TCP Sockets: Double messages - python

I'm having a problem with sockets in python.
I have a a TCP server and client that send each other data in a while 1 loop.
It packages up 2 shorts in the struct module (struct.pack("hh", mousex, mousey)). But sometimes when recving the data on the other computer, it seems like 2 messages have been glued together. Is this nagle's algorithm?
What exactly is going on here? Thanks in advance.

I agree with other posters, that "TCP just does that". TCP guarantees that your bytes arrive in the right order, but makes no guarantees about the sizes of the chunks they arrive in. I would add that TCP is also allowed to split a single send into multiple recv's, or even for example to split aabb, ccdd into aab, bcc, dd.
I put together this module for dealing with the relevant issues in python:
http://stromberg.dnsalias.org/~strombrg/bufsock.html
It's under an opensource license and is owned by UCI. It's been tested on CPython 2.x, CPython 3.x, Pypy and Jython.
HTH

To be sure I'd have to see actual code, but it sounds like you are expecting a send of n bytes to show up on the receiver as exactly n bytes all the time, every time.
TCP streams don't work that way. It's a "streaming" protocol, as opposed to a "datagram" (record-oriented) one like UDP or STCP or RDS.
For fixed-data-size protocols (or any where the next chunk size is predictable in advance), you can build your own "datagram-like receiver" on a stream socket by simply recv()ing in a loop until you get exactly n bytes:
def recv_n_bytes(socket, n):
"attempt to receive exactly n bytes; return what we got"
data = []
while True:
have = sum(len(x) for x in data)
if have >= n:
break
want = n - have
got = socket.recv(want)
if got == '':
break
return ''.join(data)
(untested; python 2.x code; not necessarily efficient; etc).

You may not assume that data will become available for reading from the local socket in the same size pieces it was provided for sending at the other source end. As you have seen, this might sometimes be usually true, but by no means reliably so. Rather, what TCP guarantees is that what goes in one end will eventually come out the other, in order without anything missing or if that cannot be achieved by means built into the protocol such as retries, then whole thing will break with an error.
Nagle is one possible cause, but not the only one.

Related

Why do sockets behave like this?

So I was just messing around with sockets in python. I discovered that setting the socket option SO_RECVBUF to N makes the sockets recv buffer become 2N bytes large, according to the getsockopt function. For example:
import socket
a, b = socket.socketpair()
a.setsockopt(socket.SOL_SOCKET, socket.SO_RCVBUF, 4096)
print a.getsockopt(socket.SOL_SOCKET, socket.SO_RCVBUF) #prints 8192
b.send('1'*5000)
print len(a.recv(5000)) #prints 5000 instead of 4096 or something else.
a.setsockopt(socket.SOL_SOCKET, socket.SO_RCVBUF, 8192)
print a.getsockopt(socket.SOL_SOCKET, socket.SO_RCVBUF) #prints 16384
Can someone explain this to me? I am writing an HTTP server and I want to strictly limit the size of a request to protect my preciously scarce amount of RAM.
Internally this is making an OS level operation which according to man 7 socket says the following:
SO_RCVBUF
Sets or gets the maximum socket receive buffer in bytes. *The kernel doubles this value (to allow space for bookkeeping overhead) when it is set using setsockopt(2), and this doubled value is returned by getsockopt(2).** The default value is set by the /proc/sys/net/core/rmem_default file, and the maximum allowed value is set by the /proc/sys/net/core/rmem_max file. The minimum (doubled) value for this option is 256.
Liberally copied from this wonderful answer to a slightly different question: https://stackoverflow.com/a/11827867/758446

