Messaging/Service Bus for the harmonic convergence of node-based enterprise entities.
Harcon - Proven and reliable microservice solution for the harmonic convergence of JS entities. Designed to be scalable from prototypes till highly structured node-based enterprise components of real-time systems.
Harcon is strong to build an endlessly scalable execution-chain or workflow-based or machine-learning solution for NodeJS applications.
Need help? Join me on
and level your problem with me freely. :)
!Note: From version 3.0.0, harcon provides callback interface with Promise support in its functions.
!Note: From version 5.0.0, harcon has powerful toolset to define and perform execution chains and workflows.
========
harcon is a microservice solution for NodeJS/Browser giving superior abstraction layer for interoperability between entities in a highly structured and fragmented ecosystem. It allows you to design and implement complex workflows and microservices where context and causality of messages are important.
The library has a stunning feature list beyond basic messaging functionality.
Channel-agnostic: harcon represents a very abstract messaging framework allowing you to use any underlaying technology your application requires: AMQP, MQTT, Amazon SQS, etc... For amqp integration, please check this: harcon-amqp For sqs integration, please check this: harcon-sqs For mqtt integration, please check this: harcon-mqtt
Tracking: you can monitor every message delivered (request or response) by only few lines of code
Flow control / Reproducibility: A flow of communication / messages can be halted / continued / reinitiated anytime with no effort
Free orchestration: your system can be orchestrated and distributed as you wish, message delivery is not limited to nodes or hosts
Short learning curve: no need to learn hundred of pages, communication has to be simple after all
Log-free coding: no more mixture of logging and business logic. Harcon logs all messages exchanged.
Transparent: although harcon introduces lots of complex types and structures, your code and callbacks will be kept clean and pure, everything is (un)packed in the background in a transparent way
Smooth infiltration: your objects / functions will possess the necessary services via injection, no need to create complex structures and compounds
Advanced routing & listening: system fragmentation, qualified names, regular expressions, wildcards, etc.
Execution-chain: toolset to easily define execution paths between entities to relay messages and results.
Business-transaction: small helpers to manage business executions paths as (distributed) transactions.
!Note: Harcon's concept is to introduce a clean and high abstraction layer over messaging between entities. Like in case of every abstraction tool, for webapps which are simple as 1, it can be proven as a liability.
This library starts to shine in a highly structured and distributed environment.
$ npm install harcon
var Harcon = require('harcon')
new Harcon( { /* opts */ } )
// define a listener function listening every message related to "greet" like "greet.goodmorning" or "greet.goodday"
harcon.addict( null, 'peter', 'greet.*', function (greetings1, greetings2, callback) {
callback(null, 'Hi there!')
} )
// define an plain object serving as listener withing the context "greet" to messages "warm"
marie = {
name: 'marie',
context: 'greet',
warm: function (greetings1, greetings2, callback) {
callback( null, 'Bonjour!' )
}
}
harcon.addicts( marie )
// sends a communication 'greet.everyone' with parameters and defines a callback to handle responses
// will receive back 2 answers: 'Hi there!' and 'Bonjour!'
harcon.simpleIgnite( 'greet.everyone', 'Whatsup?', 'How do you do?', function (err, res) {
console.log( err, res )
} )
In an enterprise-level system, one has to realize complex communication structure where lots of entities are following business logic and rules, involving subsystems and external resources, policies and other considerations, in short form: workflows. I take the liberty to define the workflow now as well defined routes and causality of messages. In a workflow, you are not dependent on the response timeframe, workflows manage distance in time and space. The recipient of a message can be on another server or city or planet. Recipient can answer right away or tomorrow or never.
In the JS world, one should mind the introduction of microservices to start with right in the beginning. Just take the advantage of better orchestration, simpler development and debugging, easier deployment, scaling and monitoring.
harcon is "just" a low-level library to leverage such concept. In a simple way, you define entities and the communications among them then publish them.
