scrapy源码6：deffer和parallel的源码分析

defer

scrapy的核心的代码大量用到deffer对象，还有一些并行的东西。 这里简单去学习下deffer和并行的方法知识 。

twisted.interet.defer 这个官方的api就是下面的网址了。
http://twistedmatrix.com/documents/current/api/twisted.internet.defer.html
我们可以看到这个defer是twisted提供的internet方法的包

twisted

先了解下twisted是个什么。
Twisted: The Framework Of Your Internet. 说明twisted是一个网络框架。
http://twistedmatrix.com/documents/current/api/twisted.html 这个官方api说的很精简啊，

我们还是看看wiki上怎么说的: Twisted is an event-driven network programming framework written in Python and licensed under the MIT
License.
Twisted projects variously support TCP, UDP, SSL/TLS, IP multicast, Unix domain sockets, a large number of protocols (including HTTP, XMPP, NNTP, IMAP, SSH, IRC, FTP, and others), and much more. Twisted is based on the event-driven programming paradigm, which means that users of Twisted write short callbacks which are called by the framework.
大概意思如下：
Twisted是一个以Python编写的事件驱动的网络编程框架，并根据MIT许可证进行许可。
twisted的项目各种支持TCP，UDP，SSL / TLS，IP组播，Unix域套接字，大量协议
（包括HTTP，XMPP，NNTP，IMAP，SSH，IRC，FTP等）等等。
Twisted是基于事件驱动的编程范例，这意味着Twisted的用户写入框架调用的简短回调。

知道twisted了， 还是应该了解下twisted的异步处理。
网址如下: http://blog.csdn.net/fxjtoday/article/details/6396932
这个人的博客写的不错。 http://blog.csdn.net/fangjian1204/article/details/42084273

Method addCallbacks Add a pair of callbacks (success and error) to this Deferred.
Method addCallback Convenience method for adding just a callback.
Method addErrback Convenience method for adding just an errback.
Method addBoth Convenience method for adding a single callable as both a callback and an errback.

这里有4个方法， addCallbacks,addBoth 一次添加成功和错误的2个回调。addCallback 添加成功的回调 ， addErrback 添加错误的回调。
Method addTimeout Time out this Deferred by scheduling it to be cancelled after timeout seconds.
这个是设置下deferred的时间，超时的话会被取消的。
Method chainDeferred Chain another Deferred to this Deferred.

Method callback Run all success callbacks that have been added to this Deferred.
运行成功的回调
Method errback Run all error callbacks that have been added to this Deferred.
运行所有的错误回调
Method pause Stop processing on a Deferred until unpause() is called.
暂定deferred执行， 一直到unpause调用才继续执行
Method unpause Process all callbacks made since pause() was called.
解除defered的pause的暂停操作
Method cancel Cancel this Deferred.
取消这个defferd
Method str Return a string representation of this Deferred.
Method iter Undocumented
Method send Undocumented
Method asFuture Adapt a Deferred into a asyncio.Future which is bound to loop.
Class Method fromFuture Adapt an asyncio.Future to a Deferred.

 
# 我们在defer.py里面有个方法如下 
def parallel(iterable, count, callable, *args, **named):
    """Execute a callable over the objects in the given iterable, in parallel,
    using no more than ``count`` concurrent calls.

    Taken from: http://jcalderone.livejournal.com/24285.html
    """
    coop = task.Cooperator()
    work = (callable(elem, *args, **named) for elem in iterable)
    return defer.DeferredList([coop.coiterate(work) for _ in range(count)])

