Warning As of 2023-04-20 this repository is Archived and in read-only mode. It will receive no further updates.
As it is MIT licensed, anyone is welcome to fork this repo under the terms of that license.
Thanks to all past contributors!
A simplified way to setup an AMQP connection/channel with amqplib. It's a function that takes an AMQP url as the first parameter and an optional second parameter that defines which methods and arguments should be called on the channel
. The function returns a Promise
that will resolve with the a channel
object once all the method invocations defined in the second parameter have been resolved. Please see amqplib's documentation for the channel
API.
npm install amqp.channel --save
amqplib syntax:
require('amqplib').connect(url).then(function(connection){
return connection.createConfirmChannel();
}).then(function(channel){
return require('bluebird').all([
channel.assertExchange('exchange', 'fanout', { durable: true }),
channel.checkExchange('exchange'),
channel.bindExchange('alt.exchange', 'exchange', ''),
channel.unbindExchange('alt.exchange', 'exchange', ''),
channel.deleteExchange('alt.exchange', { ifEmpty: true }),
channel.assertQueue('first', { durable: true }),
channel.assertQueue('second'),
channel.checkQueue('first'),
channel.bindQueue('first', 'exchange', ''),
channel.unbindQueue('first', 'exchange', ''),
channel.purgeQueue('first'),
channel.deleteQueue('first', { ifEmpty: true }),
channel.deleteQueue('second')
]);
}).then(function(channel){
// Do stuff with the channel
});
amqp.channel syntax:
require('amqp.channel')(url, {
assertExchange : [['exchange', 'fanout', { durable: true }]],
checkExchange : [['exchange']],
bindExchange : [['alt.exchange', 'exchange', '']],
unbindExchange : [['alt.exchange', 'exchange', '']],
deleteExchange : [['alt.exchange', { ifEmpty: true }]],
assertQueue : [['first', { durable: true }], ['second']],
checkQueue : [['first']],
bindQueue : [['first', 'exchange', '']],
unbindQueue : [['first', 'exchange', '']],
purgeQueue : [['first']],
deleteQueue : [['first', { ifEmpty: true }], ['second']]
}).then(function(channel){
// Do stuff with the channel
});
The channel
object resolved by the returned Promise
will behave differently from a normal channel
object returned by the amqplib library in a few (hopefully convenient) ways:
- The
consume
,publish
, andsendToQueue
channel methods have been changed to explicitly handle JSON. - The
publish
andsendToQueue
methods have been "promisified" in a way that will still provide information to know whether or not the write buffer is full (and therefore, whether or not you should continue writing to it) by adding an additionalok
boolean property to the promise. - A
channel
consumer callback will no longer receivenull
when that consumer had been cancelled by Rabbit MQ. Instead, thechannel
object will emit a'cancelled'
event with all the arguments passed to thechannel.consume()
call for the consumer that was cancelled.
Automatic translation of JS object to JSON string to Buffer for sending/publishing:
channel.sendToQueue('someQueue', { hello: 'world' });
channel.publish('someExchange', 'routingKey', { hello: 'world' });
Promisification of sendToQueue
and publish
methods:
return channel.sendToQueue('someQueue', { hello: 'world' }).then(function(){
return channel.publish('someExchange', 'routingKey', { hello: 'world' });
});
Automatic translation of message Buffer to JSON string to JS object for consuming:
channel.sendToQueue('someQueue', { hello: 'world' });
channel.consume('someQueue', function(parsedMessage, originalMessage){
console.log('hello', parsedMessage.hello); // => hello world
channel.ack(originalMessage);
});
Handling a consumer getting cancelled by Rabbit MQ:
channel.on('cancelled', function(queue, callback, options){
// When the consumer below gets cancelled by Rabbit MQ
console.log(queue, callback.name, options); // 'someQueue', 'onMessage', { noAck: true }
});
channel.consume('someQueue', function onMessage(parsedMessage, originalMessage){
console.log(parsedMessage);
}, { noAck: true });
The ok
property on the promises returned by the sendToQueue
and publish
methods:
var sent = channel.sendToQueue('someQueue', { hello: 'world' });
if (sent.ok) {
// continue sending
} else {
// maybe pause sending until unblocked?
channel.once('drain', function(){
// continue sending
});
}
amqp.channel supports two parameters meant to modify connection options passed on to amqplib.
The first is an object called socketOptions
is the equivalent to the same named property on the
channel#connect method.
The second is called defaultServernameToHostname
. This property is meant to signal that the servername property on the
socketoptions should be defaulted to the hostname from the url passed in. This is meant to simplify correctly configuring
connections over TLS for SNI. See the following for more details:
nodejs/node#28167 (comment) amqp-node/amqplib#331
Say you wanted to listen to the 'foo'
exchange and send a different message to the 'bar'
queue every time the message's baz
property contained the word 'qux'
.
In your config.js:
var env = process.env;
var cfg = {
exchange: env.EXCHANGE_TO_BIND_TO || 'foo',
queue: {
toSendTo: env.QUEUE_TO_SEND_TO || 'bar',
toConsumeFrom: env.QUEUE_TO_CONSUME_FROM || 'baz',
}
amqpUrl: env.RABBIT_MQ_URL || 'amqp://test:[email protected]:5672'
};
cfg.channelMethodsToCall = {
assertQueue: // Channel method to invoke
[ // Array of channel method invocations
[ // channel.assertQueue( cfg.queue.toConsumeFrom )
cfg.queue.toConsumeFrom
],
[ // channel.assertQueue( cfg.queue.toSendTo, { durable: true } );
cfg.queue.toSendTo, { durable: true }
]
],
assertExchange: [
[ cfg.exchange, 'fanout' ] // channel.assetExchange(cfg.exchange, 'fanout')
],
bindQueue: [
[ cfg.queue.toConsumeFrom, cfg.exchange, '' ]
]
}
module.exports = cfg;
In your app.js:
var cfg = require('./config');
var amqp = require('amqp.channel');
module.exports = amqp(cfg.amqpUrl, cfg.channelMethodsToCall)
.then(consumeAtMost(1))
.then(consumeFrom(cfg.queue.toConsumeFrom));
function consumeAtMost(maxMessages){
return function(channel){
// Only process `maxMessages` at a time and don't consume another
// message until we've either `ack` or `nack` the current one.
return channel.prefetch(maxMessage).then(function(){
return channel;
});
}
}
function consumeFrom(queue){
return function(channel){
channel.consume(queue, function onMessage(parsed, msg){
if (/baz/.test(parsed.baz)) {
var msgToSend = { hello: 'world' };
var options = { persistent: true };
var sendMsg = channel.sendToQueue(cfg.queue.toSendTo, msgToSend, options);
sendMsg.catch(function(e){
console.error(e);
// Try to process message again?
// onMessage(parsed, msg);
});
if (sendMsg.ok) {
channel.ack(msg);
} else {
sendMsg.then(function(){
channel.ack(msg);
});
}
} else {
channel.ack(msg);
}
});
channel.on('cancelled', function onConsumerCancelled(queue, cb, options){
console.warn('RabbitMQ cancelled your consumer for %s', queue);
// Try to setup the consumer again?
// channel.consume(queue, cb, options);
});
return channel;
}
}