In a microservice architecture, the key to design and orchestrate a working and living system is to abstract out the control in some form and write business logic fitting that abstract description.
harcon provides the Bender chapter for details, but please learn the basics of harcon before entering the rabbit hole, it is not without a reason that chapter is at the end of the documentation... ;)
When execution is defined by execution-chains, harcon is capable to tell when a given flow really terminated and inform all entities about the termination opening the gate to the distributed transaction management, when all entities participating to a given flow might react to its termination, performing DB operations for example.
Please check chapter Transactions for detail.
In harcon, the communication unit is simply called entity. One can define 2 type of entities:
// Qualified name - will answer to only this message
harcon.addict( null, 'hugh', 'allocate.ip', function (callback) {
callback(null, 'Done.')
} )
// Wildcards - will answer anything within the context greet
harcon.addict( null, 'peter', 'greet.*', function (callback) {
callback(null, 'Done.')
} )
// Regular expression - will answer anything where message name start with string 'job'
harcon.addict( null, 'john', /job.*/, function ( partner, callback) {
callback(null, 'Done.')
} )
...
harcon.simpleIgnite( 'job.order', {name: 'Stephen', customerID:123}, function (err, res) {
console.log( 'Finished.', err, res )
} )
harcon.simpleIgnite( 'john.job', {name: 'Stephen', customerID:123}, function (err, res) {
console.log( 'Finished.', err, res )
} )
var bookKeeper = {
name: 'BookKeeper',
...
newOrder: function ( customer, callback ) {
callback( null, 'Done.' )
},
ordersOfToday: function ( callback ) {
callback( null, [] )
},
searchForOrders: async function ( ) {
return new Promise( (resolve, reject) => {
...
} )
}
}
...
harcon.simpleIgnite( 'BookKeeper.newOrder', {name: 'Stephen', customerID:123}, function (err, res) {
console.log( 'Finished', err, res )
} )
harcon.simpleIgnite( 'BookKeeper.ordersOfToday', function (err, res) {
console.log( 'Finished.', err, res )
} )
The simplest but not the only way to address is to quality entities with their names.
!Note: The following names are strictly forbidden to be used as entity names in any lower- or uppercase version: 'Barrel', 'Bender', 'Blower', 'Communication', 'Fire', 'Firestarter', 'Firestormstarter', 'FireBender', 'Flamestarter', 'FlowBuilder', 'FlowReader', 'Inflicter', 'Mortar', 'Warper'
Those are used by harcon itself.
!Note: Async functions are available only in NodeJS v7+
By default, harcon returns and array of response objects returned by the entities addressed by a message sent.
Let's have 2 simple entities:
harcon.addict( null, 'peter', 'greet.*', function (callback) {
callback(null, 'Hi.')
} )
harcon.addict( null, 'camille', 'greet.*', function (callback) {
callback(null, 'Hello.')
} )
harcon.simpleIgnite( 'greet.simple', function (err, res) {
console.error( err, res )
} )
Returns with the following:
[ 'Hi.', 'Hello.' ]
(Note: for consistency reason, you got an array even if one entity has been addressed.)
In some cases, you might find useful to know which answer comes from which entity. If you add a single parameter to the harcon:
new Harcon( { namedResponses: true } )
The returned object will look like this:
{ peter: 'Hi.', camille: 'Hello.' }
!Note: Harcon support special cases, where all message exchange is performed in 1:1 constellation and the result of a given call is always a single object. Please read Unfolding for details.
Your callback might receive an error object in unwanted situations. The default transport channel of harcon will stop the message processing at the first error occurring as follows:
harcon.addict( null, 'peter', 'greet.*', function (callback) {
callback( new Error('Stay away, please.') )
} )
harcon.addict( null, 'camille', 'greet.*', function (callback) {
callback( new Error('Do not bother me.') )
} )
harcon.simpleIgnite( 'greet.simple', function (err, res) {
console.error( err, res )
} )
will result the following on your console:
[Error: Stay away, please.] null
The default transport layer is designed for development purposes only. In an EE, environment, please mind the introduction of a message queue solution like: AMQP, SQS or MQTT using official plugins: harcon-amqp, harcon-sqs and harcon-mqtt accordingly.
By using a real transport layer, all occurred error messages will be delegated. In such cases, harcon will retrieve an Error object encapsulating all error object received from entities.
