2 :mod:`SocketServer` --- A framework for network servers
3 =======================================================
5 .. module:: SocketServer
6 :synopsis: A framework for network servers.
10 The :mod:`SocketServer` module has been renamed to :mod:`socketserver` in
11 Python 3.0. The :term:`2to3` tool will automatically adapt imports when
12 converting your sources to 3.0.
15 The :mod:`SocketServer` module simplifies the task of writing network servers.
17 There are four basic server classes: :class:`TCPServer` uses the Internet TCP
18 protocol, which provides for continuous streams of data between the client and
19 server. :class:`UDPServer` uses datagrams, which are discrete packets of
20 information that may arrive out of order or be lost while in transit. The more
21 infrequently used :class:`UnixStreamServer` and :class:`UnixDatagramServer`
22 classes are similar, but use Unix domain sockets; they're not available on
23 non-Unix platforms. For more details on network programming, consult a book
25 W. Richard Steven's UNIX Network Programming or Ralph Davis's Win32 Network
28 These four classes process requests :dfn:`synchronously`; each request must be
29 completed before the next request can be started. This isn't suitable if each
30 request takes a long time to complete, because it requires a lot of computation,
31 or because it returns a lot of data which the client is slow to process. The
32 solution is to create a separate process or thread to handle each request; the
33 :class:`ForkingMixIn` and :class:`ThreadingMixIn` mix-in classes can be used to
34 support asynchronous behaviour.
36 Creating a server requires several steps. First, you must create a request
37 handler class by subclassing the :class:`BaseRequestHandler` class and
38 overriding its :meth:`handle` method; this method will process incoming
39 requests. Second, you must instantiate one of the server classes, passing it
40 the server's address and the request handler class. Finally, call the
41 :meth:`handle_request` or :meth:`serve_forever` method of the server object to
42 process one or many requests.
44 When inheriting from :class:`ThreadingMixIn` for threaded connection behavior,
45 you should explicitly declare how you want your threads to behave on an abrupt
46 shutdown. The :class:`ThreadingMixIn` class defines an attribute
47 *daemon_threads*, which indicates whether or not the server should wait for
48 thread termination. You should set the flag explicitly if you would like threads
49 to behave autonomously; the default is :const:`False`, meaning that Python will
50 not exit until all threads created by :class:`ThreadingMixIn` have exited.
52 Server classes have the same external methods and attributes, no matter what
53 network protocol they use.
59 There are five classes in an inheritance diagram, four of which represent
60 synchronous servers of four types::
67 +-----------+ +------------------+
68 | TCPServer |------->| UnixStreamServer |
69 +-----------+ +------------------+
72 +-----------+ +--------------------+
73 | UDPServer |------->| UnixDatagramServer |
74 +-----------+ +--------------------+
76 Note that :class:`UnixDatagramServer` derives from :class:`UDPServer`, not from
77 :class:`UnixStreamServer` --- the only difference between an IP and a Unix
78 stream server is the address family, which is simply repeated in both Unix
81 Forking and threading versions of each type of server can be created using the
82 :class:`ForkingMixIn` and :class:`ThreadingMixIn` mix-in classes. For instance,
83 a threading UDP server class is created as follows::
85 class ThreadingUDPServer(ThreadingMixIn, UDPServer): pass
87 The mix-in class must come first, since it overrides a method defined in
88 :class:`UDPServer`. Setting the various member variables also changes the
89 behavior of the underlying server mechanism.
91 To implement a service, you must derive a class from :class:`BaseRequestHandler`
92 and redefine its :meth:`handle` method. You can then run various versions of
93 the service by combining one of the server classes with your request handler
94 class. The request handler class must be different for datagram or stream
95 services. This can be hidden by using the handler subclasses
96 :class:`StreamRequestHandler` or :class:`DatagramRequestHandler`.
98 Of course, you still have to use your head! For instance, it makes no sense to
99 use a forking server if the service contains state in memory that can be
100 modified by different requests, since the modifications in the child process
101 would never reach the initial state kept in the parent process and passed to
102 each child. In this case, you can use a threading server, but you will probably
103 have to use locks to protect the integrity of the shared data.
