class LogStats(object):
def __init__(self, engine, clock=None):
self.interval = engine.settings.get_float('LOG_STATS_INTERVAL')
if not self.interval:
raise NotConfigured
self.multiplier = 60.0 / self.interval
self.logging = LoopingCall(self.log, clock=clock)
engine.signals.connect(self.engine_started,
signal=signals.engine_started)
engine.signals.connect(self.engine_stopped,
signal=signals.engine_stopped)
engine.signals.connect(self.response_downloaded,
signal=signals.response_downloaded)
self.downloaded = 0
self.downloaded_prev = 0
def engine_started(self):
self.logging.schedule(self.interval)
def engine_stopped(self):
self.logging.cancel()
def response_downloaded(self):
self.downloaded += 1
def log(self):
downloaded_speed = (self.downloaded -
self.downloaded_prev) * self.multiplier
self.downloaded_prev = self.downloaded
log.msg(format='Crawled %(down)d pages (at %(speed)d pages/min).',
level=log.INFO,
down=self.downloaded,
speed=downloaded_speed)
class LogStats(object):
def __init__(self, engine, clock=None):
self.interval = engine.settings.get_float('LOG_STATS_INTERVAL')
if not self.interval:
raise NotConfigured
self.multiplier = 60.0 / self.interval
self.logging = LoopingCall(self.log, clock=clock)
engine.signals.connect(self.engine_started, signal=signals.engine_started)
engine.signals.connect(self.engine_stopped, signal=signals.engine_stopped)
engine.signals.connect(self.response_downloaded, signal=signals.response_downloaded)
self.downloaded = 0
self.downloaded_prev = 0
def engine_started(self):
self.logging.schedule(self.interval)
def engine_stopped(self):
self.logging.cancel()
def response_downloaded(self):
self.downloaded += 1
def log(self):
downloaded_speed = (self.downloaded - self.downloaded_prev) * self.multiplier
self.downloaded_prev = self.downloaded
log.msg(format='Crawled %(down)d pages (at %(speed)d pages/min).',
level=log.INFO, down=self.downloaded, speed=downloaded_speed)
def test_init(self):
# test initializing LoopingCall without overriding its clock
sc = LoopingCall(self.obj.func, *self.default_args,
**self.default_kwargs)
sc.schedule()
sc.cancel()
class LoopingCallTest(unittest.TestCase):
default_args = (10, 'hello')
default_kwargs = {'a': 47, 'b': 'c'}
def setUp(self):
self.clock = Clock()
self.obj = ModifiedObject()
self.sc = LoopingCall(self.obj.func, clock=self.clock,
*self.default_args,
**self.default_kwargs)
def _check(self, args, kwargs):
if args is None:
self.assertIsNone(self.obj.args)
else:
self.assertTupleEqual(self.obj.args, args)
if kwargs is None:
self.assertIsNone(self.obj.kwargs)
else:
self.assertEqual(self.obj.kwargs, kwargs)
def test_init(self):
# test initializing LoopingCall without overriding its clock
sc = LoopingCall(self.obj.func, *self.default_args,
**self.default_kwargs)
sc.schedule()
sc.cancel()
def test_basic(self):
# scheduling
self.assertFalse(self.sc.is_scheduled())
self.sc.schedule(2, count=2, now=False)
# before the first call
self.assertTrue(self.sc.is_scheduled())
self.assertEqual(self.sc.get_time(), 2)
self.assertEqual(self.sc.calls_left(), 2)
self.clock.advance(1)
self.assertEqual(self.obj.num_calls, 0)
# after the first call
self.clock.advance(1)
self.assertEqual(self.obj.num_calls, 1)
self.assertEqual(self.sc.calls_left(), 1)
self._check(self.default_args, self.default_kwargs)
# after the second call
self.clock.advance(2)
self.assertEqual(self.obj.num_calls, 2)
self.assertEqual(self.sc.calls_left(), 0)
self.assertFalse(self.sc.is_scheduled())
# no more calls
self.clock.advance(20)
self.assertEqual(self.obj.num_calls, 2)
def test_now(self):
self.sc.schedule(2, count=2, now=True)
self.assertEqual(self.obj.num_calls, 0)
self.clock.advance(0)
self.assertEqual(self.obj.num_calls, 1)
self.clock.advance(20)
self.assertEqual(self.obj.num_calls, 2)
def test_infinite(self):
self.sc.schedule(2)
self.clock.pump([2] * 100)
self.assertEqual(self.obj.num_calls, 100)
self.assertTrue(self.sc.is_scheduled())
self.assertIsNone(self.sc.calls_left())
def test_cancel(self):
self.sc.schedule(2)
