happn

HAPPN-3 released!

happn version 3 has been released, the repo is here - this version of happn has a bunch of maintainability updates whereby the code has been better structured. The intra-process client is now using the same code as the websockets one, now just spoofs the primus spark. There is also now notably a protocol abstraction layer, so we can now use different protocols to interact with the happn database and subscriptions, such as an MQTT plugin etc.

Introduction

Happn is a mini database combined with pub/sub, the system stores json objects on paths. Paths can be queried using wildcard syntax. The happn client can run in the browser or in a node process. Happn clients can subscribe to events on paths, events happn when data is changed by a client on a path, either by a set or a remove operation.

Happn stores its data in a collection called 'happn' by default on your mongodb/nedb. The happn system is actually built to be a module, this is because the idea is that you will be able to initialize a server in your own code, and possibly attach your own plugins to various system events.

A paid for alternative to happn would be firebase

Technologies used: Happn uses Primus to power websockets for its pub/sub framework and mongo or nedb depending on the mode it is running in as its data store, the API uses connect. nedb as the embedded database, although we have forked it happn's purposes here

Getting started

npm install happn

You need NodeJS and NPM of course, you also need to know how node works (as my setup instructions are pretty minimal) To run the tests, clone the repo, npm install then npm test:

git clone https://github.com/happner/happn.git
npm install
npm test

But if you want to run your own service do the following: Create a directory you want to run your happn in, create a node application in it - with some kind of main.js and a package.json

In node_modules/happn/test in your folder, the e2e_test.js script demonstrates the server and client interactions shown in the following code snippets

starting service:

The service runs on port 55000 by default - the following code snippet demonstrates how to instantiate a server.

var happn = require('happn')
var happnInstance; //this will be your server instance
 
//we are using a compact default config here, port defaults to 55000
 
 happn.service.create({
  utils: {
    logLevel: 'error',
    // see happn-logger module for more config options
  }
},
function (e, happn) {
  if (e)
    return callback(e);
 
  happnInstance = happn; //here it is, your server instance
  happnInstance.log.info('server up');
 
});
 

In your console, go to your application folder and runnode mainyour server should start up and be listening on your port of choice.

Connecting to Happn

Using node:

var happn = require('happn');
var my_client_instance; //this will be your client instance
 
happn.client.create([options], function(e, instance) {
 
    //instance is now connected to the server listening on port 55000
    my_client_instance = instance;
 
});
 

To use the browser client, make sure the server is running, and reference the client javascript with the url pointing to the running server instances port and ip address like so:

<script type="text/javascript" src="http://localhost:55000/browser_client"></script>
<script>
 
var my_client_instance;
 
HappnClient.create([options], function(e, instance) {
 
    //instance is now connected to the server listening on port 55000
    my_client_instance = instance;
 
});
 
</script> 

SET

Puts the json in the branch e2e_test1/testsubscribe/data, creates the branch if it does not exist

 
//the noPublish parameter means this data change wont be published to other subscribers, it is false by default
//there are a bunch other parameters - like noStore (the json isnt persisted, but the message is published)
 
my_client_instance.set('e2e_test1/testsubscribe/data/', {property1:'property1',property2:'property2',property3:'property3'}, {noPublish:true}, function(e, result){
 
    //your result object has a special _meta property (not enumerable) that contains its actual _id, path, created and modified dates
    //so you get back {property1:'property1',property2:'property2',property3:'property3', _meta:{path:'e2e_test1/testsubscribe/data/', created:20151011893020}}
 
 
});
 

NB - by setting the option merge:true, the data at the end of the path is not overwritten by your json, it is rather merged with the data in your json, overwriting the fields you specify in your set data, but leaving the fields that are already at that branch.