You have seen how to call service functions using qualified names and regular expressions (function-based entity). If a much structured system must be orchestrated, a set of finer toolset is at you disposal: contexts and divisions, representing different abstraction levels.
Context: is a named set of object-based entities and contexts. a qualified name identifying the field/purpose the entity is operating. To refine the structure of your service. For example you can have multiple entities providing service like parse but in different context like: "xml" or "json". You can structure your entities like the following:
var parser = {
name: 'JSONParser',
context: 'transfer.json',
parse: function ( document, callback ) {
callback( null, 'Done.' )
}
}
var observer = {
name: 'XMLParser',
context: 'transfer.xml',
parse: function ( document, callback ) {
callback( null, null )
}
}
In this case, such messages can be sent:
harcon.simpleIgnite( 'transfer.json.parse', function (err, res) { ... } )
or
harcon.simpleIgnite( 'transfer.xml.parse', function (err, res) { ... } )
or
harcon.simpleIgnite( 'transfer.'+document.type+'.parse', function (err, res) { ... } )
addressing directly to one of those entities, depending the type of the document you want to parse. Let's define the following entity:
var observer = {
name: 'Observer',
context: 'transfer',
parse: function ( document, callback ) {
callback( null, null )
}
}
Such entity will also receive those message and might do logging or measuring or perform preliminary actions. In short form: you can address multiple entities with a single message. The matching algorithm is simple: spliting the title by '.' matching with the context you specify and fails only if the compared strings are not equals. This means, that:
The title 'transfer.xml.parse' will target the XMLParser and Observer entities.
and
The title 'transfer.parse' will target the XMLParser, JSONParser and Observer entities.
Contexts are very good way to refine your structures and entities and express overlapping functional behaviors.
Division: divisions is a different angle on the plane of orchestration. A division is a closed "box" of entities, meaning that an entity can operate only within the division it is a member of. In fact, every entity belongs to a division defined explicitly or implicitly by the harcon. Divisions can be encapsulated, so a complete division-tree can be built-up in a harcon application. The reason why divisions are important, because it represents a responsibility and/or legal unit. Entities within it (in normal cases) cannot see outside and an entity published to a container division can answer to messages initiated by an entity somewhere lower in the tree. This gives you a control to define surveillance-like or control-like features and much higher complexity of communication-management.
Please find the Divisions chapter for details.
Note: contexts and divisions are not mandatory to be used. The complexity will tell you how to orchestrate your app. It might happen, that simple function-based named entities are fitting your need. Feel free to act on your own, there is no pattern to follow.
If you work with workflows, the sequence/order of your messages will get an importance. To chain messages, define the next point in the workflow you have to add another parameter to your service function:
var order = {
name: 'Order',
context: 'order',
newVPN: function ( customer, ignite, callback ) {
ignite( 'allocate.address', '127.0.0.1', function (err, res) {
callback(err, res)
} )
}
}
...
harcon.simpleIgnite( 'order.newVPN', {name: 'Stephen', customerID:123}, function (err, res) {
console.log( 'Finished', err, res )
} )
That will initiate a small workflow. harcon.simpleIgnite sends a message to entity Order who will send within the same workflow to the Allocator. When it answeres, then the message of the beginning will be answered. harcon will know if you initiate a message within the processing of another one and considers it as part of the ongoing workflow and tracks it. Mind the async execution to keep everything in track!
You are not forced to always expect answer, in some cases a quite entity is desired. If you do not pass a callback when a communication is initiated, harcon will consider it as a one-ways message sending.
// Qualified name - will answer to only this message
harcon.addict( null, 'karl', 'reserve.address', function ( address, callback ) {
// Do something...
callback(null, 'Done.')
} )
...
harcon.simpleIgnite( 'reserve.address', '127.0.0.1' )
Harcon will know, that sender does not expect any answer at all. But, considering that harcon is designed for highly structured and fragmented context, receiver entities always MUST have a callback, to let harcon know about the termination of the given operation for operational prudence!