self.clock.advance(1)
self.sc.cancel()
self.clock.advance(20)
self.assertEqual(self.obj.num_calls, 0)
def test_reschedule(self):
self.sc.schedule(2)
self.clock.advance(1)
self.sc.schedule(5)
self.clock.advance(4)
self.assertEqual(self.obj.num_calls, 0)
self.clock.advance(1)
self.assertEqual(self.obj.num_calls, 1)
def test_no_delay(self):
self.sc.schedule()
self.assertEqual(self.obj.num_calls, 0)
self.clock.advance(0)
self.assertEqual(self.obj.num_calls, 1)
self.clock.advance(0)
self.assertEqual(self.obj.num_calls, 2)
def test_nested_schedule(self):
def func(*args, **kwargs):
self.obj.func(*args, **kwargs)
self.sc.schedule()
self.sc.func = func
self.sc.schedule()
self.assertEqual(self.obj.num_calls, 0)
self.clock.advance(0)
self.assertEqual(self.obj.num_calls, 1)
self.clock.advance(0)
self.assertEqual(self.obj.num_calls, 2)
def test_init(self):
# test initializing LoopingCall without overriding its clock
sc = LoopingCall(self.obj.func, *self.default_args,
**self.default_kwargs)
sc.schedule()
sc.cancel()
class LoopingCallTest(unittest.TestCase):
default_args = (10, 'hello')
default_kwargs = {'a': 47, 'b': 'c'}
def setUp(self):
self.clock = Clock()
self.obj = ModifiedObject()
self.sc = LoopingCall(self.obj.func,
clock=self.clock,
*self.default_args,
**self.default_kwargs)
def _check(self, args, kwargs):
if args is None:
self.assertIsNone(self.obj.args)
else:
self.assertTupleEqual(self.obj.args, args)
if kwargs is None:
self.assertIsNone(self.obj.kwargs)
else:
self.assertEqual(self.obj.kwargs, kwargs)
def test_init(self):
# test initializing LoopingCall without overriding its clock
sc = LoopingCall(self.obj.func, *self.default_args,
**self.default_kwargs)
sc.schedule()
sc.cancel()
def test_basic(self):
# scheduling
self.assertFalse(self.sc.is_scheduled())
self.sc.schedule(2, count=2, now=False)
# before the first call
self.assertTrue(self.sc.is_scheduled())
self.assertEqual(self.sc.get_time(), 2)
self.assertEqual(self.sc.calls_left(), 2)
self.clock.advance(1)
self.assertEqual(self.obj.num_calls, 0)
# after the first call
self.clock.advance(1)
self.assertEqual(self.obj.num_calls, 1)
self.assertEqual(self.sc.calls_left(), 1)
self._check(self.default_args, self.default_kwargs)
# after the second call
self.clock.advance(2)
self.assertEqual(self.obj.num_calls, 2)
self.assertEqual(self.sc.calls_left(), 0)
self.assertFalse(self.sc.is_scheduled())
# no more calls
self.clock.advance(20)
self.assertEqual(self.obj.num_calls, 2)
def test_now(self):
self.sc.schedule(2, count=2, now=True)
self.assertEqual(self.obj.num_calls, 0)
self.clock.advance(0)
self.assertEqual(self.obj.num_calls, 1)
self.clock.advance(20)
self.assertEqual(self.obj.num_calls, 2)
def test_infinite(self):
self.sc.schedule(2)
self.clock.pump([2] * 100)
self.assertEqual(self.obj.num_calls, 100)
self.assertTrue(self.sc.is_scheduled())
self.assertIsNone(self.sc.calls_left())
def test_cancel(self):
self.sc.schedule(2)
self.clock.advance(1)
self.sc.cancel()
self.clock.advance(20)
self.assertEqual(self.obj.num_calls, 0)
def test_reschedule(self):
self.sc.schedule(2)
self.clock.advance(1)
self.sc.schedule(5)
self.clock.advance(4)
self.assertEqual(self.obj.num_calls, 0)
self.clock.advance(1)
self.assertEqual(self.obj.num_calls, 1)
def test_no_delay(self):
self.sc.schedule()
self.assertEqual(self.obj.num_calls, 0)
self.clock.advance(0)
self.assertEqual(self.obj.num_calls, 1)
self.clock.advance(0)
self.assertEqual(self.obj.num_calls, 2)
def test_nested_schedule(self):
def func(*args, **kwargs):
self.obj.func(*args, **kwargs)
self.sc.schedule()
self.sc.func = func
self.sc.schedule()
self.assertEqual(self.obj.num_calls, 0)
self.clock.advance(0)
self.assertEqual(self.obj.num_calls, 1)
self.clock.advance(0)
self.assertEqual(self.obj.num_calls, 2)
class Downloader(object):
'''Fetch requests from `request_queue` queue. When downloaded,
put the results into `response_queue` queue. Respect CONCURRENT_REQUESTS
setting.
Requests are further divided into specific slots, based on their domains.