SET SIBLING

sets your data to a unique path starting with the path you passed in as a parameter, suffixed with a random short id

    my_client_instance.setSibling('e2e_test1/siblings', {property1:'sib_post_property1',property2:'sib_post_property2'}, function(e, results){
        //you get back {property1:'sib_post_property1',property2:'sib_post_property2', _meta:{path:'e2e_test1/siblings/yCZ678__'}}
        //you would get all siblings by querying the path e2e_test1/siblings*

GET

Gets the data living at the specified branch

my_client_instance.get('e2e_test1/testsubscribe/data',
    null, //options
    function(e, results){
    //results is your data, if you used a wildcard in your path, you get back an array
    //if you used an explicit path, you get back your data as the object on that path
 

You can also use wildcards, gets all items with the path starting e2e_test1/testsubscribe/data

my_client_instance.get('e2e_test1/testsubscribe/data*',
    null,
    function(e, results){
    //results is your data
    results.map(function(item){
 
    });

You can also just get paths, without data

my_client_instance.getPaths('e2e_test1/testwildcard/*', function(e, results){

SEARCH

You can pass mongo style search parameters to look for data sets within specific key ranges

 
    var options = {
      fields: {"name": 1},
      sort: {"name": 1},
      limit: 1
    }
 
    var criteria = {
      $or: [{"region": {$in: ['North', 'South', 'East', 'West']}},
        	{"town": {$in: ['North.Cape Town', 'South.East London']}}],
      "surname": {$in: ["Bishop", "Emslie"]}
    }
 
    publisherclient.get('/users/*', {
        criteria: criteria,
        options: options
      },
      function (e, search_results) {
      	//and your results are here
      	search_results.map(function(user){
      		if (user.name == 'simon')
      			throw new Error('stay away from this chap, he is dodgy');
      	});
      }
    );
 

DELETE

deletes the data living at the specified branch

    my_client_instance.remove('/e2e_test1/testsubscribe/data/delete_me', null, function(e, result){
    if (!e)
        //your item was deleted, result.payload is an object that lists the amount of objects deleted

EVENTS

you can listen to any SET & REMOVE events happening in your data - you can specifiy a path you want to listen on or you can listen to all SET and DELETE events using a catch-all listener

Specific listener:

my_client_instance.on('/e2e_test1/testsubscribe/data/delete_me', //the path you are listening on
                    {event_type:'remove', // either set, remove or all - defaults to all
                     count:0},// how many times you want your handler to handle for before it is removed - default is 0 (infinity)
                    function(//your listener event handler
                        message, //the actual object data being set or removed
                        meta){ //the meta data - path, modified,created _id etc.
 
 
                    },
                    function(e){
                        //passes in an error if you were unable to register your listener
                    });

Catch all listener:

my_client_instance.onAll(function(//your listener event handler
                        message, //the actual object data being set or removed
                        meta){ //the meta data - path, modified,created _id, also tells you what type of operation happened - ie. GET, SET etc.
                    },
                    function(e){
                        //passes in an error if you were unable to register your listener
                    });
 

EVENT DATA

• you can grab the data you are listening for immediately either by causing the events to be emitted immediately on successful subscription or you can have the data returned as part of the subscription callback using the initialCallback and initialEmit options respectively*

//get the data back as part of the subscription callback
listenerclient.on('/e2e_test1/testsubscribe/data/values_on_callback_test/*', 
  {"event_type": "set", 
  "initialCallback":true //set to true, causes data to be passed back
  }, function (message) {
 
          expect(message.updated).to.be(true);
          callback();
 
        }, function(e, reference, response){
          if (e) return callback(e);
          try{
 
            //the response is your data, ordered by modified - will always be in an array even if only one or none is found
 
            expect(response.length).to.be(2);
            expect(response[0].test).to.be('data');
            expect(response[1].test).to.be('data1');
 
            listenerclient.set('/e2e_test1/testsubscribe/data/values_on_callback_test/1', {"test":"data", "updated":true}, function(e){
              if (e) return callback(e);
            });
 
          }catch(e){
            return callback(e);
          }
        });
 

//get the data emitted back immediately
 
listenerclient.on('/e2e_test1/testsubscribe/data/values_emitted_test/*', 
  {"event_type": "set", 
  "initialEmit":true //set to true causes emit to happen on successful subscription
  }, function (message, meta) {
          //this emit handler runs immediately
          caughtEmitted++;
 
          if (caughtEmitted == 2){
            expect(message.test).to.be("data1");
            callback();
          }
 
 
        }, function(e){
          if (e) return callback(e);
        });
 

UNSUBSCRIBING FROM EVENTS

//use the .off method to unsubscribe from a specific event (the handle is returned by the .on callback) or the .offPath method to unsubscribe from all listeners on a path:

 
 var currentListenerId;
    var onRan = false;
    var pathOnRan = false;
 
    listenerclient.on('/e2e_test1/testsubscribe/data/on_off_test', {event_type: 'set', count: 0}, function (message) {
 
      if (pathOnRan) return callback(new Error('subscription was not removed by path'));
      else pathOnRan = true;
 
