self):
bot = build_spider(self.SimpleSpider)
self.setup_queue(bot)
bot.task_queue.clear()
for delay in six.moves.range(5):
bot.add_task(Task('page', url=self.server.get_url(),
delay=delay + 1))
self.assertEqual(5, len(bot.task_queue.schedule_list))
bot.task_queue.clear()
self.assertEqual(0, len(bot.task_queue.schedule_list))
def test_task_useragent(self):
bot = build_spider(SimpleSpider, )
bot.setup_queue()
grab = Grab()
grab.setup(url=self.server.get_url())
grab.setup(user_agent='Foo')
task = Task('baz', grab=grab)
bot.add_task(task.clone())
bot.run()
self.assertEqual(self.server.request['headers']['User-Agent'], 'Foo')
def get_configured_spider(self, pause=None, spider_options=None):
bot = build_spider(SimpleSpider,
meta={
'server': self.server,
'pause': (pause or [])
},
parser_pool_size=1,
**(spider_options or {}))
self.setup_cache(bot)
bot.cache_reader_service.backend.clear()
bot.setup_queue()
return bot
def test_task_queue_clear(self):
class TestSpider(Spider):
def task_page(self, unused_grab, unused_task):
self.stop()
bot = build_spider(TestSpider)
bot.setup_queue()
for _ in six.moves.range(5):
bot.add_task(Task('page', url=self.server.get_url()))
self.assertEqual(5, bot.task_queue.size())
bot.run()
self.assertEqual(0, bot.task_queue.size())
def test_task_queue_clear(self):
class TestSpider(Spider):
def task_page(self, unused_grab, unused_task):
self.stop()
bot = build_spider(TestSpider)
bot.setup_queue()
for _ in six.moves.range(5):
bot.add_task(Task('page', url=self.server.get_url()))
self.assertEqual(5, bot.task_queue.size())
bot.run()
self.assertEqual(0, bot.task_queue.size())
def test_exception_from_data_handler(self):
class TestSpider(Spider):
def task_page(self, unused_grab, unused_task):
yield Data('foo', num=1)
def data_foo(self, num): # pylint: disable=unused-argument
raise Exception('Shit happens!')
bot = build_spider(TestSpider)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url()))
bot.run()
self.assertTrue('data_foo' in bot.stat.collections['fatal'][0])
def test_initial_urls(self):
url = self.server.get_url()
class TestSpider(Spider):
initial_urls = [url]
def task_initial(self, unused_grab, unused_task):
self.stat.inc('foo', 1)
bot = build_spider(TestSpider)
bot.run()
self.assertEqual(1, bot.stat.counters['foo'])
def test_handler_result_none(self):
class TestSpider(Spider):
def prepare(self):
# pylint: disable=attribute-defined-outside-init
self.points = []
def task_page(self, unused_grab, unused_task):
yield None
bot = build_spider(TestSpider)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url()))
bot.run()
def test_handler_result_none(self):
class TestSpider(Spider):
def prepare(self):
# pylint: disable=attribute-defined-outside-init
self.points = []
def task_page(self, unused_grab, unused_task):
yield None
bot = build_spider(TestSpider)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url()))
bot.run()
def test_check_task_limits_invalid_value(self):
class TestSpider(Spider):
def task_page(self, grab, task):
pass
def check_task_limits(self, task):
return False, 'zz'
bot = build_spider(TestSpider)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url(),
fallback_name='fallback_zz'))
self.assertRaises(SpiderError, bot.run)
def test_generator(self):
server = self.server
class TestSpider(Spider):
def task_generator(self):
for _ in six.moves.range(1111):
yield Task('page', url=server.get_url())
def task_page(self, unused_grab, unused_task):
self.stat.inc('count')
bot = build_spider(TestSpider)
bot.run()
self.assertEqual(bot.stat.counters['count'], 1111)
def test_worker_restored(self):
class TestSpider(Spider):
def task_page(self, unused_grab, unused_task):
pass
bot = build_spider(
TestSpider,
parser_requests_per_process=2,
)
bot.setup_queue()
for _ in range(5):
bot.add_task(Task('page', url=self.server.get_url()))
bot.run()
self.assertTrue(bot.stat.counters['parser:worker-restarted'] == 2)
def test_task_url(self):
bot = build_spider(SimpleSpider, )
bot.setup_queue()
task = Task('baz', url='http://xxx.com')
self.assertEqual('http://xxx.com', task.url)
bot.add_task(task)
self.assertEqual('http://xxx.com', task.url)
