Python在一个单独的过程中运行。我可以统一包装函数

时间:2014-10-19 18:16:32

标签: python multiprocessing wrapper

我是Python的新手,我想知道如何在语法上有效地实现以下问题。 我有函数f1,f2 ... fN这些函数是生成新进程的包装器(带有目标_f1,_f2,.. _ fN), 将其参数(arg1,arg2,...)传递给子进程并接收返回值
使用这样的代码,我希望模块功能在不同的进程中执行,然后调用者(模块的用户)进程 函数f1,f2,... fN(分别为_f1,f2,... _fN)可能有不同的原型。

in a module

def _f1(arg1, arg2, ... argn,  connection):
    ...
    connection.send(return_value)
    connection.close()
def f1(arg1, arg2, ... argn):
    parent_conn, child_conn = Pipe()
    p = Process(target=_f1, args=(arg1, arg2, ... argn, child_conn))
    p.start()
    p.join() 
    return parent_conn.recv()


def _f2(arg1, arg2, ... argm,  connection):
    ...
    connection.send(return_value)
    connection.close()    
def f2(arg1, arg2, ... argn):
    parent_conn, child_conn = Pipe()
    p = Process(target=_f2, args=(arg1, arg2, ... argm, child_conn))
    p.start()
    p.join() 
    return parent_conn.recv()

...

def _fn(arg1, arg2, ... argk,  connection):
    ...
    connection.send(return_value)
    connection.close()    
def fN(arg1, arg2, ... argn):
    parent_conn, child_conn = Pipe()
    p = Process(target=_fN, args=(arg1, arg2, ... argk, child_conn))
    p.start()
    p.join() 
    return parent_conn.recv()

很明显,包装函数f1,f2,fN大致相同。我可以将它们实现为单个包装函数吗? 我希望执行不会阻塞。例如,模块的用户应该能够同时执行f1和f2。

我希望我已经设法解释了我的问题。

这里有两个函数sum()和sin()的具体例子:

def _sum(a, b,  connection):
   return_value=a+b
   connection.send(return_value)
   connection.close()
def sum(a, b):
   parent_conn, child_conn = Pipe()
   p = Process(target=_sum, args=(a, b, child_conn))
   p.start()
   p.join() 
   return parent_conn.recv()

def _sin(x,  connection):
   return_value=sin(x)
   connection.send(return_value)
   connection.close()    
def sin(x):
   parent_conn, child_conn = Pipe()
   p = Process(target=_sin, args=(x, child_conn))
   p.start()
   p.join() 
   return parent_conn.recv() 

采取srj关于使用装饰的想法我来到下面发布的解决方案。 我试图进一步扩展它以装饰connection.send(return_value)和connection.close(),但它对我不起作用。代码下方。通过评论,我指定了什么是有效的,以及(我认为)不起作用的等价物。有什么帮助吗?

from multiprocessing import Process, Pipe

def process_wrapper1(func):
    def wrapper(*args):
        parent_conn, child_conn = Pipe()
        f_args = args + (child_conn,)
        p = Process(target=func, args=f_args)
        p.start()
        p.join() 
        return parent_conn.recv()
    return wrapper

def process_wrapper2(func):
    def wrapper(*args):
        res=func(*args[0:len(args)-1])
        args[-1].send(res)
        args[-1].close()
    return wrapper



#def _sum(a, b,  connection):            #Working 
#   return_value=a+b
#   connection.send(return_value)
#   connection.close()
def __sum(a, b):                       #Doesn't work, see the error bellow
    return(a+b)    
_sum=process_wrapper2(__sum)

sum=process_wrapper1(_sum) 

Pyzo ipython shell中的上述代码生成以下结果:

In [3]: import test1
In [4]: test1.sum(2,3)
---------------------------------------------------------------------------
PicklingError                             Traceback (most recent call last)
<ipython-input-4-8c542dc5e11a> in <module>()
----> 1 test1.sum(2,3)