      //NB - unsubscribing by path
      listenerclient.offPath('/e2e_test1/testsubscribe/data/on_off_test', function (e) {
 
        if (e)
          return callback(new Error(e));
 
        listenerclient.on('/e2e_test1/testsubscribe/data/on_off_test', {event_type: 'set', count: 0},
          function (message) {
            if (onRan) return callback(new Error('subscription was not removed'));
            else {
              onRan = true;
              //NB - unsubscribing by listener handle
              listenerclient.off(currentListenerId, function (e) {
                if (e)
                  return callback(new Error(e));
 
                publisherclient.set('/e2e_test1/testsubscribe/data/on_off_test', {"test":"data"}, function (e, setresult) {
                  if (e) return callback(new Error(e));
                  setTimeout(callback, 2000);
                });
              });
            }
          },
          function (e, listenerId) {
          
            //NB - listener id is passed in on the .on callback
          
            if (e) return callback(new Error(e));
 
            currentListenerId = listenerId;
 
            publisherclient.set('/e2e_test1/testsubscribe/data/on_off_test', {"test":"data"}, function (e, setresult) {
              if (e) return callback(new Error(e));
            });
          });
      });
 
    }, function (e, listenerId) {
      if (e) return callback(new Error(e));
 
      currentListenerId = listenerId;
 
      publisherclient.set('/e2e_test1/testsubscribe/data/on_off_test', {"test":"data"}, function (e) {
        if (e) return callback(new Error(e));
      });
 
    });
 

TAGGING

You can do a set command and specify that you want to tag the data at the path. Tagging will take a snapshot of the data as it currently stands, and will save the snapshot to a new path in /_TAGS

 
my_client_instance.set('path/with/existing/data', null, {tag:'tagName'}, function(e, result) {});
my_client_instance.get('_TAGS/path/with/existing/data/*', function(e, result) {});
 

MERGING

you can do a set command and specify that you want to merge the json you are pushing with the existing dataset, this means any existing values that are not in the set json but exist in the database are persisted

 
my_client_instance.set('e2e_test1/testsubscribe/data/', {property1:'property1',property2:'property2',property3:'property3'}, {merge:true}, function(e, result){
 
});
 

SECURITY SERVER

happn server instances can be secured with user and group authentication, a default user and group called _ADMIN is created per happn instance, the admin password is 'happn' but is configurable (MAKE SURE PRODUCTION INSTANCES DO NOT RUN OFF THE DEFAULT PASSWORD)

 
var happn = require('happn')
var happnInstance; //this will be your server instance
 
happn.service.create({secure:true, adminUser:{password:'testPWD'}},
function (e, instance) {
 
  if (e)
    return callback(e);
 
  happnInstance = instance; //here it is, your server instance
 
});
 
 

at the moment, adding users, groups and permissions can only be done by directly accessing the security service, to see how this is done - please look at the functional test for security

SECURITY CLIENT

the client needs to be instantiated with user credentials and with the secure option set to true to connect to a secure server

 
//logging in with the _ADMIN user
 
var happn = require('happn');
happn.client.create({config:{username:'_ADMIN', password:'testPWD'}, secure:true},function(e, instance) {
 
 

SECURITY PROFILES

profiles can be configured to fit different session types, profiles are ordered sets of rules that match incoming sessions with specific policies, the first matching rule in the set is selected when a session is profiled, so the order they are configured in the array is important

 
//there are 2 default profiles that exist in secure systems - here is an example configuration 
//showing how profiles can be configured for a service:
 
