线程池版的SocketServer

python的SocketServer模块提供了一个开发socket server的捷径。但是其自带的异步handler ThreadingMixIn每次处理请求是都会产生一个新的线程，在负载较重的时候会有一些问题。需要一个使用线程池的TCPServer.

具体的代码如下：

from threading import Thread
from SocketServer import ThreadingMixIn, TCPServer, BaseRequestHandler
from Queue import Queue, Empty

class PooledThreadTCPServer(TCPServer, object):
    def __init__(self, server_address, RequestHandlerClass, thread_num):
        super(PooledThreadTCPServer, self).__init__(server_address, RequestHandlerClass)
        self.__thread_num = thread_num
        self.__queue = Queue()
       
    def __thread_main(self):
        while self.__serving:
            try:
                request, client_address = self.__queue.get(True, 0.5)
            except Empty:
                continue
            self.process_request_thread(request, client_address)
        db.close()
    
    def process_request_thread(self, request, client_address):
        try:
            self.finish_request(request, client_address)
        except:
            self.handle_error(request, client_address)
        finally:
            self.close_request(request)
    
    def process_request(self, request, client_address):
        self.__queue.put((request, client_address))
        
    def serve_forever(self):
        self.__serving = True
        for i in range(self.__thread_num):
            t = Thread(target=self.__thread_main)
            t.setDaemon(True)
            t.start()
        super(PooledThreadTCPServer, self).serve_forever()

继承自TCPServer和object。
继承object的原因是需要使用super调用父类的方法，不继承object就可能报错。
process_request方法在有请求到来的时候被调用。
它将request和客户端地址放到一个队列里。

配合一个RequestHandler，就可以轻松的实现一个线程池版的TCPServer

class EchoRequestHandler(BaseRequestHandler):
    def handle(self):
        data = self.request.recv(1024)
        if data == "quit":
            self.server.shutdown()
        self.request.send(data)

if __name__ == "__main__":
    addr = ("localhost", 12345)
    server = PooledThreadTCPServer(addr, EchoRequestHandler, 10)
    server.serve_forever()

需要注意的是，如果RequestHandler中执行的是CPU密集型的操作，那么线程池
无助于提高性能。python同一时间只能有一个线程在运行，无论你有多少CPU。