Passing a record over a socket

I have basic socket communication set up between python and Delphi code (text only). Now I would like to send/receive a record of data on both sides. I have a Record "C compatible" and would like to pass records back and forth have it in a usable format in python.
I use conn.send("text") in python to send the text but how do I send/receive a buffer with python and access the record items sent in python?
Record
TPacketData = record
pID : Integer;
dataType : Integer;
size : Integer;
value : Double;
end;
I don't know much about python, but I have done a lot between Delphi, C++, C# and Java even with COBOL.Anyway, to send a record from Delphi to C first you need to pack the record at both ends,
in Deplhi
MyRecord = pack record
in C++
#pragma pack(1)
I don’t know in python but I guess there must be a similar one. Make sure that at both sides the sizeof(MyRecord) is the same length.Also, before sending the records, you should take care about byte ordering (you know, Little-Endian vs Big-Endian), use the Socket.htonl() and Socket.ntohl() in python and the equivalent in Deplhi which are in WinSock unit. Also a "double" in Delphi could not be the same as in python, in Delphi is 8 bytes check this as well, and change it to Single(4 bytes) or Extended (10 bytes) whichever matches.
If all that match then you could send/receive binary records in one shut, otherwise, I'm afraid, you have to send the individual fields one by one.
I know this answer is a bit late to the game, but may at least prove useful to other people finding this question in their search-results. Because you say the Delphi code sends and receives "C compatible data" it seems that for the sake of the answer about Python's handling it is irrelevant whether it is Delphi (or any other language) on the other end...
The python struct and socket modules have all the functionality for the basic usage you describe. To send the example record you would do something like the below. For simplicity and sanity I have presumed signed integers and doubles, and packed the data in "network order" (bigendian). This can easily be a one-liner but I have split it up for verbosity and reusability's sake:
import struct
t_packet_struc = '>iiid'
t_packet_data = struct.pack(t_packet_struc, pid, data_type, size, value)
mysocket.sendall(t_packet_data)
Of course the mentioned "presumptions" don't need to be made, given tweaks to the format string, data preparation, etc. See the struct inline help for a description of the possible format strings - which can even process things like Pascal-strings... By the way, the socket module allows packing and unpacking a couple of network-specific things which struct doesn't, like IP-address strings (to their bigendian int-blob form), and allows explicit functions for converting data bigendian-to-native and vice-versa. For completeness, here is how to unpack the data packed above, on the Python end:
t_packet_size = struct.calcsize(t_packet_struc)
t_packet_data = mysocket.recv(t_packet_size)
(pid, data_type, size, value) = struct.unpack(t_packet_struc,
t_packet_data)
I know this works in Python version 2.x, and suspect it should work without changes in Python version 3.x too. Beware of one big gotcha (because it is easy to not think about, and hard to troubleshoot after the fact): Aside from different endianness, you can also distinguish between packing things using "standard size and alignment" (portably) or using "native size and alignment" (much faster) depending on how you prefix - or don't prefix - your format string. These can often yield wildly different results than you intended, without giving you a clue as to why... (there be dragons).

How to deal with errors on select with multiple IO steams?

I have the following call to select:
try:
rlst, wlst, plst = select.select(
[x.fileno() for x in self.rlist],
[x.fileno() for x in self.wlist],
[x.fileno() for x in self.plist])
except select.error, err:
[...]
Where self.rlist, self.wlist, and self.plist are lists of IO streams (either sockets, PIPE, files, whatever). Now, I am assuming that this select could fail when one of the streams fails for whatever reason.
How can I find out which of those streams caused the error? What I really want to do is remove that IO stream from its list and continue with the select.
Quoting from the Socket Programming HOWTO:
One very nasty problem with select: if somewhere in those input lists of sockets is one which has died a nasty death, the select will fail. You then need to loop through every single damn socket in all those lists and do a select([sock],[],[],0) until you find the bad one. That timeout of 0 means it won’t take long, but it’s ugly.

How to read JSON from socket in python? (Incremental parsing of JSON)

