merging this into node-nanomsg

I've joined forces with Nick Desaulniers and the node-nanomsg team. We're maintaining this ongoing work and moved these efforts over there. -reqshark 02/24/2015

nanømsg streams

• pipe all endpoints together
• iøjs streams are domain, protocol, and transport agnostic
• combine sockets in new ways
• the socket's pipe() method is basically a more flexible zmq_proxy() or nn_device()

prerequisites

install nanomsg c lib and have iojs or node v0.10 - v0.12. on unix install pkg-config. for windows, you need visual studio and python, i would opt for anaconda installer with gitub powershell.

install

$ npm install iojs-nanomsg

example

var nano = require('iojs-nanomsg')
var pub   = nano.socket('pub'),       push  = nano.socket('push')
var sub   = nano.socket('sub'),       pull  = nano.socket('pull')

pub.bind('tcp://127.0.0.1:3333');     push.bind('tcp://127.0.0.1:4444')
sub.connect('tcp://127.0.0.1:3333');  pull.connect('tcp://127.0.0.1:4444')

sub.setEncoding('utf8') //sub socket will get utf8 strings instead of Buffers

sub.on('data', function (msg) {
  console.log(msg) //'hello from a push socket!'
})

pull.pipe(pub) //pipe readable sockets to any writeable socket or stream

setInterval( function(){ push.write('hello from a push socket!') }, 100 )

API

nano.socket(type, [options,])

Starts a new socket. The nanomsg socket can bind or connect to multiple heterogeneous endpoints as well as shutdown any of these established links.

options

• 'fam' (String, default: 'af_sp'): determines the domain of the socket. AF_SP creates a standard full-blown SP socket. AF_SP_RAW family sockets operate over internal network protocols and interfaces. Raw sockets omit the end-to-end functionality found in AF_SP sockets and thus can be used to implement intermediary devices in SP topologies, see nanomsg docs or consult your man page entry socket(2) for more info.

//ex. starting raw sockets
nano.socket('bus','raw') || nano.socket('bus', { fam: 'AF_SP_RAW' } )

• 'tcpnodelay' (boolean, default: false): see socket.tcpnodelay(boolean).
• 'linger' (number, default: 1000): see socket.linger(duration).
• 'sndbuf' (number, default: 128kB): see socket.sndbuf(size).
• 'rcvbuf' (number, default: 128kB): see socket.rcvbuf(size).
• 'sndtimeo' (number, default: -1): see socket.sndtimeo(duration).
• 'rcvtimeo' (number, default: -1): see socket.rcvtimeo(duration).
• 'reconn' (number, default: 100): see socket.reconn(duration).
• 'maxreconn' (number, default: 0): see socket.maxreconn(duration).
• 'sndprio' (number, default: 0): see socket.sndprio(priority).
• 'rcvprio' (number, default: 0): see socket.rcvprio(priority).

nano.version

require('iojs-nanomsg').version (Number): the libnanomsg beta version installed

socket.type

(String): Indicates what type of socket you have.

socket.shutdown(address)

(Function, param: String): Removes an endpoint established by calls to bind() or connect(). The nanomsg library will try to deliver any outstanding outbound messages to the endpoint for the time specified by linger.

socket.shutdown('tcp://127.0.0.1:5555')

socket.bind(address)

(Function, param: String): Adds a local endpoint to the socket. The endpoint can be then used by other applications to connect.

bind() (or connect()) may be called multiple times on the same socket thus allowing the socket to communicate with multiple heterogeneous endpoints.

socket.bind('tcp://eth0:5555')

recommend checking your machine's ifconfig first before using a named interface.

socket.connect(address)

(Function, param: String): Adds a remote endpoint to the socket. The nanomsg library would then try to connect to the specified remote endpoint.

connect() (as well as bind()) may be called multiple times on the same socket thus allowing the socket to communicate with multiple heterogeneous endpoints.

socket.connect('tcp://127.0.0.1:5555')

When connecting over remote TCP allow 100ms or more depending on round trip time for the operation to complete.

a note on address strings

socket.close(callback)

(Function, param: Function): Closes the socket. Any buffered inbound messages that were not yet received by the application will be discarded. The nanomsg library will try to deliver any outstanding outbound messages for the time specified by linger.

the callback function fires as soon as the underlying c lib's nn_close() operation also fires, so this is when linger starts.

sending and receiving: writeable and readable

socket.send(msg)

(Function, param: String or Buffer): equivalent to the socket.send() in node.zeromq.

socket.send('hello from nanømsg!')

send(msg) is automatically invoked during Writeable consumption of some other Readable stream. In that case a pipe() method can be used to transmit from a readable data source. The flow of data distributes to endpoint(s) determined by the particular socket type.

var fs = require('fs')
var source = fs.createReadStream(__dirname + 'filename.ext')

source.pipe(socket) //sends each chunk as a msg to socket's particular endpoint

socket.on(data, callback)

