由于我不得不自己解决问题，而且似乎很少有关于这个问题的信息，所以我要回答自己的问题。
思路如下。使用 GRequests 的每个请求对象在创建时可以将会话对象作为参数传入。另一方面，会话对象可以挂载 HTTP 适配器，在进行请求时使用。通过创建自己的适配器，我们可以拦截请求并以最适合我们的应用程序的方式对它们进行速率限制。在我的情况下，我最终得到了以下代码。
用于限流的对象：
DEFAULT_BURST_WINDOW = datetime.timedelta(seconds=5)
DEFAULT_WAIT_WINDOW = datetime.timedelta(seconds=15)
class BurstThrottle(object):
max_hits = None
hits = None
burst_window = None
total_window = None
timestamp = None
def __init__(self, max_hits, burst_window, wait_window):
self.max_hits = max_hits
self.hits = 0
self.burst_window = burst_window
self.total_window = burst_window + wait_window
self.timestamp = datetime.datetime.min
def throttle(self):
now = datetime.datetime.utcnow()
if now < self.timestamp + self.total_window:
if (now < self.timestamp + self.burst_window) and (self.hits < self.max_hits):
self.hits += 1
return datetime.timedelta(0)
else:
return self.timestamp + self.total_window - now
else:
self.timestamp = now
self.hits = 1
return datetime.timedelta(0)
HTTP适配器：
class MyHttpAdapter(requests.adapters.HTTPAdapter):
throttle = None
def __init__(self, pool_connections=requests.adapters.DEFAULT_POOLSIZE,
pool_maxsize=requests.adapters.DEFAULT_POOLSIZE, max_retries=requests.adapters.DEFAULT_RETRIES,
pool_block=requests.adapters.DEFAULT_POOLBLOCK, burst_window=DEFAULT_BURST_WINDOW,
wait_window=DEFAULT_WAIT_WINDOW):
self.throttle = BurstThrottle(pool_maxsize, burst_window, wait_window)
super(MyHttpAdapter, self).__init__(pool_connections=pool_connections, pool_maxsize=pool_maxsize,
max_retries=max_retries, pool_block=pool_block)
def send(self, request, stream=False, timeout=None, verify=True, cert=None, proxies=None):
request_successful = False
response = None
while not request_successful:
wait_time = self.throttle.throttle()
while wait_time > datetime.timedelta(0):
gevent.sleep(wait_time.total_seconds(), ref=True)
wait_time = self.throttle.throttle()
response = super(MyHttpAdapter, self).send(request, stream=stream, timeout=timeout,
verify=verify, cert=cert, proxies=proxies)
if response.status_code != 429:
request_successful = True
return response
设置：
requests_adapter = adapter.MyHttpAdapter(
pool_connections=__CONCURRENT_LIMIT__,
pool_maxsize=__CONCURRENT_LIMIT__,
max_retries=0,
pool_block=False,
burst_window=datetime.timedelta(seconds=5),
wait_window=datetime.timedelta(seconds=20))
requests_session = requests.session()
requests_session.mount('http://', requests_adapter)
requests_session.mount('https://', requests_adapter)
unsent_requests = (grequests.get(url,
hooks={'response': handle_response},
session=requests_session) for url in urls)
grequests.map(unsent_requests, size=__CONCURRENT_LIMIT__)
查看这个自动请求限流的工具：https://pypi.python.org/pypi/RequestsThrottler/0.2.2
你可以设置固定延迟时间或者在固定时间内发送一定数量的请求来限制请求速率（本质上是相同的）：
import requests
from requests_throttler import BaseThrottler
request = requests.Request(method='GET', url='http://www.google.com')
reqs = [request for i in range(0, 5)] # An example list of requests
with BaseThrottler(name='base-throttler', delay=1.5) as bt:
throttled_requests = bt.multi_submit(reqs)
multi_submit函数返回ThrottledRequest的列表（请查看文档结尾处的链接）。
然后您可以访问响应：
for tr in throttled_requests:
print tr.response
或者您可以通过指定在固定时间内发送的请求数量（例如每60秒15个请求）来实现相同的效果：
import requests
from requests_throttler import BaseThrottler
request = requests.Request(method='GET', url='http://www.google.com')
reqs = [request for i in range(0, 5)] # An example list of requests
with BaseThrottler(name='base-throttler', reqs_over_time=(15, 60)) as bt:
throttled_requests = bt.multi_submit(reqs)
这两种解决方案都可以在不使用 with 语句的情况下实现：
import requests
from requests_throttler import BaseThrottler
request = requests.Request(method='GET', url='http://www.google.com')
reqs = [request for i in range(0, 5)] # An example list of requests
bt = BaseThrottler(name='base-throttler', delay=1.5)
bt.start()
throttled_requests = bt.multi_submit(reqs)
bt.shutdown()
了解更多详情请访问：http://pythonhosted.org/RequestsThrottler/index.html
from ratelimit import limits
import requests
FIFTEEN_MINUTES = 900
@limits(calls=15, period=FIFTEEN_MINUTES)
def call_api(url):
response = requests.get(url)
if response.status_code != 200:
raise Exception('API response: {}'.format(response.status_code))
return response
def worker():
with rate_limit('slow.domain.com', 2):
response = requests.get('https://slow.domain.com/path')
text = response.text
# Use `text`
(domain, delay)，以便您不会触发速率限制。from contextlib import contextmanager
from gevent.event import Event
from gevent.queue import Queue
from time import time
def rate_limit(resource, delay, _queues={}):
"""Delay use of `resource` until after `delay` seconds have passed.
Example usage:
def worker():
with rate_limit('foo.bar.com', 1):
response = requests.get('https://foo.bar.com/path')
text = response.text
# use `text`
This will serialize and delay requests from multiple workers for resource
'foo.bar.com' by 1 second.
"""
if resource not in _queues:
queue = Queue()
gevent.spawn(_watch, queue)
_queues[resource] = queue
return _resource_manager(_queues[resource], delay)
def _watch(queue):
"Watch `queue` and wake event listeners after delay."
last = 0
while True:
event, delay = queue.get()
now = time()
if (now - last) < delay:
gevent.sleep(delay - (now - last))
event.set() # Wake worker but keep control.
event.clear()
event.wait() # Yield control until woken.
last = time()
@contextmanager
def _resource_manager(queue, delay):
"`with` statement support for `rate_limit`."
event = Event()
queue.put((event, delay))
event.wait() # Wait for queue watcher to wake us.
yield
event.set() # Wake queue watcher.
看起来requests或grequests代码中没有任何处理这个问题的简单机制。唯一似乎可用的挂钩是用于响应。
这里有一个超级hacky的解决方法，至少证明它是可能的 - 我修改了grequests以保持请求发出的时间列表，并睡眠创建AsyncRequest，直到每秒请求数低于最大值。
class AsyncRequest(object):
def __init__(self, method, url, **kwargs):
print self,'init'
waiting=True
while waiting:
if len([x for x in q if x > time.time()-15]) < 8:
q.append(time.time())
waiting=False
else:
print self,'snoozing'
gevent.sleep(1)
import time
import rg
urls = [
'http://www.heroku.com',
'http://python-tablib.org',
'http://httpbin.org',
'http://python-requests.org',
'http://kennethreitz.com',
'http://www.cnn.com',
]
def print_url(r, *args, **kwargs):
print(r.url),time.time()
hook_dict=dict(response=print_url)
rs = (rg.get(u, hooks=hook_dict) for u in urls)
for r in rg.imap(rs):
print r