在之前的文章中已经讲过了gevent的使用、gevent的底层greenlet的使用以及gevent调度的源码分析,可以阅读文章回顾一下:python之gevent(1),python之greenlet,python之gevent(2)。本文将带大家一起学习几个gevent比较重要的模块,包括Timeout,Event/AsynResult,Semphore,socket patch,这些模块都涉及当前协程与hub的切换。
Timeout
这个类在gevent.timeout模块,其作用是超时后在当前协程抛出异常,这样执行流程也强制回到了当前协程。看一个简单的例子:
import time
import gevent
from gevent.timeout import Timeout
SLEEP = 3
TIMEOUT = 2
timeout = Timeout(TIMEOUT)
timeout.start()
def wait():
gevent.sleep(SLEEP)
print('log in wait')
begin = time.time()
try:
gevent.spawn(wait).join()
except Timeout:
print('after %s catch Timeout Exception' % (time.time() - begin))
finally:
timeout.cancel()
运行这段代码,输出如下:
可以看出,在2s之后在main协程抛出了Timeout异常(继承自BaseException)。Timeout的实现的核心在start函数:
def start(self):
"""Schedule the timeout."""
if self.pending:
raise AssertionError('%r is already started; to restart it, cancel it first' % self)
if self.seconds is None:
# "fake" timeout (never expires)
return
if self.exception is None or self.exception is False or isinstance(self.exception, string_types):
# timeout that raises self
throws = self
else:
# regular timeout with user-provided exception
throws = self.exception
# Make sure the timer updates the current time so that we don't
# expire prematurely.
self.timer.start(getcurrent().throw, throws, update=True)
从代码中可以看到,在超时之后调用了getcurrent().throw(),throw方法会切换协程,并抛出异常(在上面的代码中默认抛出Timeout异常)。使用Timeout有几点需要注意:
第一:一定要记得在finally调用cancel,否则如果协程先于TIMEOUT时间恢复,之后还会抛出异常,例如下面的代码:
import gevent
from gevent import Timeout
SLEEP = 4
TIMEOUT = 5
timeout = Timeout(TIMEOUT)
timeout.start()
def wait():
gevent.sleep(SLEEP)
print('log in wait')
begin = time.time()
try:
gevent.spawn(wait).join()
except Timeout:
print('after %s catch Timeout Exception' % (time.time() - begin))
# finally:
# timeout.cancel()
gevent.sleep(2)
print 'program will finish'
#协程先于超时恢复
上述的代码运行会抛出Timeout异常,在这个例子中,协程先于超时恢复(SLEEP < TIMEOUT),且没有在finally中调用Timeout.cancel。最后的两行保证程序不要过早结束退出,那么在hub调度的时候会重新抛出异常。
由于Timeout实现了with协议(enter和exit方法),更好的写法是将TImeout写在with语句中,如下面的代码:
import gevent
from gevent import Timeout
SLEEP = 4
TIMEOUT = 5
def wait():
gevent.sleep(SLEEP)
print('log in wait')
with Timeout(TIMEOUT):
begin = time.time()
try:
gevent.spawn(wait).join()
except Timeout:
print('after %s catch Timeout Exception' % (time.time() - begin))
gevent.sleep(2)
print 'program will finish'
#Timeout with
第二:Timeout只是切换到当前协程,并不会取消已经注册的协程(上面通过spawn发起的协程),我们改改代码:
import gevent
from gevent import Timeout
SLEEP = 6
TIMEOUT = 5
timeout = Timeout(TIMEOUT)
timeout.start()
def wait():
gevent.sleep(SLEEP)
print('log in wait')
begin = time.time()
try:
gevent.spawn(wait).join()
except Timeout:
print('after %s catch Timeout Exception' % (time.time() - begin))
finally:
timeout.cancel()
gevent.sleep(2)
print 'program will finish'
# output:
# after 5.00100016594 catch Timeout Exception
# log in wait
# program will finish
#Timeout不影响发起的协程
从输出可以看到,即使因为超时切回了main greenlet,但spawn发起的协程并不受影响。如果希望超时取消之前发起的协程,那么可以在捕获到异常之后调用 Greenlet.kill
第三:gevent对可能导致当前协程挂起的函数都提供了timeout参数,用于在指定时间到达之后恢复被挂起的协程。在函数内部会捕获Timeout异常,并不会抛出。例如:
SLEEP = 6
TIMEOUT = 5
def wait():
gevent.sleep(SLEEP)
print('log in wait')
begin = time.time()
try:
gevent.spawn(wait).join(TIMEOUT)
except Timeout:
print('after %s catch Timeout Exception' % (time.time() - begin))
print 'program will exit', time.time() - begin
#函数的timeout参数
Event & AsyncResult
Event用来在Greenlet之间同步,tutorial上的例子简单明了:
import gevent
from gevent.event import Event
'''
Illustrates the use of events
'''
evt = Event()
def setter():
'''After 3 seconds, wake all threads waiting on the value of evt'''
print('A: Hey wait for me, I have to do something')
gevent.sleep(3)
print("Ok, I'm done")
evt.set()
def waiter():
'''After 3 seconds the get call will unblock'''
print("I'll wait for you")
evt.wait() # blocking
print("It's about time")
def main():
gevent.joinall([
gevent.spawn(setter),
gevent.spawn(waiter),
gevent.spawn(waiter),
])
if __name__ == '__main__': main()
Event Example
Event主要的两个方法是set和wait:wait等待事件发生,如果事件未发生那么挂起该协程;set通知事件发生,然后hub会唤醒所有wait在该事件的协程。从输出可知, 一次event触发可以唤醒所有在该event上等待的协程。AsyncResult同Event类似,只不过可以在协程唤醒的时候传值(有点类似generator的next send的区别)。接下来大致看看Event的set和wait方法。
Event.wait的核心代码在gevent.event._AbstractLinkable._wait_core,其中_AbstractLinkable是Event的基类。_wait_core源码如下:
def _wait_core(self, timeout, catch=Timeout):