# 先来看看这个task.Cooperator 是什么
# Cooperator的官方api 网址： http://twistedmatrix.com/documents/current/api/twisted.internet.task.Cooperator.html
# google翻译了下如下： 
# 合作任务调度程序。
# 合作任务是一个迭代器，每个迭代代表一个原子工作单元。当迭代器产生时，它允许合作者决定下一个执行其任务。如果迭代器产生defer.Deferred，
# 则工作将暂停，直到defer.Deferred触发并完成其回调链。
# 当合作者有多个任务时，它会在所有任务之间分配工作。
# 有两种方法可以向协作者添加任务，进行合作和协调。合作是更有用的两个，因为它返回一个CooperativeTask，可以暂停，恢复和等待。 
# coiterate具有相同的效果，但只返回一个延迟。当任务完成时触发。
# 合作者可以用于许多事情，包括但不限于：

#     运行一个或多个计算密集型任务而不阻止
#     通过同时运行总任务的子集来限制并行性
#     做一件事，等待延期开火，做下一件事，重复（即序列化一系列异步任务）

# 多个合作者不能相互合作，所以在大多数情况下，您应该使用全局的合作伙伴。

# Method 	__init__ 	Create a scheduler-like object to which iterators may be added.
#     创建可以添加迭代器的类似于调度程序的对象。
# Method 	coiterate 	Add an iterator to the list of iterators this Cooperator is #currently running.
#     将迭代器添加到此Cooperator当前正在运行的迭代器列表中。
# Method 	cooperate 	Start running the given iterator as a long-running cooperative task, by calling next() on it as a periodic timed event.
#     开始运行给定的迭代器作为一个长期运行的协作任务，通过调用next（）作为定期定时事件。
# Method 	start 	Begin scheduling steps.
#     开始安排步骤
# Method 	stop 	Stop scheduling steps. Errback the completion Deferreds of all iterators which have been added and forget about them.
#     停止调度步骤。 Errback完成所有迭代器已被添加并忘记了它们。
# Method 	running 	Is this Cooperator is currently running?
#     这个合作伙伴是否正在运行？

# 回到那个代码吧。 第一句    coop = task.Cooperator() 创建一个合作者对象。
#  work = (callable(elem, *args, **named) for elem in iterable) 遍历下iterable方法列表， 构造works元组
# coop.coiterate(work) 每一个work都去调用这个方法添加到coop中去， 返回一个deferred数组， 然后转化为defer.DeferredList

def process_chain(callbacks, input, *a, **kw):
    """Return a Deferred built by chaining the given callbacks"""
    d = defer.Deferred()
    for x in callbacks:
        d.addCallback(x, *a, **kw)
    d.callback(input)
    return d

# 创建一个deferred对象， 添加一系列的成功回调方法， 然后执行所有的成功回调。 返回deffered对象。 


def process_chain_both(callbacks, errbacks, input, *a, **kw):
    """Return a Deferred built by chaining the given callbacks and errbacks"""
    d = defer.Deferred()
    for cb, eb in zip(callbacks, errbacks):
        d.addCallbacks(cb, eb, callbackArgs=a, callbackKeywords=kw,
            errbackArgs=a, errbackKeywords=kw)
    if isinstance(input, failure.Failure):
        d.errback(input)
    else:
        d.callback(input)
    return d

# 这个方法， 创建一个deferred对象， 添加成功和失败回调对。 
# 如果input是成功的就调用成功回调， 失败调用失败的回调。 

def process_parallel(callbacks, input, *a, **kw):
    """Return a Deferred with the output of all successful calls to the given
    callbacks
    """
    dfds = [defer.succeed(input).addCallback(x, *a, **kw) for x in callbacks]
    d = defer.DeferredList(dfds, fireOnOneErrback=1, consumeErrors=1)
    d.addCallbacks(lambda r: [x[1] for x in r], lambda f: f.value.subFailure)
    return d
# 遍历 callbacks列表， 使用addcallback添加成功的回调。 
# defer.succeed(input) 应该是返回一个deferred对象。 
# 查看下官方api。 
# Function 	succeed 	Return a Deferred that has already had .callback(result) called.
# 现获取一个dfds 的数组， 然后转为list ， 再有2个参数。
# 添加2个匿名的回调方法。