The need to pass contextual parameters to entities might rise. The options object passed to the constructure of Harcon allows you to specify parameters for entities which will be passed while the init method defined in the entity is called.
harcon = new Harcon( { /* ... */ marie: {greetings: 'Hi!'} } )
var marie = {
name: 'marie',
context: 'test',
init: function (options) {
// {greetings: 'Hi!'} will be passed
}
// services ...
}
There is an option for an object-based entity to enforce uniqueness:
module.exports = {
name: 'Charlotte',
distinguish: 'Unique',
access: function ( callback ) {
callback( null, 'D\'accord?')
}
}
The attribute 'distinguish' can be a boolean or a string which will be added to its name as a postfix. In case of boolean value, a random value will be generated.
The entity can be called with 'Charlotte-Unique.access' or 'Charlotte-1231441123.access' where '1231441123' is the random string generated by the harcon.
This feature is useful in a largely scaled environment, where a given entity might appear in multiple instances and distinguishing them is a must-have behavior.
The entity will be still available with its own name: 'Charlotte'. You can consider it as a synonym or alternative name.
When you create the Harcon instance, you can pass a logger object which will be respected and used to do loggings. If not set, harcon will log everything to the console. So in production, setting up a logging facility ( like winston or bunyan ) is strongly adviced.
harcon = new Harcon( { logger: logger /* ... */ } )
That logger instance will be used as logging facility all over the system, including internal services and entities. Each entity receives a function: harconlog with the signature:
function( err, message, obj, level ) { ... }
That function can be used anything within your entity object:
var Marie = {
name: 'Marie',
context: 'greet',
whiny: function (greetings, callback) {
this.harconlog( null, 'Some logging', { data: greetings }, 'silly' )
callback( null, 'Pas du tout!' )
}
}
That function should be used for any logging activity you need during the flow of your app.
Important: Harcon logs all incoming and outgoing messages at the level 'igniteLevel' set in the configuration file making every-day logging activities unnecessary. You might want to change the default level of this behaviour in order to refine your logging strategy.
harcon = new Harcon( { igniteLevel: 'info' /* ... */ } )
An entity might exclude itself from the logging process by possessing the attribute "concealed":
var Marie = {
name: 'Marie',
concealed: true,
...
}
That will turn off the logging of all messages sent or received by the entity 'Marie'.
Another option is to set the the list of entities unlogged.
harcon = new Harcon( { seals: [ 'Marie' ] /* ... */ } )
That will turn off the logging of all messages sent to Marie. So no answers received by other entities from Marie will be logged.
Every communication exchanged possesses the following properties (not exclusively):
Any time you sends a message or receives an answer, such objects are bypassing through the harcon system which logs and tracks all of them.
By default, harcon uses 32 as length of the IDs which are unique over time and among computer nodes. You can override this default when initiating Harcon
harcon = new Harcon( { /* ... */ idLength: 32 } )
When you defined your components, the need to send and receive messages arises. In a workflow, your component might initiate a message, response one or while responding one sends other ones. The function-based components can perform only the latter 2 cases, cannot initiate anything by its own. This type of components are present to define services, listeners, definitely not serious business entities. As you saw above, the serices functions might possess a parameter before the callback: ignite
var order = {
name: 'Order',
context: 'order',
newVPN: function ( customer, ignite, callback ) {
ignite( 'allocate.address', '127.0.0.1', function (err, res) {
callback(err, res)
} )
}
}
That ignite can be used to chain messages, which means to send messages during the processing of a received one. The tool to initiate sub-workflows.
Of course components are not just reacting entities, they might launch new workflows as well. Object-based components possesses an injected function: ignite and can be used as follows:
var timer = {
name: 'Timer',
scheduling: function ( ) {
this.ignite( 'validate.accounts', function (err, res) {
} )
}
}
That ignite function is injected by the harcon when you publish the components.
In a workflow, a contextual object is much desired to set the context where business entities are participating and share some environmental state or values. If you implement a financial transaction management system, currencies should be added to the terms object making the information accessible to all entities interoperating within the workflow.
