1.. _socket-howto: 2 3**************************** 4 Socket Programming HOWTO 5**************************** 6 7:Author: Gordon McMillan 8 9 10.. topic:: Abstract 11 12 Sockets are used nearly everywhere, but are one of the most severely 13 misunderstood technologies around. This is a 10,000 foot overview of sockets. 14 It's not really a tutorial - you'll still have work to do in getting things 15 operational. It doesn't cover the fine points (and there are a lot of them), but 16 I hope it will give you enough background to begin using them decently. 17 18 19Sockets 20======= 21 22I'm only going to talk about INET (i.e. IPv4) sockets, but they account for at least 99% of 23the sockets in use. And I'll only talk about STREAM (i.e. TCP) sockets - unless you really 24know what you're doing (in which case this HOWTO isn't for you!), you'll get 25better behavior and performance from a STREAM socket than anything else. I will 26try to clear up the mystery of what a socket is, as well as some hints on how to 27work with blocking and non-blocking sockets. But I'll start by talking about 28blocking sockets. You'll need to know how they work before dealing with 29non-blocking sockets. 30 31Part of the trouble with understanding these things is that "socket" can mean a 32number of subtly different things, depending on context. So first, let's make a 33distinction between a "client" socket - an endpoint of a conversation, and a 34"server" socket, which is more like a switchboard operator. The client 35application (your browser, for example) uses "client" sockets exclusively; the 36web server it's talking to uses both "server" sockets and "client" sockets. 37 38 39History 40------- 41 42Of the various forms of :abbr:`IPC (Inter Process Communication)`, 43sockets are by far the most popular. On any given platform, there are 44likely to be other forms of IPC that are faster, but for 45cross-platform communication, sockets are about the only game in town. 46 47They were invented in Berkeley as part of the BSD flavor of Unix. They spread 48like wildfire with the internet. With good reason --- the combination of sockets 49with INET makes talking to arbitrary machines around the world unbelievably easy 50(at least compared to other schemes). 51 52 53Creating a Socket 54================= 55 56Roughly speaking, when you clicked on the link that brought you to this page, 57your browser did something like the following:: 58 59 # create an INET, STREAMing socket 60 s = socket.socket(socket.AF_INET, socket.SOCK_STREAM) 61 # now connect to the web server on port 80 - the normal http port 62 s.connect(("www.python.org", 80)) 63 64When the ``connect`` completes, the socket ``s`` can be used to send 65in a request for the text of the page. The same socket will read the 66reply, and then be destroyed. That's right, destroyed. Client sockets 67are normally only used for one exchange (or a small set of sequential 68exchanges). 69 70What happens in the web server is a bit more complex. First, the web server 71creates a "server socket":: 72 73 # create an INET, STREAMing socket 74 serversocket = socket.socket(socket.AF_INET, socket.SOCK_STREAM) 75 # bind the socket to a public host, and a well-known port 76 serversocket.bind((socket.gethostname(), 80)) 77 # become a server socket 78 serversocket.listen(5) 79 80A couple things to notice: we used ``socket.gethostname()`` so that the socket 81would be visible to the outside world. If we had used ``s.bind(('localhost', 8280))`` or ``s.bind(('127.0.0.1', 80))`` we would still have a "server" socket, 83but one that was only visible within the same machine. ``s.bind(('', 80))`` 84specifies that the socket is reachable by any address the machine happens to 85have. 86 87A second thing to note: low number ports are usually reserved for "well known" 88services (HTTP, SNMP etc). If you're playing around, use a nice high number (4 89digits). 90 91Finally, the argument to ``listen`` tells the socket library that we want it to 92queue up as many as 5 connect requests (the normal max) before refusing outside 93connections. If the rest of the code is written properly, that should be plenty. 