self.assertEqual(None, task.grab_config)
grab = Grab(url='http://yyy.com')
task = Task('baz', grab=grab)
bot.add_task(task)
self.assertEqual('http://yyy.com', task.url)
self.assertEqual('http://yyy.com', task.grab_config['url'])
def test_render_stats(self):
class TestSpider(Spider):
def prepare(self):
self.stat.logging_period = 0
self.stat.inc('foo')
def task_page(self, grab, task):
pass
bot = build_spider(TestSpider)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url()))
bot.run()
bot.render_stats()
def test_generator(self):
server = self.server
class TestSpider(Spider):
def task_generator(self):
for _ in six.moves.range(1111):
yield Task('page', url=server.get_url())
def task_page(self, unused_grab, unused_task):
self.stat.inc('count')
bot = build_spider(TestSpider)
bot.run()
self.assertEqual(bot.stat.counters['count'], 1111)
def test_handler_result_invalid(self):
class TestSpider(Spider):
def prepare(self):
# pylint: disable=attribute-defined-outside-init
self.points = []
def task_page(self, unused_grab, unused_task):
yield 1
bot = build_spider(TestSpider)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url()))
#bot.run()
#self.assertEqual(1, bot.stat.counters['spider:error-spidererror'])
self.assertRaises(SpiderError, bot.run)
def test_task_clone_post_request(self):
class TestSpider(Spider):
def task_foo(self, unused_grab, task):
if not task.get('fin'):
yield task.clone(fin=True)
bot = build_spider(TestSpider)
bot.setup_queue()
grab = Grab()
grab.setup(url=self.server.get_url(), post={'x': 'y'})
task = Task('foo', grab=grab)
bot.add_task(task)
bot.run()
self.assertEqual('POST', self.server.request['method'])
def test_create_table(self):
self.server.response['get.data'] = content_generator()
class TestSpider(Spider):
def task_page(self, grab, task):
pass
bot = build_spider(TestSpider)
self.setup_cache(bot)
bot.cache_reader_service.backend.cursor.execute('begin')
bot.cache_reader_service.backend.cursor.execute('DROP TABLE cache')
bot.cache_reader_service.backend.cursor.execute('commit')
self.setup_cache(bot)
bot.cache_reader_service.backend.clear()
self.assertEqual(0, bot.cache_reader_service.backend.size())
def test_handler_result_invalid(self):
class TestSpider(Spider):
def prepare(self):
# pylint: disable=attribute-defined-outside-init
self.points = []
def task_page(self, unused_grab, unused_task):
yield 1
bot = build_spider(TestSpider)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url()))
#bot.run()
#self.assertEqual(1, bot.stat.counters['spider:error-spidererror'])
self.assertRaises(SpiderError, bot.run)
def test_basic_priority(self):
bot = build_spider(self.SimpleSpider, parser_pool_size=1,
thread_number=1)
self.setup_queue(bot)
bot.task_queue.clear()
requested_urls = {}
for priority in (4, 2, 1, 5):
url = self.server.get_url() + '?p=%d' % priority
requested_urls[priority] = url
bot.add_task(Task('page', url=url,
priority=priority))
bot.run()
urls = [x[1] for x in sorted(requested_urls.items(),
key=lambda x: x[0])]
self.assertEqual(urls, bot.stat.collections['url_history'])
def test_data_simple_case(self):
class TestSpider(Spider):
def prepare(self):
# pylint: disable=attribute-defined-outside-init
self.data_processed = []
def task_page(self, unused_grab, unused_task):
yield Data('foo', number=1)
def data_foo(self, number):
self.data_processed.append(number)
bot = build_spider(TestSpider)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url()))
bot.run()
self.assertEqual(bot.data_processed, [1])
def test_setup_proxylist(self):
with temp_file() as proxy_file:
content = '\n'.join(x['proxy'] for x in
self.extra_servers.values())
with open(proxy_file, 'w') as out:
out.write(content)
# Simple test, one task
bot = build_spider(SimpleSpider, thread_number=1)
bot.load_proxylist(proxy_file, 'text_file')
bot.setup_queue()
bot.add_task(Task('baz', grab=Grab(url='http://yandex.ru',
debug=True)))
bot.run()
serv = [x['server'] for x in self.extra_servers.values()
if x['server'].request['done']][0]
self.assertEqual(serv.request['headers']['host'], 'yandex.ru')
self.assertEqual(1, len(set(bot.stat.collections['ports'])))