(Function, param order: String, Function): The Readable stream's on() function is an event listener registered with the nanomsg c lib that emits 'data' events. To receive messages, pass the string 'data' followed a callback containing a single data parameter.

socket.on('data', function (msg) {
  console.log(String(msg)) //'hello from nanømsg!'
})

the readable stream's data event is automatically invoked when piped to a Writeable or Transform consumer stream. Here msgprocessor is a transform you could pipe to a writeable or the next transform:

var through = require('through')

var msgprocessor = through(function(msg){
  var str = String(msg); console.log(str) //'hello from nanømsg!'
  this.queue(str + ' and cheers from nanomsg.iojs!')
})

socket.pipe(msgprocessor)

socket.tcpnodelay(boolean)

(Function, param: Boolean, default: false): When set, disables Nagle’s algorithm. It also disables delaying of TCP acknowledgments. Using this option improves latency at the expense of throughput.

Pass no parameter for current tcp nodelay setting.

//default
console.log(socket.tcpnodelay()) //tcp nodelay: off

socket.tcpnodelay(true) //disabling Nagle's algorithm

console.log(socket.tcpnodelay()) //tcp nodelay: on

socket.linger(duration)

(Function, param: Number, default: 1000): Specifies how long the socket should try to send pending outbound messages after socket.close() or socket.shutdown() is called, in milliseconds.

Pass no parameter for the linger duration.

socket.linger(5000)
console.log(socket.linger()) //5000

socket.sndbuf(size)

(Function, param: Number, default: 128kB): Size of the send buffer, in bytes. To prevent blocking for messages larger than the buffer, exactly one message may be buffered in addition to the data in the send buffer.

Pass no parameter for the socket's send buffer size.

socket.sndbuf(131072)
console.log(socket.sndbuf()) // 131072

socket.rcvbuf(size)

(Function, param: Number, default: 128kB): Size of the receive buffer, in bytes. To prevent blocking for messages larger than the buffer, exactly one message may be buffered in addition to the data in the receive buffer.

Pass no parameter for the socket's receive buffer size.

socket.rcvbuf(20480)
console.log(socket.rcvbuf()) // 20480

socket.sndtimeo(duration)

(Function, param: Number, default: -1): The timeout for send operation on the socket, in milliseconds.

Pass no parameter for the socket's send timeout.

socket.sndtimeo(200)
console.log(socket.sndtimeo()) // 200

socket.rcvtimeo(duration)

(Function, param: Number, default: -1): The timeout for recv operation on the socket, in milliseconds.

Pass no parameter for the socket's recv timeout.

socket.rcvtimeo(50)
console.log(socket.rcvtimeo()) // 50

socket.reconn(duration)

(Function, param: Number, default: 100): For connection-based transports such as TCP, this option specifies how long to wait, in milliseconds, when connection is broken before trying to re-establish it. Note that actual reconnect interval may be randomized to some extent to prevent severe reconnection storms.

Pass no parameter for the socket's reconnect interval.

socket.reconn(600)
console.log(socket.reconn()) // 600

socket.maxreconn(duration)

(Function, param: Number, default: 0): Only to be used in addition to socket.reconn(). maxreconn() specifies maximum reconnection interval. On each reconnect attempt, the previous interval is doubled until maxreconn is reached. Value of zero means that no exponential backoff is performed and reconnect interval is based only on reconn. If maxreconn is less than reconn, it is ignored.

Pass no parameter for the socket's maxreconn interval.

socket.maxreconn(60000)
console.log(socket.maxreconn()) // 60000

socket.sndprio(priority)

(Function, param: Number, default: 8): Sets outbound priority for endpoints subsequently added to the socket.

This option has no effect on socket types that send messages to all the peers. However, if the socket type sends each message to a single peer (or a limited set of peers), peers with high priority take precedence over peers with low priority.

Highest priority is 1, lowest is 16. Pass no parameter for the socket's current outbound priority.

socket.sndprio(2)
console.log(socket.sndprio()) // 2

socket.rcvprio(priority)

(Function, param: Number, default: 8): Sets inbound priority for endpoints subsequently added to the socket.

This option has no effect on socket types that are not able to receive messages.

When receiving a message, messages from peer with higher priority are received before messages from peer with lower priority.

Highest priority is 1, lowest is 16. Pass no parameter for the socket's current inbound priority.

socket.rcvprio(10)
console.log(socket.rcvprio()) // 10

test

on unix systems:

$ make clean && make && make check

# or if u want all that just run:
$ make full

on windows:

git submodule update --init
make full

disabling node v0.08.x for now... but v0.10 and up should be fine

performance

run benchmarks:

$ make perf

for more info how to do that and your own custom comparisons check out: running benchmarks

and if you want you can also run:

$ make bench

:)