C:\projects\PYnGUInLib\test1.py in wrapper(*args)
     11         f_args = (child_conn,) + args
     12         p = Process(target=func, args=f_args)
---> 13         p.start()
     14         p.join()
     15         return parent_conn.recv()

C:\pyzo2014a_64b\lib\multiprocessing\process.py in start(self)
    103                'daemonic processes are not allowed to have children'
    104         _cleanup()
--> 105         self._popen = self._Popen(self)
    106         self._sentinel = self._popen.sentinel
    107         _children.add(self)

C:\pyzo2014a_64b\lib\multiprocessing\context.py in _Popen(process_obj)
    210     @staticmethod
    211     def _Popen(process_obj):
--> 212         return _default_context.get_context().Process._Popen(process_obj)
    213 
    214 class DefaultContext(BaseContext):

C:\pyzo2014a_64b\lib\multiprocessing\context.py in _Popen(process_obj)
    311         def _Popen(process_obj):
    312             from .popen_spawn_win32 import Popen
--> 313             return Popen(process_obj)
    314 
    315     class SpawnContext(BaseContext):

C:\pyzo2014a_64b\lib\multiprocessing\popen_spawn_win32.py in __init__(self, process_obj)
     64             try:
     65                 reduction.dump(prep_data, to_child)
---> 66                 reduction.dump(process_obj, to_child)
     67             finally:
     68                 context.set_spawning_popen(None)

C:\pyzo2014a_64b\lib\multiprocessing\reduction.py in dump(obj, file, protocol)
     57 def dump(obj, file, protocol=None):
     58     '''Replacement for pickle.dump() using ForkingPickler.'''
---> 59     ForkingPickler(file, protocol).dump(obj)
     60 
     61 #

PicklingError: Can't pickle <function process_wrapper2.<locals>.wrapper at 0x0000000005541048>: attribute lookup wrapper on test1 failed
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "C:\pyzo2014a_64b\lib\multiprocessing\spawn.py", line 106, in spawn_main
   exitcode = _main(fd)
  File "C:\pyzo2014a_64b\lib\multiprocessing\spawn.py", line 116, in _main
   self = pickle.load(from_parent)
EOFError: Ran out of input

In [5]: 

2 个答案:

答案 0 :(得分:1)

您可以使用decorator将函数包装为创建流程并执行它的样板。

def process_wrapper(func):
    def wrapper(*args):
        parent_conn, child_conn = Pipe()
        #attach the connection to the arguments
        f_args = args + (child_conn,)
        p = Process(target=func, args=f_args)
        p.start()
        p.join() 
        return parent_conn.recv()
    return wrapper

并将函数定义为

@process_wrapper
def _f2(arg1, arg2, ... argm,  connection):
    ...
    connection.send(return_value)
    connection.close()

说明 process_wrapper函数接受一个具有N个位置参数的函数,其中最后一个始终是管道连接。它返回一个带有N-1个参数的函数,并在其中预填充连接。

如果你的具体功能,

@process_wrapper
def sin(x,  connection):
   return_value=sin(x)
   connection.send(return_value)
   connection.close()  

@process_wrapper
def sum(a, b,  connection):
   return_value=a+b
   connection.send(return_value)
   connection.close()

您可以将该功能称为

sum(a,b)

对python装饰器的更多引用 http://www.jeffknupp.com/blog/2013/11/29/improve-your-python-decorators-explained/

答案 1 :(得分:0)

您应该使用multiprocessing.Pool。这是一个例子:

def f1(*args):
    rv = do_calculations()
    return rv 

def f2(*args):
    ...

...
def fN(*args):
    ...

def worker(args):
    fn = args[0]
    return fn(*args[1:])

inputs = [
    [f1, f1_args],
    [f2, f2_args],
    ...
    [fN, fN_args]
]

pool = multiprocessing.Pool(processes=multiprocessing.cpu_count())
results = pool.map(worker, inputs)