94 95Now that we have a "server" socket, listening on port 80, we can enter the 96mainloop of the web server:: 97 98 while True: 99 # accept connections from outside 100 (clientsocket, address) = serversocket.accept() 101 # now do something with the clientsocket 102 # in this case, we'll pretend this is a threaded server 103 ct = client_thread(clientsocket) 104 ct.run() 105 106There's actually 3 general ways in which this loop could work - dispatching a 107thread to handle ``clientsocket``, create a new process to handle 108``clientsocket``, or restructure this app to use non-blocking sockets, and 109multiplex between our "server" socket and any active ``clientsocket``\ s using 110``select``. More about that later. The important thing to understand now is 111this: this is *all* a "server" socket does. It doesn't send any data. It doesn't 112receive any data. It just produces "client" sockets. Each ``clientsocket`` is 113created in response to some *other* "client" socket doing a ``connect()`` to the 114host and port we're bound to. As soon as we've created that ``clientsocket``, we 115go back to listening for more connections. The two "clients" are free to chat it 116up - they are using some dynamically allocated port which will be recycled when 117the conversation ends. 118 119 120IPC 121--- 122 123If you need fast IPC between two processes on one machine, you should look into 124pipes or shared memory. If you do decide to use AF_INET sockets, bind the 125"server" socket to ``'localhost'``. On most platforms, this will take a 126shortcut around a couple of layers of network code and be quite a bit faster. 127 128.. seealso:: 129 The :mod:`multiprocessing` integrates cross-platform IPC into a higher-level 130 API. 131 132 133Using a Socket 134============== 135 136The first thing to note, is that the web browser's "client" socket and the web 137server's "client" socket are identical beasts. That is, this is a "peer to peer" 138conversation. Or to put it another way, *as the designer, you will have to 139decide what the rules of etiquette are for a conversation*. Normally, the 140``connect``\ ing socket starts the conversation, by sending in a request, or 141perhaps a signon. But that's a design decision - it's not a rule of sockets. 142 143Now there are two sets of verbs to use for communication. You can use ``send`` 144and ``recv``, or you can transform your client socket into a file-like beast and 145use ``read`` and ``write``. The latter is the way Java presents its sockets. 146I'm not going to talk about it here, except to warn you that you need to use 147``flush`` on sockets. These are buffered "files", and a common mistake is to 148``write`` something, and then ``read`` for a reply. Without a ``flush`` in 149there, you may wait forever for the reply, because the request may still be in 150your output buffer. 151 152Now we come to the major stumbling block of sockets - ``send`` and ``recv`` operate 153on the network buffers. They do not necessarily handle all the bytes you hand 154them (or expect from them), because their major focus is handling the network 155buffers. In general, they return when the associated network buffers have been 156filled (``send``) or emptied (``recv``). They then tell you how many bytes they 157handled. It is *your* responsibility to call them again until your message has 158been completely dealt with. 159 160When a ``recv`` returns 0 bytes, it means the other side has closed (or is in 161the process of closing) the connection. You will not receive any more data on 162this connection. Ever. You may be able to send data successfully; I'll talk 163more about this later. 164 165A protocol like HTTP uses a socket for only one transfer. The client sends a 166request, then reads a reply. That's it. The socket is discarded. This means that 167a client can detect the end of the reply by receiving 0 bytes. 168 169But if you plan to reuse your socket for further transfers, you need to realize 170that *there is no* :abbr:`EOT (End of Transfer)` *on a socket.* I repeat: if a socket 171``send`` or ``recv`` returns after handling 0 bytes, the connection has been 172broken. If the connection has *not* been broken, you may wait on a ``recv`` 173forever, because the socket will *not* tell you that there's nothing more to 174read (for now). Now if you think about that a bit, you'll come to realize a 175fundamental truth of sockets: *messages must either be fixed length* (yuck), *or 176be delimited* (shrug), *or indicate how long they are* (much better), *or end by 177shutting down the connection*. The choice is entirely yours, (but some ways are 178righter than others). 