105 On the other hand, if you are building an HTTP server where all data is stored
106 externally (for instance, in the file system), a synchronous class will
107 essentially render the service "deaf" while one request is being handled --
108 which may be for a very long time if a client is slow to receive all the data it
109 has requested. Here a threading or forking server is appropriate.
111 In some cases, it may be appropriate to process part of a request synchronously,
112 but to finish processing in a forked child depending on the request data. This
113 can be implemented by using a synchronous server and doing an explicit fork in
114 the request handler class :meth:`handle` method.
116 Another approach to handling multiple simultaneous requests in an environment
117 that supports neither threads nor :func:`fork` (or where these are too expensive
118 or inappropriate for the service) is to maintain an explicit table of partially
119 finished requests and to use :func:`select` to decide which request to work on
120 next (or whether to handle a new incoming request). This is particularly
121 important for stream services where each client can potentially be connected for
122 a long time (if threads or subprocesses cannot be used). See :mod:`asyncore` for
123 another way to manage this.
125 .. XXX should data and methods be intermingled, or separate?
126 how should the distinction between class and instance variables be drawn?
133 .. function:: fileno()
135 Return an integer file descriptor for the socket on which the server is
136 listening. This function is most commonly passed to :func:`select.select`, to
137 allow monitoring multiple servers in the same process.
140 .. function:: handle_request()
142 Process a single request. This function calls the following methods in
143 order: :meth:`get_request`, :meth:`verify_request`, and
144 :meth:`process_request`. If the user-provided :meth:`handle` method of the
145 handler class raises an exception, the server's :meth:`handle_error` method
146 will be called. If no request is received within :attr:`self.timeout`
147 seconds, :meth:`handle_timeout` will be called and :meth:`handle_request`
151 .. function:: serve_forever(poll_interval=0.5)
153 Handle requests until an explicit :meth:`shutdown` request. Polls for
154 shutdown every *poll_interval* seconds.
157 .. function:: shutdown()
159 Tells the :meth:`serve_forever` loop to stop and waits until it does.
161 .. versionadded:: 2.6
164 .. data:: address_family
166 The family of protocols to which the server's socket belongs.
167 Common examples are :const:`socket.AF_INET` and :const:`socket.AF_UNIX`.
170 .. data:: RequestHandlerClass
172 The user-provided request handler class; an instance of this class is created
176 .. data:: server_address
178 The address on which the server is listening. The format of addresses varies
179 depending on the protocol family; see the documentation for the socket module
180 for details. For Internet protocols, this is a tuple containing a string giving
181 the address, and an integer port number: ``('127.0.0.1', 80)``, for example.
186 The socket object on which the server will listen for incoming requests.
188 The server classes support the following class variables:
190 .. XXX should class variables be covered before instance variables, or vice versa?
193 .. data:: allow_reuse_address
195 Whether the server will allow the reuse of an address. This defaults to
196 :const:`False`, and can be set in subclasses to change the policy.
199 .. data:: request_queue_size
201 The size of the request queue. If it takes a long time to process a single
202 request, any requests that arrive while the server is busy are placed into a
203 queue, up to :attr:`request_queue_size` requests. Once the queue is full,
204 further requests from clients will get a "Connection denied" error. The default
205 value is usually 5, but this can be overridden by subclasses.
208 .. data:: socket_type
210 The type of socket used by the server; :const:`socket.SOCK_STREAM` and
211 :const:`socket.SOCK_DGRAM` are two common values.
215 Timeout duration, measured in seconds, or :const:`None` if no timeout is
216 desired. If :meth:`handle_request` receives no incoming requests within the
217 timeout period, the :meth:`handle_timeout` method is called.
219 There are various server methods that can be overridden by subclasses of base
220 server classes like :class:`TCPServer`; these methods aren't useful to external
221 users of the server object.
223 .. XXX should the default implementations of these be documented, or should
224 it be assumed that the user will look at SocketServer.py?
227 .. function:: finish_request()
229 Actually processes the request by instantiating :attr:`RequestHandlerClass` and
230 calling its :meth:`handle` method.
