I have a process that sits in a loop and receives commands.
receive
increase ->
...
decrease ->
...
after 5000 ->
...
end
But when I bombard it with thousands of messages it breaks down and receives these warnings.
Warning Message:
***WARNING*** Unexp msg {<0.106.0>,rec_acked}, info {running,
[{'_UserConnections',20}],
{ieval,3994,34,log,
clientLogging,
[20],
false}}
Is there anyway to handle this? And does it cause any issue?
Thank you for your answer!
This code is good just for example and practice, but don't run in production environment.
You should always receive all messages from process mailbox and select what you want after getting.
handle_message() ->
receive
Msg ->
handle_message(Msg)
after 5000 ->
handle_timeout()
end.
handle_message(increase) ->
...;
handle_message(decrease) ->
...;
handle_message(_) ->
%% Back to receiving loop
handle_message().
You should prevent filling process mailbox.
In production-ready application, often nobody uses receive statement, often they use some standard codes which those codes handle receiving, timeouts, replies, hibernation, etc. we call those codes behavior, for example one of OTP standad behaviors is gen_server behavior
Because OTP behaviors are for general purposes if you need very efficient code for doing some special duty, you have to write something named Special process which should handle your own messages and Erlang system messages.
I'm working on a protocol handling data exchange that somehow a little complex, then I found TLV is the one I need. Is there a formal way to read and write TLV in erlang? or some lib / code example handling this? thanks.
The "default" in Erlang is LTV rather than TLV, but it is rather easy to handle:
case gen_tcp:recv(Socket, 8) of
<<Type:32/integer, Len:32/integer>> ->
Payload = gen_tcp:recv(Socket, Len),
{type_of(Type), Payload};
...
end,
You will need passive sockets to get this to work, but it is rather easy to do. If you have the freedom to pick your format, the LTV encoding is better because you can then put the socket in {active, once} mode which means the underlying layer decodes stuff for you.
I haven't actually used it, but how about this one: https://github.com/essiene/smpp34pdu/blob/master/src/tlv.erl
I am trying to make a simple UDP packet decoder.
packet_decoder(Packet)->
<<Opts:8,MobIdLength:8,MobId:64,MobIdType:8,MgeType:8,SeqNum:16,Rest/binary>> = Packet,
io:format("Options:~p~n",Opts),
io:format("MobIdLength:~p~n",MobIdLength),
io:format("MobId:~p~n",MobId),
io:format("MobIdType:~p~n",MobIdType),
io:format("MgeType:~p~n",MgeType),
io:format("SeqNum:~p~n",SeqNum).
Packet is passed by a receive loop:
rcv_loop(Socket) ->
inet:setopts(Socket, [{active, once}, binary]),
io:format("rcvr started:~n"),
receive
{udp, Socket, Host, Port, Bin} ->
packet_decoder(Bin),
rcv_loop(Socket)
end.
I keep getting(following error edited 9/7/12 9:30 EST):
** exception error: no match of right hand side value
<<131,8,53,134,150,4,149,0,80,15,1,2,1,2,0,16,80,71,115,
52,80,71,115,53,24,63,227,197,211,...>>
in function udp_server:packet_decoder/1
called as udp_server:packet_decoder(<<131,8,53,134,150,4,149,0,80,15,
1,2,1,2,0,16,80,71,115,52,80,71,
115,53,24,63,227,197,...>>)
in call from udp_server:rcv_loop/1
in call from udp_server:init/0
If I create the same variable in the Erlang shell as a binary, i.e.
Packet = <<131,8,53,134,150,4,149,0,80,15,1,2,1,2,0,16,80,71,115,52,80,71,115,53,24,63,227,197,211,228,89,72,0,0,0,0,0,0,0,16,0,5,5,32,1,4,255,159,15,18,28,0,34,62,2,0,0,0,0,0,0,0,47,67>>.
<<Opts:8,MobIdLength:8,MobId:64,MobIdType:8,MgeType:8,SeqNum:16,Rest/binary>> = Packet.
It works just fine. Is there some subtlety in passing this to a function that I am missing? I have tried what I think is everything(except the right way). I tried setting the type and size. I also just tried
<<Rest/binary>> = Packet.
To no avail. Any help much appreciated.
The error you are getting when you run your code does not match your code. The error you are getting:
** exception error: no match of right hand side value ...
is a badmatch error and comes from an explicit = match where the pattern does not match the value from the RHS. There is no = in the code for rcv_loop/1. This implies that the loop you are running is not this code. So there are some questions to be asked:
When you have recompiled the module containing rcv_loop/1 have you restarted the loop so you run the new code? This is not done automagically.
Are you sure you are loading/running the code you think you are? I know this question sounds stupid but it is very easy, and not uncommon, to work on one version of the code and load another. You need to get the paths right.
The other things about mentioned about your code would not give this error. The calls to io:format/2 are wrong but would result in errors when you make the actual calls to io:format/2. Using the variable Socket as you do is not an error, it just means that you only want to receive UDP packets from just that socket.
EDIT : the first part of my answer was completely wrong so in order to not mislead, I deleted it.
Like spotted Alexey Kachayev io:format takes as second parameter a list, so :
packet_decoder(Packet)->
<<Opts:8,MobIdLength:8,MobId:64,MobIdType:8,MgeType:8,SeqNum:16,Rest/binary>> = Packet,
io:format("Options:~p~n",[Opts]),
io:format("MobIdLength:~p~n",[MobIdLength]),
io:format("MobId:~p~n",[MobId]),
io:format("MobIdType:~p~n",[MobIdType]),
io:format("MgeType:~p~n",[MgeType]),
io:format("SeqNum:~p~n",[SeqNum]).
I figured it out(kinda). I had been working on this in erlide in eclipse which had worked fine for all of the other parts of the. I tried compiling it from the erl shell and it worked fine. There must be some minor difference in the way eclipse is representing the source or the way it invokes the erlang compiler and shell. I will take it up with erlide.org. Thanks for the help!
Using Nitrogen, the Erlang web framework, I have the following method that receives messages and adds them to the bottom of an html element:
receive_messages() ->
receive
Message ->
io:format("~p received ~p", [self(), Message]),
wf:insert_bottom(messages, [#p{}, #span { text=io_lib:format("~p", [Message]) }])
end,
wf:comet_flush(),
receive_messages().
It is set as the usual way for comet:
wf:comet(fun() -> receive_messages() end)
It receives two messages very quickly:
<0.907.0> received {starting_chat,<0.905.0>}
<0.907.0> received {connected_to,<0.902.0>}
This is what I see in the HTML:
{connected_to, <0.902.0>}
{starting_chat, <0.905.0>}
Somehow, they ended in the reverse order.
I've started adding timer:sleep() calls to this method. With 50 milliseconds, they are in the correct order, with 20 they are not. When they are in the incorrect order they seem to be always in the incorrect order, it seems very deterministic.
Any ideas why? Is this a bug? Any ideas what should I do to get them in the correct order other than sleeping?
Also asked on the mailing list, in case there's more info there.
wf_comet.erl has a "strange" reverse call after applying inner_collect_content. My guess is that inner_collect_content used to be tail-recursive sometime ago.
If you apply the timeout, your messages are collected in different loops, one-by-one, so the order is not reversed anymore.
There seems to be a commit fixing this on another repository:
http://github.com/gersh/nitrogen/commit/a8bfcb23d003e68f7394a0455285beeb0fbf9b09