179 180Assuming you don't want to end the connection, the simplest solution is a fixed 181length message:: 182 183 class MySocket: 184 """demonstration class only 185 - coded for clarity, not efficiency 186 """ 187 188 def __init__(self, sock=None): 189 if sock is None: 190 self.sock = socket.socket( 191 socket.AF_INET, socket.SOCK_STREAM) 192 else: 193 self.sock = sock 194 195 def connect(self, host, port): 196 self.sock.connect((host, port)) 197 198 def mysend(self, msg): 199 totalsent = 0 200 while totalsent < MSGLEN: 201 sent = self.sock.send(msg[totalsent:]) 202 if sent == 0: 203 raise RuntimeError("socket connection broken") 204 totalsent = totalsent + sent 205 206 def myreceive(self): 207 chunks = [] 208 bytes_recd = 0 209 while bytes_recd < MSGLEN: 210 chunk = self.sock.recv(min(MSGLEN - bytes_recd, 2048)) 211 if chunk == b'': 212 raise RuntimeError("socket connection broken") 213 chunks.append(chunk) 214 bytes_recd = bytes_recd + len(chunk) 215 return b''.join(chunks) 216 217The sending code here is usable for almost any messaging scheme - in Python you 218send strings, and you can use ``len()`` to determine its length (even if it has 219embedded ``\0`` characters). It's mostly the receiving code that gets more 220complex. (And in C, it's not much worse, except you can't use ``strlen`` if the 221message has embedded ``\0``\ s.) 222 223The easiest enhancement is to make the first character of the message an 224indicator of message type, and have the type determine the length. Now you have 225two ``recv``\ s - the first to get (at least) that first character so you can 226look up the length, and the second in a loop to get the rest. If you decide to 227go the delimited route, you'll be receiving in some arbitrary chunk size, (4096 228or 8192 is frequently a good match for network buffer sizes), and scanning what 229you've received for a delimiter. 230 231One complication to be aware of: if your conversational protocol allows multiple 232messages to be sent back to back (without some kind of reply), and you pass 233``recv`` an arbitrary chunk size, you may end up reading the start of a 234following message. You'll need to put that aside and hold onto it, until it's 235needed. 236 237Prefixing the message with its length (say, as 5 numeric characters) gets more 238complex, because (believe it or not), you may not get all 5 characters in one 239``recv``. In playing around, you'll get away with it; but in high network loads, 240your code will very quickly break unless you use two ``recv`` loops - the first 241to determine the length, the second to get the data part of the message. Nasty. 242This is also when you'll discover that ``send`` does not always manage to get 243rid of everything in one pass. And despite having read this, you will eventually 244get bit by it! 245 246In the interests of space, building your character, (and preserving my 247competitive position), these enhancements are left as an exercise for the 248reader. Lets move on to cleaning up. 249 250 251Binary Data 252----------- 253 254It is perfectly possible to send binary data over a socket. The major problem is 255that not all machines use the same formats for binary data. For example, a 256Motorola chip will represent a 16 bit integer with the value 1 as the two hex 257bytes 00 01. Intel and DEC, however, are byte-reversed - that same 1 is 01 00. 258Socket libraries have calls for converting 16 and 32 bit integers - ``ntohl, 259htonl, ntohs, htons`` where "n" means *network* and "h" means *host*, "s" means 260*short* and "l" means *long*. Where network order is host order, these do 261nothing, but where the machine is byte-reversed, these swap the bytes around 262appropriately. 263 264In these days of 32 bit machines, the ascii representation of binary data is 265frequently smaller than the binary representation. That's because a surprising 266amount of the time, all those longs have the value 0, or maybe 1. The string "0" 267would be two bytes, while binary is four. Of course, this doesn't fit well with 268fixed-length messages. Decisions, decisions. 269 270 271Disconnecting 272============= 273 274Strictly speaking, you're supposed to use ``shutdown`` on a socket before you 275``close`` it. The ``shutdown`` is an advisory to the socket at the other end. 276Depending on the argument you pass it, it can mean "I'm not going to send 277anymore, but I'll still listen", or "I'm not listening, good riddance!". Most 278socket libraries, however, are so used to programmers neglecting to use this 279piece of etiquette that normally a ``close`` is the same as ``shutdown(); 280close()``. So in most situations, an explicit ``shutdown`` is not needed. 281 282One way to use ``shutdown`` effectively is in an HTTP-like exchange. The client 283sends a request and then does a ``shutdown(1)``. This tells the server "This 284client is done sending, but can still receive." The server can detect "EOF" by 285a receive of 0 bytes. It can assume it has the complete request. The server 286sends a reply. If the ``send`` completes successfully then, indeed, the client 287was still receiving. 