Replyer is a wrapper of MQTT.js client that expands its API with common use cases. It is completely compatible and interoperable, so in most projects you can upgrade by dropping:
var mqtt = require('replyer')
var mqtt = require('mqtt')
var mqtt =var client = mqttclientclient
var mqtt =var client = mqttclientclientclient
This package is real time communication similar trying to join the best of both socket.io and MQTT.
It implements request and response with a similar API as in
socket.io to ease refactor. All clients are listening on their own id path as:
The ‘at’ symbol identifies a message as a MQTT request: it has to be addressed only to that particular client and request. #msg-id is an optional parameter to ensure that a certain reply is for a very particular request. Under some conditions we can count on this (qos > 0). Messages may get lost, for which timeouts are implemented. In case of a series of packets lost before a timeout, a client may interpret a wrong order, thus failing. So this little overhead for request/reply is really useful.
Replyer requests must indicate –somehow– the
clientId and the
messageId. We have to take into account the the messageId parameter in the packet is not always mandatory (qos 0), so depending in
qos for a request we have or have not to append it.
On a request if
qos === 2 message id would never be necessary.
Now that we can make requests to an API and receive an answer on plain MQTT protocol, we could make use of HTTP URL scheme to indicate resources.
This is the skeleton of a MQTT URL
mqtt://host:port/path/to/api/call/@issuer/mid as emitter URL
mqtt://host:port/@issuer/mid/ as listener URL
The code has tried to follow MQTT best practices as in http://www.hivemq.com/blog/mqtt-essentials-part-5-mqtt-topics-best-practices