# The core of the wait implementation, handling
# switching and linking. If *catch* is set to (),
# a timeout that elapses will be allowed to be raised.
# Returns a true value if the wait succeeded without timing out.
switch = getcurrent().switch
self.rawlink(switch)
try:
timer = Timeout._start_new_or_dummy(timeout)
try:
try:
result = self.hub.switch()
if result is not self: # pragma: no cover
raise InvalidSwitchError('Invalid switch into Event.wait(): %r' % (result, ))
return True
except catch as ex:
if ex is not timer:
raise
# test_set_and_clear and test_timeout in test_threading
# rely on the exact return values, not just truthish-ness
return False
finally:
timer.cancel()
finally:
self.unlink(switch)
首先是将当前协程的switch加入到Event的callback列表,然后切换到hub。
接下来是set函数:
def set(self):
self._flag = True # make event ready
self._check_and_notify()
def _check_and_notify(self):
# If this object is ready to be notified, begin the process.
if self.ready():
if self._links and not self._notifier:
self._notifier = self.hub.loop.run_callback(self._notify_links)
_check_and_notify函数通知hub调用_notify_links, 在这个函数中将调用Event的callback列表(记录的是之前各个协程的switch函数),这样就恢复了所有wait的协程。
Semaphore & Lock
Semaphore是gevent提供的信号量,实例化为Semaphore(value), value代表了可以并发的量。当value为1,就变成了互斥锁(Lock)。Semaphore提供了两个函数,acquire(P操作)和release(V操作)。当acquire操作导致资源数量将为0之后,就会在当前协程wait,源代码如下(gevent._semaphore.Semaphore.acquire):
def acquire(self, blocking=True, timeout=None):
if self.counter > 0:
self.counter -= 1
return True
if not blocking:
return False
timeout = self._do_wait(timeout)
if timeout is not None:
# Our timer expired.
return False
# Neither our timer no another one expired, so we blocked until
# awoke. Therefore, the counter is ours
self.counter -= 1
assert self.counter >= 0
return True
逻辑比较简单,如果counter数量大于0,那么表示可并发。否则进入wait,_do_wait的实现与Event.wait十分类似,都是记录当前协程的switch,并切换到hub。当资源足够切换回到当前协程,此时counter一定是大于0的。由于协程的并发并不等同于线程的并发,在任意时刻,一个线程内只可能有一个协程在调度,所以上面对counter的操作也不用加锁。
Monkey-Patch
对于python这种动态语言,在运行时替换模块、类、实例的属性都是非常容易的。我们以patch_socket为例:
可见在patch前后,同一个名字(socket)所指向的对象是不一样的。在python2.x环境下,patch后的socket源码在gevent._socket2.py,如果是python3.x,那么对应的源码在gevent._socket3.py.。至于为什么patch之后就让原生的socket操作可以在协程之间协作,看两个函数socket.init 和 socket.recv就明白了。
init函数(gevent._socket2.socket.init):
def __init__(self, family=AF_INET, type=SOCK_STREAM, proto=0, _sock=None):
if _sock is None:
self._sock = _realsocket(family, type, proto) # 原生的socket
self.timeout = _socket.getdefaulttimeout()
else:
if hasattr(_sock, '_sock'):
self._sock = _sock._sock
self.timeout = getattr(_sock, 'timeout', False)
if self.timeout is False:
self.timeout = _socket.getdefaulttimeout()
else:
self._sock = _sock
self.timeout = _socket.getdefaulttimeout()
if PYPY:
self._sock._reuse()
self._sock.setblocking(0) #设置成非阻塞
fileno = self._sock.fileno()
self.hub = get_hub() # hub
io = self.hub.loop.io
self._read_event = io(fileno, 1) # 监听事件
self._write_event = io(fileno, 2)
从init函数可以看到,patch后的socket还是会维护原生的socket对象,并且将原生的socket设置成非阻塞,当一个socket是非阻塞时,如果读写数据没有准备好,那么会抛出EWOULDBLOCK\EAGIN异常。最后两行注册socket的可读和可写事件。再来看看recv函数(gevent._socket2.socket.recv):
def recv(self, *args):
sock = self._sock # keeping the reference so that fd is not closed during waiting
while True:
try:
return sock.recv(*args) # 如果数据准备好了,直接返回
except error as ex:
if ex.args[0] != EWOULDBLOCK or self.timeout == 0.0:
raise
# QQQ without clearing exc_info test__refcount.test_clean_exit fails
sys.exc_clear()
self._wait(self._read_event) # 等待数据可读的watcher
如果在while循环中读到了数据,那么直接返回。但实际很大概率数据并没有准备好,对于非阻塞socket,抛出EWOULDBLOCK异常。在最后一行,调用wait,注册当前协程switch,并切换到hub,当read_event触发时,表示socket可读,这个时候就会切回当前协程,进入下一次while循环。