288 289Python takes the automatic shutdown a step further, and says that when a socket 290is garbage collected, it will automatically do a ``close`` if it's needed. But 291relying on this is a very bad habit. If your socket just disappears without 292doing a ``close``, the socket at the other end may hang indefinitely, thinking 293you're just being slow. *Please* ``close`` your sockets when you're done. 294 295 296When Sockets Die 297---------------- 298 299Probably the worst thing about using blocking sockets is what happens when the 300other side comes down hard (without doing a ``close``). Your socket is likely to 301hang. TCP is a reliable protocol, and it will wait a long, long time 302before giving up on a connection. If you're using threads, the entire thread is 303essentially dead. There's not much you can do about it. As long as you aren't 304doing something dumb, like holding a lock while doing a blocking read, the 305thread isn't really consuming much in the way of resources. Do *not* try to kill 306the thread - part of the reason that threads are more efficient than processes 307is that they avoid the overhead associated with the automatic recycling of 308resources. In other words, if you do manage to kill the thread, your whole 309process is likely to be screwed up. 310 311 312Non-blocking Sockets 313==================== 314 315If you've understood the preceding, you already know most of what you need to 316know about the mechanics of using sockets. You'll still use the same calls, in 317much the same ways. It's just that, if you do it right, your app will be almost 318inside-out. 319 320In Python, you use ``socket.setblocking(False)`` to make it non-blocking. In C, it's 321more complex, (for one thing, you'll need to choose between the BSD flavor 322``O_NONBLOCK`` and the almost indistinguishable POSIX flavor ``O_NDELAY``, which 323is completely different from ``TCP_NODELAY``), but it's the exact same idea. You 324do this after creating the socket, but before using it. (Actually, if you're 325nuts, you can switch back and forth.) 326 327The major mechanical difference is that ``send``, ``recv``, ``connect`` and 328``accept`` can return without having done anything. You have (of course) a 329number of choices. You can check return code and error codes and generally drive 330yourself crazy. If you don't believe me, try it sometime. Your app will grow 331large, buggy and suck CPU. So let's skip the brain-dead solutions and do it 332right. 333 334Use ``select``. 335 336In C, coding ``select`` is fairly complex. In Python, it's a piece of cake, but 337it's close enough to the C version that if you understand ``select`` in Python, 338you'll have little trouble with it in C:: 339 340 ready_to_read, ready_to_write, in_error = \ 341 select.select( 342 potential_readers, 343 potential_writers, 344 potential_errs, 345 timeout) 346 347You pass ``select`` three lists: the first contains all sockets that you might 348want to try reading; the second all the sockets you might want to try writing 349to, and the last (normally left empty) those that you want to check for errors. 350You should note that a socket can go into more than one list. The ``select`` 351call is blocking, but you can give it a timeout. This is generally a sensible 352thing to do - give it a nice long timeout (say a minute) unless you have good 353reason to do otherwise. 354 355In return, you will get three lists. They contain the sockets that are actually 356readable, writable and in error. Each of these lists is a subset (possibly 357empty) of the corresponding list you passed in. 358 359If a socket is in the output readable list, you can be 360as-close-to-certain-as-we-ever-get-in-this-business that a ``recv`` on that 361socket will return *something*. Same idea for the writable list. You'll be able 362to send *something*. Maybe not all you want to, but *something* is better than 363nothing. (Actually, any reasonably healthy socket will return as writable - it 364just means outbound network buffer space is available.) 365 366If you have a "server" socket, put it in the potential_readers list. If it comes 367out in the readable list, your ``accept`` will (almost certainly) work. If you 368have created a new socket to ``connect`` to someone else, put it in the 369potential_writers list. If it shows up in the writable list, you have a decent 370chance that it has connected. 371 372Actually, ``select`` can be handy even with blocking sockets. It's one way of 373determining whether you will block - the socket returns as readable when there's 374something in the buffers. However, this still doesn't help with the problem of 375determining whether the other end is done, or just busy with something else. 376 377**Portability alert**: On Unix, ``select`` works both with the sockets and 378files. Don't try this on Windows. On Windows, ``select`` works with sockets 379only. Also note that in C, many of the more advanced socket options are done 380differently on Windows. In fact, on Windows I usually use threads (which work 381very, very well) with my sockets. 382 383 384