def test_stat_error_name_threaded_urllib3(self):
server = self.server
server.response['sleep'] = 2
class SimpleSpider(Spider):
def prepare(self):
self.network_try_limit = 1
def task_generator(self):
grab = Grab(url=server.get_url(), timeout=1)
yield Task('page', grab=grab)
def task_page(self, grab, unused_task):
pass
bot = build_spider(SimpleSpider)
bot.run()
self.assertTrue('error:read-timeout-error' in bot.stat.counters)
def test_complex_data(self):
class TestSpider(Spider):
def prepare(self):
# pylint: disable=attribute-defined-outside-init
self.data_processed = []
def task_page(self, unused_grab, unused_task):
yield Data('foo', one=1, two=2, bar='gaz')
def data_foo(self, one, two, **kwargs):
self.data_processed.append(one)
self.data_processed.append(two)
self.data_processed.append(kwargs)
bot = build_spider(TestSpider)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url()))
bot.run()
self.assertEqual(bot.data_processed, [1, 2, {'bar': 'gaz'}])
def test_fallback_handler_by_default_name(self):
class TestSpider(Spider):
def prepare(self):
# pylint: disable=attribute-defined-outside-init
self.points = []
def task_page(self, grab, task):
pass
def task_page_fallback(self, unused_task):
self.points.append(1)
self.server.response['code'] = 403
bot = build_spider(TestSpider, network_try_limit=1)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url()))
bot.run()
self.assertEqual(bot.points, [1])
def test_too_many_redirects(self):
class TestSpider(Spider):
def task_page(self, unused_grab, unused_task):
pass
bot = build_spider(TestSpider)
bot.setup_queue()
bot.add_task(Task('page', url=self.server.get_url()))
self.server.response['headers'] = [
('Location', self.server.get_url()),
]
self.server.response['code'] = 302
bot.run()
self.assertEqual(
1, len(bot.stat.collections['network-count-rejected'])
)
self.assertTrue('error:too-many-redirects' in bot.stat.counters)
def test_grab_attribute_exception(self):
server = self.server
server.response['sleep'] = 2
class SimpleSpider(Spider):
def task_generator(self):
grab = Grab()
grab.setup(url=server.get_url(),
timeout=1)
yield Task('page', grab=grab,
raw=True)
def task_page(self, grab, unused_task):
self.meta['exc'] = grab.exception
bot = build_spider(SimpleSpider)
bot.run()
self.assertTrue(isinstance(bot.meta['exc'], GrabTimeoutError))
def test_redirect_with_invalid_url(self):
server = self.server
class TestSpider(Spider):
def task_generator(self):
# pylint: disable=attribute-defined-outside-init
self.done_counter = 0
# pylint: enable=attribute-defined-outside-init
yield Task('page', url=server.get_url())
def task_page(self, grab, task):
pass
self.server.response_once['code'] = 301
self.server.response_once['headers'] = [
('Location', INVALID_URL),
]
bot = build_spider(TestSpider, network_try_limit=1)
bot.run()
def test_setup_proxylist2(self):
with temp_file() as proxy_file:
content = '\n'.join(x['proxy'] for x in
self.extra_servers.values())
with open(proxy_file, 'w') as out:
out.write(content)
# By default auto_change is True
bot = build_spider(SimpleSpider, thread_number=1)
bot.load_proxylist(proxy_file, 'text_file')
bot.setup_queue()
for _ in six.moves.range(10):
bot.add_task(Task('baz', 'http://yandex.ru'))
bot.run()
servers = [x['server'] for x in self.extra_servers.values()
if x['server'].request['done']]
for serv in servers:
self.assertEqual(serv.request['headers']['host'], 'yandex.ru')
self.assertTrue(len(set(bot.stat.collections['ports'])) > 1)
def test_schedule(self):
"""
In this test I create a number of delayed task
and then check the order in which they was executed
"""
server = self.server
class TestSpider(Spider):
def task_generator(self):
yield Task('page', url=server.get_url(), delay=1.5, num=3)
yield Task('page', url=server.get_url(), delay=4.5, num=2)
yield Task('page', url=server.get_url(), delay=3, num=4)
yield Task('page', url=server.get_url(), num=1)
def task_page(self, unused_grab, task):
self.stat.collect('numbers', task.num)
bot = build_spider(TestSpider, thread_number=1)
self.setup_queue(bot)
bot.run()
self.assertEqual(bot.stat.collections['numbers'], [1, 3, 4, 2])