Should you define your business functions as below, the terms will be passed and managed by harcon
module.exports = {
name: 'Claire',
init: function (options, callback) {
callback()
},
simple: function (greetings1, greetings2, terms, ignite, callback) {
callback( null, 'Pas du tout!' )
}
}
Along the parameters you need for your business logic, and the ignite function and callback object introduced earlier, the terms can be added to the parameter list in the right order shown above.
module.exports = {
name: 'Domina',
force: function ( terms, ignite, callback ) {
var self = this
terms.tree = 'grow'
ignite( 'Claire.simple', 'It is morning!', 'Time to wake up!', callback )
}
}
The 'tree' attribute set by entity 'Domina' will be seen by the entity 'Claire' when it receives the message.
When you orchastrate your system as a microservice architure over a clustered message bus for example, the individual microservicSystems can be orchastrated into divisions which is a tree structure actually. One can create divisions following the control-flow or responsibility-chain of the application. Every component you deploy will belong to a division. If not declared, then to the system division where all system-level components are put. Divisions is not just a logical grouping of components, but also an encapsulation-model. A component cannot send messages outside the own division but can send to the inner ones. This means, that system components can send to any component, but non-system components cannot reach the level of the main system or other branches of the division-tree.
Divisions give you a very easy-to-use structure to orchestrate your system. Of course, you can use the harcon without using divisions, the complexity of your system will show if you needed it or not.
Let's define components and add them to divisions:
// This will add John to the division 'workers'
harcon.addict( 'workers', 'john', /job.*/, function (callback) {
callback(null, 'Done.')
} )
// This will add Claire to the division 'entrance'
var claire = {
name: 'claire',
division: 'entrance',
context: 'greet',
simple: function (greetings1, greetings2, callback) {
callback(null, 'Enchanté, mon plaisir!')
}
}
Components in a division can be called to:
harcon.ignite( null, 'entrance', 'greet.simple', 'Hi', 'Ca vas?', function (err, res) {
} )
Note: please keep in mind, that harcon.ignite can be and should be used only when you initiate a workflow from outside the harcon!
If you initiate a communication through the harcon instance, it means, that you wants to drop in a message "from outside" which could mean an integration with an external system or just kick off a workflow. There are 2 methods to be called:
ignite and simpleIgnite
The different between them is the parameter list. The later does not require to specify external messageId or division, the prior one does as you can see in the example just above.
External ID is very useful, when the workflow is initiated by some external event possessing an id which must be kept for further logging or tracking or just because a communication harmonization across the complete company.
By default, harcon presumes to have one division per node following the concept of microservices. That division name can be given via its config object or will be derived from its unique name. The created division will serve as the root for every divisions and entities defined.
harcon = new Harcon( { division: 'District 8', ... } )
Entities can be configured easily. The init method of an entity if present, will be called at the end of its publishing process withing the harcon. The function can be used to set up DB connections or anything required for your entities.
module.exports = {
name: 'Claire',
context: 'greet',
init: function (options, callback) {
console.log('Init...', options)
callback()
}
}
The method receives a configuration object and a callback to be called when the init method finishes. That configuration object can be passed when an entity is published:
harcon.addicts( Claire, config )
or even before, when harcon is created.
harcon = new Harcon( { logger: logger, idLength: 32, Claire: {greetings: 'Hi!'} } )
An entity's configuration is an merged object made from the followings (in order):
Should you use none of them, and your entity shall be initiated with empty object.
harcon has an internal handler to deal with messages not answered within a reasonable timeframe.
harcon = new Harcon( { ..., blower: { commTimeout: 2000 } } )
This configuration will tell your running harcon instance to check if the messages sent to some entities have been answered within 2000 millisecs. If not, an error will be sent to the sender with the message: 'Communication has not been received answer withing the given timeframe.'
The nature of your app might urge you to distinguish time management following some business logic. For example: operations involving third party APIs have to be performed with a wider or without time limitation. harcon supports such demand in many ways.
In the simplest case:
harcon = new Harcon( { ..., blower: { commTimeout: 2000, tolerates: [ 'Alizee.superFlegme' ] } } )
The field 'tolerates' can be defined as an array of strings turning off the global timeout management for the communication enclosing the given messages.