233 .. function:: get_request()
235 Must accept a request from the socket, and return a 2-tuple containing the *new*
236 socket object to be used to communicate with the client, and the client's
240 .. function:: handle_error(request, client_address)
242 This function is called if the :attr:`RequestHandlerClass`'s :meth:`handle`
243 method raises an exception. The default action is to print the traceback to
244 standard output and continue handling further requests.
246 .. function:: handle_timeout()
248 This function is called when the :attr:`timeout` attribute has been set to a
249 value other than :const:`None` and the timeout period has passed with no
250 requests being received. The default action for forking servers is
251 to collect the status of any child processes that have exited, while
252 in threading servers this method does nothing.
254 .. function:: process_request(request, client_address)
256 Calls :meth:`finish_request` to create an instance of the
257 :attr:`RequestHandlerClass`. If desired, this function can create a new process
258 or thread to handle the request; the :class:`ForkingMixIn` and
259 :class:`ThreadingMixIn` classes do this.
261 .. Is there any point in documenting the following two functions?
262 What would the purpose of overriding them be: initializing server
263 instance variables, adding new network families?
266 .. function:: server_activate()
268 Called by the server's constructor to activate the server. The default behavior
269 just :meth:`listen`\ s to the server's socket. May be overridden.
272 .. function:: server_bind()
274 Called by the server's constructor to bind the socket to the desired address.
278 .. function:: verify_request(request, client_address)
280 Must return a Boolean value; if the value is :const:`True`, the request will be
281 processed, and if it's :const:`False`, the request will be denied. This function
282 can be overridden to implement access controls for a server. The default
283 implementation always returns :const:`True`.
286 RequestHandler Objects
287 ----------------------
289 The request handler class must define a new :meth:`handle` method, and can
290 override any of the following methods. A new instance is created for each
294 .. function:: finish()
296 Called after the :meth:`handle` method to perform any clean-up actions
297 required. The default implementation does nothing. If :meth:`setup` or
298 :meth:`handle` raise an exception, this function will not be called.
301 .. function:: handle()
303 This function must do all the work required to service a request. The
304 default implementation does nothing. Several instance attributes are
305 available to it; the request is available as :attr:`self.request`; the client
306 address as :attr:`self.client_address`; and the server instance as
307 :attr:`self.server`, in case it needs access to per-server information.
309 The type of :attr:`self.request` is different for datagram or stream
310 services. For stream services, :attr:`self.request` is a socket object; for
311 datagram services, :attr:`self.request` is a pair of string and socket.
312 However, this can be hidden by using the request handler subclasses
313 :class:`StreamRequestHandler` or :class:`DatagramRequestHandler`, which
314 override the :meth:`setup` and :meth:`finish` methods, and provide
315 :attr:`self.rfile` and :attr:`self.wfile` attributes. :attr:`self.rfile` and
316 :attr:`self.wfile` can be read or written, respectively, to get the request
317 data or return data to the client.
320 .. function:: setup()
322 Called before the :meth:`handle` method to perform any initialization actions
323 required. The default implementation does nothing.
329 :class:`SocketServer.TCPServer` Example
330 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
332 This is the server side::
336 class MyTCPHandler(SocketServer.BaseRequestHandler):
338 The RequestHandler class for our server.
340 It is instantiated once per connection to the server, and must
341 override the handle() method to implement communication to the
346 # self.request is the TCP socket connected to the client
347 self.data = self.request.recv(1024).strip()
348 print "%s wrote:" % self.client_address[0]
350 # just send back the same data, but upper-cased
351 self.request.send(self.data.upper())
353 if __name__ == "__main__":
354 HOST, PORT = "localhost", 9999
356 # Create the server, binding to localhost on port 9999
357 server = SocketServer.TCPServer((HOST, PORT), MyTCPHandler)
359 # Activate the server; this will keep running until you
360 # interrupt the program with Ctrl-C
361 server.serve_forever()
363 An alternative request handler class that makes use of streams (file-like
364 objects that simplify communication by providing the standard file interface)::
366 class MyTCPHandler(SocketServer.StreamRequestHandler):
369 # self.rfile is a file-like object created by the handler;
370 # we can now use e.g. readline() instead of raw recv() calls
371 self.data = self.rfile.readline().strip()
372 print "%s wrote:" % self.client_address[0]
374 # Likewise, self.wfile is a file-like object used to write back
376 self.wfile.write(self.data.upper())
378 The difference is that the ``readline()`` call in the second handler will call
379 ``recv()`` multiple times until it encounters a newline character, while the
380 single ``recv()`` call in the first handler will just return what has been sent
381 from the client in one ``send()`` call.