I have a socket opened and I'd like to read some json data from it. The problem is that the json module from standard library can only parse from strings (load only reads the whole file and calls loads inside) It even looks that all the way inside the module it all depends on the parameter being string.
This is a real problem with sockets since you can never read it all to string and you don't know how many bytes to read before you actually parse it.
So my questions are: Is there a (simple and elegant) workaround? Is there another json library that can parse data incrementally? Is it worth writing it myself?
Edit: It is XBMC jsonrpc api. There are no message envelopes, and I have no control over the format. Each message may be on a single line or on several lines.
I could write some simple parser that needs only getc function in some form and feed it using s.recv(1), but this doesn't as a very pythonic solution and I'm a little lazy to do that :-)
Edit: given that you aren't defining the protocol, this isn't useful, but it might be useful in other contexts.
Assuming it's a stream (TCP) socket, you need to implement your own message framing mechanism (or use an existing higher level protocol that does so). One straightforward way is to define each message as a 32-bit integer length field, followed by that many bytes of data.
Sender: take the length of the JSON packet, pack it into 4 bytes with the struct module, send it on the socket, then send the JSON packet.
Receiver: Repeatedly read from the socket until you have at least 4 bytes of data, use struct.unpack to unpack the length. Read from the socket until you have at least that much data and that's your JSON packet; anything left over is the length for the next message.
If at some point you're going to want to send messages that consist of something other than JSON over the same socket, you may want to send a message type code between the length and the data payload; congratulations, you've invented yet another protocol.
Another, slightly more standard, method is DJB's Netstrings protocol; it's very similar to the system proposed above, but with text-encoded lengths instead of binary; it's directly supported by frameworks such as Twisted.
If you're getting the JSON from an HTTP stream, use the Content-Length header to get the length of the JSON data. For example:
import httplib
import json
h = httplib.HTTPConnection('graph.facebook.com')
h.request('GET', '/19292868552')
response = h.getresponse()
content_length = int(response.getheader('Content-Length','0'))
# Read data until we've read Content-Length bytes or the socket is closed
data = ''
while len(data) < content_length or content_length == 0:
s = response.read(content_length - len(data))
if not s:
break
data += s
# We now have the full data -- decode it
j = json.loads(data)
print j
What you want(ed) is ijson, an incremental json parser.
It is available here: https://pypi.python.org/pypi/ijson/ . The usage should be simple as (copying from that page):
import ijson.backends.python as ijson
for item in ijson.items(file_obj):
# ...
(for those who prefer something self-contained - in the sense that it relies only on the standard library: I wrote yesterday a small wrapper around json - but just because I didn't know about ijson. It is probably much less efficient.)
EDIT: since I found out that in fact (a cythonized version of) my approach was much more efficient than ijson, I have packaged it as an independent library - see here also for some rough benchmarks: http://pietrobattiston.it/jsaone
Do you have control over the json? Try writing each object as a single line. Then do a readline call on the socket as described here.
infile = sock.makefile()
while True:
line = infile.readline()
if not line: break
# ...
result = json.loads(line)
Skimming the XBMC JSON RPC docs, I think you want an existing JSON-RPC library - you could take a look at:
http://www.freenet.org.nz/dojo/pyjson/
If that's not suitable for whatever reason, it looks to me like each request and response is contained in a JSON object (rather than a loose JSON primitive that might be a string, array, or number), so the envelope you're looking for is the '{ ... }' that defines a JSON object.
I would, therefore, try something like (pseudocode):
while not dead:
read from the socket and append it to a string buffer
set a depth counter to zero
walk each character in the string buffer:
if you encounter a '{':
increment depth
if you encounter a '}':
decrement depth
if depth is zero:
remove what you have read so far from the buffer
pass that to json.loads()
You may find JSON-RPC useful for this situation. It is a remote procedure call protocol that should allow you to call the methods exposed by the XBMC JSON-RPC. You can find the specification on Trac.
res = str(s.recv(4096), 'utf-8') # Getting a response as string
res_lines = res.splitlines() # Split the string to an array
last_line = res_lines[-1] # Normally, the last one is the json data
pair = json.loads(last_line)
https://github.com/A1vinSmith/arbitrary-python/blob/master/sockets/loopHost.py

How do I dump the TCP client's buffer in order to accept more data?

I've got a simple TCP server and client. The client receives data:
received = sock.recv(1024)
It seems trivial, but I can't figure out how to recieve data larger than the buffer. I tried chunking my data and sending it multiple times from the server (worked for UDP), but it just told me that my pipe was broken.
Suggestions?
If you have no idea how much data is going to pour over the socket, and you simply want to read everything until the socket closes, then you need to put socket.recv() in a loop:
# Assumes a blocking socket.
while True:
data = sock.recv(4096)
if not data:
break
# Do something with `data` here.
Mike's answer is the one you're looking for, but that's not a situation you want to find yourself in. You should develop an over-the-wire protocol that uses a fixed-length field that describes how much data is going to be sent. It's a Type-Length-Value protocol, which you'll find again and again and again in network protocols. It future-proofs your protocol against unforeseen requirements and helps isolate network transmission problems from programmatic ones.
The sending side becomes something like:
socket.write(struct.pack("B", type) #send a one-byte msg type
socket.write(struct.pack("H", len(data)) #send a two-byte size field
socket.write(data)
And the receiving side something like:
type = socket.read(1) # get the type of msg
dataToRead = struct.unpack("H", socket.read(2))[0] # get the len of the msg
data = socket.read(dataToRead) # read the msg
if TYPE_FOO == type:
handleFoo(data)
elif TYPE_BAR == type:
handleBar(data)
else:
raise UnknownTypeException(type)
You end up with an over-the-wire message format that looks like:
struct {
unsigned char type;
unsigned short length;
void *data;
}
Keep in mind that:
Your operating system has it's own idea of what it's TCP/IP socket buffer size is.
TCP/IP packet maximum size (generally is 1500 bytes)
pydoc for socket suggests that 4096 is a good buffer size
With that said, it'd really be helpful to see the code around that one line. There are a few things that could play into this, if you're using select or just polling, is the socket non-blocking, etc.
It also matters how you're sending the data, if your remote end disconnects. More details.

Categories