The array tolerate might be enlisting strings, regexp values, functions or objects evaluated as follows:
Please find some examples below:
harcon = new Harcon( { ..., blower: { commTimeout: 2000, tolerates: [ 'Alizee.superFlegme' ] } } )
... or ...
harcon = new Harcon( { ..., blower: { commTimeout: 2000, tolerates: [ /Alizee\.\w+/g ] } } )
... or ...
harcon = new Harcon( { ..., blower: { commTimeout: 2000, tolerates: [ function (comm) { return comm.event === 'Alizee.flegme' } ] } } )
... or ...
harcon = new Harcon( { ..., blower: { commTimeout: 2000, tolerates: [ { event: 'Alizee.flegme', timeout: 2000 } ] } } )
You can mix the techniques as your business logic demands.
!Note: Please, keep it mind, that harcon is transport layer agnostic, so a preferred message broker can also provide some security over interaction.
harcon can be easily extended by using pure harcon components listening to system events:
var extension = {
name: 'As you design it',
context: harcon.name,
castOf: function ( name, firestarter ) {
},
affiliate: function ( firestarter ) {
},
close: function () {
}
}
harcon.addicts( extension )
In the current version, the harcon instance you are using will send to your components events about system closing, entity publishing and revoking. For a working example, please check harcon-radiation.
To think in terms of the current trends in the JS-world, the services of harcon might give you back a promise of no callback is provided as last parameter.
harcon.ignite( '0', 'Inflicter.click', 'greet.simple', 'Hi', 'Ca vas?' )
.then( function ( res ) {
})
.catch( function ( reason ) {
} )
There is a "shared" environmental object added to all entities deployed within harcon:
harcon = new Harcon( { millieu: { workDir: '/temp' } } )
That object will be inserted into the initial configuration of all Entities published:
var Marie = {
name: 'Marie',
init: function (options) {
// options = { workDir: '/temp' }
self.options = options
}
}
In any service of entity 'Marie', the 'self.options.workDir' will be a valid object.
!Note: please keep in mind, that some transport layers might restrict the size or the content of the packets relayed. Use millieu object wisely.
There is a reverse-directed-like feature in harcon, allowing an entity to shift "internal state" and let other entities to know about it. Let's say, entities of a system want to know when the DB component becomes "available", or when a third party connector entity becomes "connected".
To build such constellation between entities, is pretty straightforward:
module.exports = {
name: 'Lina',
...
registerForShift: function () {
self.ignite('Marie.notify', 'data', 'Lina.marieChanged', function (err) { if (err) { self.harconlog(err) } } )
}, ...
marieChanged: function ( payload, callback ) {
console.log( '>>>>>>>>>>>', payload )
callback( null, 'OK' )
}
}
In function 'registerForShift' 'Lina' tells 'Marie' to notify her if the state 'data' is changing. The function to be called is 'marieChanged'. The new value of the given state should be passed. State is a string. It does not need to be represented as attribute or whatever, just a pure string - object pair.
How Marie can "shift" state:
module.exports = {
name: 'Marie',
simple: function (greetings1, greetings2, callback) {
this.shifted( { data: 'content' } )
callback(null, 'Bonjour!')
}
}
Anytime Marie calls its internal 'shifted' function, it triggers the harcon's state changing service. You have to pass an object possessing all properties considered to be "states". So an object might shift multiple states at once if needed. The values of the attributes are the payload to send to the listener entities. That this.shifted( { data: 'content' } ) line will initiate an internal communication with the addressing 'Lina.marieChanged' as Lina specified. And the function 'marieChanged' of Lina will be called with the string 'content' passed as payload.
By default harcon blocks acts like a blackbox allowing to exchange communication within the main division / name of the system deployed. If you pass an attribute 'connectedDivisions' to the config of harcon, it will accept communication from the enlisted divisions and allow you to connect to those ones.
The basic concept of harcon is to relay messages between entities where m:n quantity relationship for delivery is much encouraged. That is why, harcon always returns with an array. It is in the nature of Javascript and harcon, that a lib can not decide in advance who will respond to a given message (domains, contextes, auditors, etc.) and if an addressed entity will respond anything.
