A javascript topic-based messaging library
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


Build Status Code Climate Dependency Status Dev Dependency Status

Topic-based Messaging (bullhorn-js)


A topic-based publish/subscribe mechanism allows agents in a system to send messages into a global bus for any other agents to consume. We generally refer to the individual topics as “channels”, using the analogy of “tuning in” to an information stream. Interested agents subscribe to a given channel by providing a callback function that will be executed when any message comes across the specified channel. In this manner, application functionality can be coordinated among agents with no direct dependency on one another, allowing for the construction of ad hoc workflows from independent units. The idea is to abstract unit dependencies by tying their functionality to topics of interest oriented around data rather than interaction specifics from a particular unit. As an example, a specific agent in the system may provide functionality to read the content of text documents. A separate agent may be responsible for providing a list display of documents available. When a document is selected from the list display agent, this agent publishes a message into the global bus announcing that a document has been selected for viewing, and includes required information for other parties to respond (i.e., the document’s identifier). The text viewing agent would then respond to this message by executing any data requests required to obtain the text content it understands for this document, and then displaying it as designed. The text viewer would therefore be relieved of the responsibility for containing its own document-opening interface. This allows for a number of data viewers or display interfaces related to documents, which can operate purely based on data notifications about the current document.

##Channel Specifications In order to maintain predictable and orderly use of bullhorn, all channels must include an accompanying specification. This specification provides a namespaced channel name which must be unique within the system, as well as the details of what data properties any messages coming across the channel must have. Incoming messages can therefore be subject to validation and rejected if non-conformant, reducing the likelihood of communication errors (“misunderstandings”) within the system. Because this is a JavaScript application, messages must be either a JSON string, or a live object that is serializable to JSON. This makes the messaging system scalable to other uses, such as cross-window or client-server. The specification for channels should be written in JSONSchema, a formal IETF specification (draft) for describing JSON objects using JSON itself (http://tools.ietf.org/html/draft-zyp-json-schema-03).

##Core API

At the core level, bullhorn is quite simple. Only three functions are required in order to accomplish all major functionality: publish, subscribe, and unsubscribe. This represents the low-level API of bullhorn.


The publish function allows any unit to send a message into the bus.


The subscribe function allows any unit to indicate interest in messages on a specific channel.


The unsubscribe function allows any unit to unsubscribe from a channel, thereby releasing its interest. This is important to keeping channels and callbacks clear from over activity or potential errors when executing callbacks on destroyed objects.

##Object-Oriented API

While the Core API is fairly simple, in an object-oriented architecture it is desirable to approach most functionality from an object-oriented perspective. This makes the components more modular, abstracts low-level functionality, and allows for injection of alternate low-level providers in order to do testing. For messaging, a simple way to do this is to create a Channel object with appropriate methods for publishing and subscribing, along with a mechanism for obtaining instances of these channels that helps reduce boilerplate and enables dependency-free testing. We will obtain Channel instances using a factory, which allows for additional shortcuts and boilerplate reductions.

###ChannelFactory The ChannelFactory is used for obtaining instances of a Channel. The factory should provide a simple method for obtaining new channels with minimal typing needed, which has the side benefit of allowing greater code minification. The ChannelFactory should be a “single instance” object that is expected to be instantiated by the application using it, with a passed-in string indicating the fully-qualified namespace to use for looking up channel names (e.g., “schema.channels”). Access to the factory instance should be mapped to a convenient alias for the application.

####get The get method returns new Channel instance objects that are tied to both a specific channel and a specific unit.

###Channel The Channel object is the core element that any units will work with in order to send and receive messages into bullhorn. The Channel object wraps calls to bullhorn, with instances holding reference information such as the channel name and the scope it is tied to, allowing for significant boilerplate reduction when pubsub is used. New channels should receive a reference to the factory that created them, in the event that they need to communicate with the factory for registration, lookups, or other communications that enable greater abstraction in the client API. Note that new Channels should never be instantiated directly by those who wish to publish or subscribe. This can be accomplished by not exposing a public constructor for Channels, rather, the ChannelFactory creates live objects with functions directly attached as needed. This makes the “Channel” object something like a private inlined class on the factory.

####recreate Creates a new channel instance using the same parameters as this channel, but with a new callback context.

####publish Publishes the specified message on the channel, to the global pubsub bus. An optional callback can be passed that is executed when the message has reached all subscribers. The scope used for retrieving the instance will be used for callback execution.

####subscribe Subscribes a callback to the channel object. The scope used for retrieving the instance will be used for callback execution. Optional Boolean to ignore messages published by source agent if also subscribed to. Returns a handle object which can be used to unsubscribe this individual subscription later.

####unsubscribe Unsubscribes this channel instance and the callbacks associated with it. The scope used for retrieving the instance will be used for unsubscribing. Optionally a handle object can be passed, to limit unsubscription to just that one subscribe instance (instead of all callbacks associated with the scope).

##OOP API Code Examples The following code examples illustrate instantiating a ChannelFactory and mapping it to project-specific aliases, as well as retrieving Channel instances and working with them from an agent.

//instantiate factory and map to aliases for access
//(this would likely happen in some app setup/bootstrap area)
var factory = new ChannelFactory();

//create new channels for publish/subscribe
function setupChannels() {
    var documentClosed = factory.get(DocumentClosed, this);

    var documentOpened = factory.get(DocumentOpened, this);
    documentOpened.publish({documentId: 1});



Builds are done with grunt.

  1. Install node.js if you don't have it
  2. Install grunt-cli if you don't have it npm install -g grunt-cli
  3. Install deps npm install
  4. Run linter/tests npm test

##License This software is licensed under the Apache 2.0 license (http://www.apache.org/licenses/LICENSE-2.0).