'''
# how many seconds to wait between the checks of request_queue
QUEUE_CHECK_FREQUENCY = 0.1
def __init__(self, settings, request_queue, response_queue,
download_handler=None, clock=None):
self.request_queue = request_queue
self.response_queue = response_queue # queue of responses
self.download_handler = download_handler or GeneralHandler(settings)
self.slots = {}
self.num_in_progress = 0
self.clock = clock or reactor
self.processing = LoopingCall(self.process, clock=self.clock)
self.processing.schedule(self.QUEUE_CHECK_FREQUENCY, now=True)
self.running = True
self.download_delay = settings.get_int('DOWNLOAD_DELAY')
self.randomize_delay = settings.get_int(
'RANDOMIZE_DOWNLOAD_DELAY')
if self.download_delay:
self.total_concurrency = self.domain_concurrency = 1
self.use_domain_specific = False
else:
self.total_concurrency = settings.get_int(
'CONCURRENT_REQUESTS')
self.domain_concurrency = settings.get_int(
'CONCURRENT_REQUESTS_PER_DOMAIN')
if (not self.domain_concurrency or
self.domain_concurrency >= self.total_concurrency):
self.use_domain_specific = False
self.domain_concurrency = self.total_concurrency
else:
self.use_domain_specific = True
def close(self):
self.processing.cancel()
self.running = False
@property
def free_slots(self):
return self.total_concurrency - self.num_in_progress
def is_idle(self):
return self.num_in_progress == 0
def process(self):
while (self.running and not self.response_queue.needs_backout() and
self.request_queue and self.free_slots > 0):
request = self.request_queue.pop()
key, slot = self._get_slot(request)
def remove_in_progress(response):
self.num_in_progress -= 1
self._clear_slots() # clear empty slots
return response
def enqueue_result(request, result):
# in a case, result is actually a Failure
result.request = request
# make sure not to modify response_queue, after stopping the downloader
if self.running:
self.response_queue.push(result)
# don't return anything from here, in a case an error occured -
# we don't want it to be logged
self.num_in_progress += 1
dfd = defer.Deferred().addBoth(remove_in_progress)
dfd.addBoth(partial(enqueue_result, request))
slot.enqueue(request, dfd)
def _get_slot(self, request):
key = request.parsed_url.hostname if self.use_domain_specific else ''
if key not in self.slots:
self.slots[key] = Slot(
self.download_handler,
self.domain_concurrency,
self.download_delay,
self.randomize_delay,
clock=self.clock)
return key, self.slots[key]
def _clear_slots(self):
'''Clear unused slots and avoid memory leaking.'''
if len(self.slots) >= 2 * self.total_concurrency:
to_delete = [k for (k, v) in self.slots.iteritems() if v.is_idle()]
for key in to_delete:
del self.slots[key]
class Downloader(object):
'''Fetch requests from `request_queue` queue. When downloaded,
put the results into `response_queue` queue. Respect CONCURRENT_REQUESTS
setting.
Requests are further divided into specific slots, based on their domains.
'''
# how many seconds to wait between the checks of request_queue
QUEUE_CHECK_FREQUENCY = 0.1
def __init__(self, settings, request_queue, response_queue,
download_handler=None, clock=None):
self.request_queue = request_queue
self.response_queue = response_queue # queue of responses
self.download_handler = download_handler or GeneralHandler(settings)
self.slots = {}
self.num_in_progress = 0
self.clock = clock or reactor
self.processing = LoopingCall(self.process, clock=self.clock)
self.processing.schedule(self.QUEUE_CHECK_FREQUENCY, now=True)
self.running = True
self.download_delay = settings.get_float('DOWNLOAD_DELAY')
self.randomize_delay = settings.get_int(
'RANDOMIZE_DOWNLOAD_DELAY')
if self.download_delay:
self.total_concurrency = self.domain_concurrency = 1
self.use_domain_specific = False
else:
self.total_concurrency = settings.get_int(
'CONCURRENT_REQUESTS')
self.domain_concurrency = settings.get_int(
'CONCURRENT_REQUESTS_PER_DOMAIN')
if (not self.domain_concurrency or
self.domain_concurrency >= self.total_concurrency):
self.use_domain_specific = False
self.domain_concurrency = self.total_concurrency
else:
self.use_domain_specific = True
def close(self):
self.processing.cancel()
self.running = False
@property
def free_slots(self):
return self.total_concurrency - self.num_in_progress
def is_idle(self):
return self.num_in_progress == 0
def process(self):
while (self.running and not self.response_queue.needs_backout() and
self.request_queue and self.free_slots > 0):
request = self.request_queue.pop()
key, slot = self._get_slot(request)
def remove_in_progress(response):
self.num_in_progress -= 1
self._clear_slots() # clear empty slots
return response
def enqueue_result(request, result):
# in a case, result is actually a Failure
result.request = request
# make sure not to modify response_queue, after stopping the downloader
if self.running:
self.response_queue.push(result)
# don't return anything from here, in a case an error occured -
# we don't want it to be logged
self.num_in_progress += 1
dfd = defer.Deferred().addBoth(remove_in_progress)
dfd.addBoth(partial(enqueue_result, request))
slot.enqueue(request, dfd)
def _get_slot(self, request):
key = request.parsed_url.hostname if self.use_domain_specific else ''
if key not in self.slots:
self.slots[key] = Slot(
self.download_handler,
self.domain_concurrency,
self.download_delay,
self.randomize_delay,
clock=self.clock)
return key, self.slots[key]
def _clear_slots(self):
'''Clear unused slots and avoid memory leaking.'''
if len(self.slots) >= 2 * self.total_concurrency:
to_delete = [k for (k, v) in self.slots.iteritems() if v.is_idle()]
for key in to_delete:
del self.slots[key]