However, if your system is designed to leverage only 1:1 interactions, harcon can be forced to unfold the array of results and return only the single answer you expect.
let harcon = new Harcon( {
name: harconName,
unfoldAnswer: true,
...
} )
That will unlock the unfolding feature of harcon as follows:
harcon.simpleIgnite( 'camille.simple', function (err, res) {
console.error( 'res will be \'Hello.\' and not [\'Hello.\'] ')
///
} )
Bender is a high-level entity over microservices aims to introduce a simple, yet powerful execution-chain-tool for architects helping to drive message relaying between entities in a regulated environment.
How it works: you define your execution logic which is a set of rules pairing
Let's say you want to see the following informal execution to be performed:
'Order.registerOrder' -> 'DB.storeOrder', 'DB.checkStore', 'Mailer.sendNotif'
The event registerOrder will trigger 3 order messages to be sent. Rolling the execution of your business workflow further. In other way, you can liberate your execution chain from low-level source code. The entity Bender is dedicated to help with you on this road.
One can activate it by using the following configuration:
let harcon = new Harcon( {
name: harconName,
bender: { enabled: true },
FireBender: {
defs: {
// rules here
}
}
...
} )
One has 2 ways to set chain definitions:
Syntax of a rule definition:
{event_name}: {
type: {execution_type},
primers: [ {target} ]
}
{execution_type} : 'series' || 'waterfall' || 'spread'
{target} : { division: {domain_name}, event: {event_name}, skipIf: {condition} } || {event_name}
{condition} : {event_name} || function
Each rule is set of a message, a list of steps to make (called primers), and an optional validation object For example :
'Order.registerOrder': { primers: [ 'DB.storeOrder', 'DB.checkStore', 'Mailer.sendNotif' ], validation: { minlength: '3' } }
or
'Order.registerOrder': { primers: [ 'DB.storeOrder', 'DB.checkStore', 'Mailer.sendNotif' ], validation: function (payload) { return true } }
That works as follows: an entity or external event triggers the message 'registerOrder' sent to entity 'Order'. The call will be performed and the result of that call will be passed as parameter to all messages enlisted in the array 'primer'. So the entity 'DB' will receive the message 'storeOrder' automatically triggered by the message 'Order.registerOrder'. When 'DB' is finished, 'checkStore' will be triggered and so on. Default execution type is 'series' means sequential execution of the given list.
Validation object can be a JS object or a function. It is used to validate the payload of the message 'registerOrder' sent to entity 'Order'. The functions should return a boolean value to tell if the payload is valid. If the validation is an object, the vindication.js will be used for validation.
Bender is turns harcon into a strict message delivery environment, which means if an event is not recognised by Bender or in other words, there is no rule for that given event, an exception will be thrown to the caller and no message delivery will be performed.
There are several type of execution defined by the {execution_type} above.
'Order.registerOrder': { type: 'series', primers: [ 'DB.storeOrder', 'DB.checkStore', 'Mailer.sendNotif' ] }
'Order.registerOrder': { type: 'waterfall', primers: [ 'DB.storeOrder', 'DB.checkStore', 'Mailer.sendNotif' ] }
'Order.registerOrder': { type: 'spread', primers: [ 'DB.storeOrder', 'DB.checkStore', 'Mailer.sendNotif' ] }
The attribute 'skipIf' defines the 'execution-gate' for a given step. It can be a message string or a function. If it is
a string, so a message, the given message will be sent to the referred entity and the answer should possess the "allowed" attribute with a value coercing to true to allow the execution of the given step. Will be ignored otherwise. If error is sent back, all remaining steps will be ignored as well.
a function, the answer coerces to true will allow to perform the step. Will be ignored otherwise. If error is thrown, all remaining steps will be ignored as well.
Attribute 'skipIf' can be used only for 'series' and 'waterfall' as logic follows.
Important: The Bender entity has a message 'completeness' telling which entities or services are still missing breaking the business flow(s) defined.
Note: harcon tries to validate the passed definitions searching for circles or crossing references. The Bender entity will fail to initialise if the validation process on the definitions falls.