384 This is the client side::
389 HOST, PORT = "localhost", 9999
390 data = " ".join(sys.argv[1:])
392 # Create a socket (SOCK_STREAM means a TCP socket)
393 sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
395 # Connect to server and send data
396 sock.connect((HOST, PORT))
397 sock.send(data + "\n")
399 # Receive data from the server and shut down
400 received = sock.recv(1024)
403 print "Sent: %s" % data
404 print "Received: %s" % received
407 The output of the example should look something like this:
411 $ python TCPServer.py
419 $ python TCPClient.py hello world with TCP
420 Sent: hello world with TCP
421 Received: HELLO WORLD WITH TCP
422 $ python TCPClient.py python is nice
424 Received: PYTHON IS NICE
427 :class:`SocketServer.UDPServer` Example
428 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
430 This is the server side::
434 class MyUDPHandler(SocketServer.BaseRequestHandler):
436 This class works similar to the TCP handler class, except that
437 self.request consists of a pair of data and client socket, and since
438 there is no connection the client address must be given explicitly
439 when sending data back via sendto().
443 data = self.request[0].strip()
444 socket = self.request[1]
445 print "%s wrote:" % self.client_address[0]
447 socket.sendto(data.upper(), self.client_address)
449 if __name__ == "__main__":
450 HOST, PORT = "localhost", 9999
451 server = SocketServer.UDPServer((HOST, PORT), MyUDPHandler)
452 server.serve_forever()
454 This is the client side::
459 HOST, PORT = "localhost"
460 data = " ".join(sys.argv[1:])
462 # SOCK_DGRAM is the socket type to use for UDP sockets
463 sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
465 # As you can see, there is no connect() call; UDP has no connections.
466 # Instead, data is directly sent to the recipient via sendto().
467 sock.sendto(data + "\n", (HOST, PORT))
468 received = sock.recv(1024)
470 print "Sent: %s" % data
471 print "Received: %s" % received
473 The output of the example should look exactly like for the TCP server example.
479 To build asynchronous handlers, use the :class:`ThreadingMixIn` and
480 :class:`ForkingMixIn` classes.
482 An example for the :class:`ThreadingMixIn` class::
488 class ThreadedTCPRequestHandler(SocketServer.BaseRequestHandler):
491 data = self.request.recv(1024)
492 cur_thread = threading.currentThread()
493 response = "%s: %s" % (cur_thread.getName(), data)
494 self.request.send(response)
496 class ThreadedTCPServer(SocketServer.ThreadingMixIn, SocketServer.TCPServer):
499 def client(ip, port, message):
500 sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
501 sock.connect((ip, port))
503 response = sock.recv(1024)
504 print "Received: %s" % response
507 if __name__ == "__main__":
508 # Port 0 means to select an arbitrary unused port
509 HOST, PORT = "localhost", 0
511 server = ThreadedTCPServer((HOST, PORT), ThreadedTCPRequestHandler)
512 ip, port = server.server_address
514 # Start a thread with the server -- that thread will then start one
515 # more thread for each request
516 server_thread = threading.Thread(target=server.serve_forever)
517 # Exit the server thread when the main thread terminates
518 server_thread.setDaemon(True)
519 server_thread.start()
520 print "Server loop running in thread:", server_thread.getName()
522 client(ip, port, "Hello World 1")
523 client(ip, port, "Hello World 2")
524 client(ip, port, "Hello World 3")
528 The output of the example should look something like this::
530 $ python ThreadedTCPServer.py
531 Server loop running in thread: Thread-1
532 Received: Thread-2: Hello World 1
533 Received: Thread-3: Hello World 2
534 Received: Thread-4: Hello World 3
537 The :class:`ForkingMixIn` class is used in the same way, except that the server
538 will spawn a new process for each request.