When Bender is activated and your flows are well defined through its services, your entities became capable of reacting the termination of your business flows. That serves to handle flows as transactions.
To be straightforward, when a business flow ends, your entities will be notified and can commit DB transactions to finalize their operations.
Each entity you define will possess the following functions:
flowFailed: function ( flowID, errMessage, terms, ignite, callback ) {
return new Promise( (resolve, reject) => {
...
} )
}
flowSucceeded: function ( flowID, result, terms, ignite, callback ) {
return new Promise( (resolve, reject) => {
...
} )
}
By default, it does not do anything. If your entity aims to react to 'flowFailed' or 'flowSucceeded' events, override those functions in the source code of your entity paying attention to the Promise-callback design principle. The function must return a promise and has to allow to pass callback as well...
During your business logic, you business logic should access the flowID through the variable 'terms':
sign: function ( document, terms, ignite, callback ) {
// terms.sourceComm.flowID hold the flowID of the current business flow
// use it to store some records in your favorite in-memory storage to access it when 'flowTerminated' is induced.
}
If you want to use Bender without transaction management, the following configuration will turn it off :
let harcon = new Harcon( {
bender: { enabled: true, igniteTermination: false },
...
} )
Note: Please keep in mind, that if you are using a scaling solution, for the safety of the transactions, harcon does not distinguish the nodes, one has to facilitate the state sharing between nodes using Redis or similar tool.
Some operations cannot be performed within a reasonable short timeframe. Let's say, a task is sent to another devision requesting for manual acknowledgement. This will make a gap between request and response for sure. Could be proven the measure of days as well...
The design you app should follow for such cases as follows: The entity receiving the request must send back 'Request accepted'-like message to the sender and store the externalID and flowID of the communication:
ignite( 'ManCanDo.sign', document, function () {
console.log('Accepted.')
} )
...
let ManCanDo = {
auditor: true,
sign: function ( document, terms, ignite, callback ) {
database.store( terms.sourceComm, document, callback )
},
_signed: function( document ) {
let self = this
database.readRequest( document, function(err, record){
self.ignite( record.externalId, record.flowId, record.sourceDivision, record.source + '.signed', document )
} )
}
}
The param terms holds references about the incoming communication by the name of sourceComm. You can store the externalID if the communication has been initiated by an external party or just the flowID if continuity must be ensured.
There is a built-in entity in harcon: Mortar dedicated to
To activate it:
harcon = new Harcon( { ..., mortar: { enable: true, folder: '', liveReload: false } } )
On the line to ease the management of workflows, harcon introduces the erupt function appearing in the published entities and in the terms object passed to the services functions during ongoing communication. It represents a simple trick to call ignite by returning the following:
function (waterfall) {
return waterfall ? function (_obj, cb) { ... } : function (cb) { ... }
}
In case, you are building up a complete chain of calls, you can use your favorite functional abstraction tool as the following example demonstrates through the popular async.js:
async.series([
erupt()( 'Marie.greet', 'Hello' ),
erupt()( 'Julie.greet', 'Hello' )
], function(err, res){ })
async.waterfall([
erupt()( 'Marie.greet', 'Hello' ),
erupt(true)( 'Julie.greet', 'Hello' )
], function(err, res){ })
No need to pass callback, the function returned by erupt call will possess an injected callback function as parameter used as THE callback of the communication. Pretty nasty, huh?
(The MIT License)
Copyright (c) 2017 Imre Fazekas
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
See https://github.com/imrefazekas/harcon/issues.
FAQs
Messaging/Service Bus for the harmonic convergence of node-based enterprise entities.
We found that harcon demonstrated a not healthy version release cadence and project activity because the last version was released a year ago. It has 1 open source maintainer collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Research
Security News
Socket’s threat research team has detected six malicious npm packages typosquatting popular libraries to insert SSH backdoors.
Security News
MITRE's 2024 CWE Top 25 highlights critical software vulnerabilities like XSS, SQL Injection, and CSRF, reflecting shifts due to a refined ranking methodology.
Security News
In this segment of the Risky Business podcast, Feross Aboukhadijeh and Patrick Gray discuss the challenges of tracking malware discovered in open source softare.