 var serviceConfig = {
    services:{
      data:{
 
      },
      security: {
        config: {
          sessionTokenSecret:"TESTTOKENSECRET",
          keyPair: {
            privateKey: 'Kd9FQzddR7G6S9nJ/BK8vLF83AzOphW2lqDOQ/LjU4M=',
            publicKey: 'AlHCtJlFthb359xOxR5kiBLJpfoC2ZLPLWYHN3+hdzf2'
          },
          profiles:[ //profiles are in an array, in descending order of priority, so if you fit more than one profile, the top profile is chosen
            {
              name:"web-session",
              session:{
                $and:[{
                  user:{username:{$eq:'WEB_SESSION'}},
                  type:{$eq:0}
                }]
              },
              policy:{
                ttl: "4 seconds",//4 seconds = 4000ms, 4 days = 1000 * 60 * 60 * 24 * 4, allow for hours/minutes
                inactivity_threshold:2000//this is costly, as we need to store state on the server side
              }
            }, {
              name:"rest-device",
              session:{
                $and:[{ //filter by the security properties of the session - check if this session user belongs to a specific group
                user:{groups:{
                  "REST_DEVICES" : { $exists: true }
                }},
                type:{$eq:0} //token stateless
              }]},
              policy: {
                ttl: 2000//stale after 2 seconds
              }
            },{
              name:"trusted-device",
              session:{
                $and:[{ //filter by the security properties of the session, so user, groups and permissions
                user:{groups:{
                  "TRUSTED_DEVICES" : { $exists: true }
                }},
                type:{$eq:1} //stateful connected device
              }]},
              policy: {
                ttl: 2000,//stale after 2 seconds
                permissions:{//permissions that the holder of this token is limited, regardless of the underlying user
                  '/TRUSTED_DEVICES/*':{actions: ['*']}
                }
              }
            },{
              name:"specific-device",
              session:{$and:[{ //instance based mapping, so what kind of session is this?
                type:{$in:[0,1]}, //any type of session
                ip_address:{$eq:'127.0.0.1'}
              }]},
              policy: {
                ttl: Infinity,//this device has this access no matter what
                inactivity_threshold:Infinity,
                permissions:{//this device has read-only access to a specific item
                  '/SPECIFIC_DEVICE/*':{actions: ['get','on']}
                }
              }
            },
            {
              name:"non-reusable",
              session:{$and:[{ //instance based mapping, so what kind of session is this?
                user:{groups:{
                  "LIMITED_REUSE" : { $exists: true }
                }},
                type:{$in:[0,1]} //stateless or stateful
              }]},
              policy: {
                usage_limit:2//you can only use this session call twice
              }
            }, {
              name:"default-stateful",// this is the default underlying profile for stateful sessions
              session:{
                $and:[{type:{$eq:1}}]
              },
              policy: {
                ttl: Infinity,
                inactivity_threshold:Infinity
              }
            }, {
              name:"default-stateless",// this is the default underlying profile for ws sessions
              session:{
                $and:[{type:{$eq:0}}]
              },
              policy: {
                ttl: 60000 * 10,//session goes stale after 10 minutes
                inactivity_threshold:Infinity
              }
            }
          ]
        }
      }
    }
  };
 

the test that clearly demonstrates profiles can be found here

the default policies look like this:

//stateful - so ws sessions:
{
    name:"default-stateful",// this is the default underlying profile for stateful sessions
    session:{
      $and:[{type:{$eq:1}}]
    },
    policy: {
      ttl: 0, //never goes stale
      inactivity_threshold:Infinity
    }
  }
  
  
//stateless - so token based http requests (REST)
{
    name:"default-stateless",// this is the default underlying profile for stateless sessions (REST)
    session:{
      $and:[{type:{$eq:0}}]
    },
    policy: {
      ttl: 0, //never goes stale
      inactivity_threshold:Infinity
    }
  }
 

NB NB - if no matching profile is found for an incoming session, one of the above is selected based on whether the session is stateful or stateless, there is no ttl or inactivity timeout on both policies - this means that tokens can be reused forever (unless the user in the token is deleted) rather push to default polcies to your policy list which would sit above these less secure ones, with a ttl and possibly inactivity timeout

WEB PATH LEVEL SECURITY

the http/s server that happn uses can also have custom routes associated with it, when the service is run in secure mode - only people who belong to groups that are granted @HTTP permissions that match wildcard patterns for the request path can access resources on the paths, here is how we grant permissions to paths:

 
var happn = require('happn')
var happnInstance; //this will be your server instance
 
happn.service.create({secure:true, adminUser:{password:'testPWD'}},
function (e, instance) {
 
  if (e)
    return callback(e);
 
  	happnInstance = instance; //here it is, your server instance
 
  	var testGroup = {
      name:'TEST GROUP',
      custom_data:{
        customString:'custom1',
        customNumber:0
      }
    }
 
    testGroup.permissions = {
        '/@HTTP/secure/route/*':{actions:['get']},//NB - we can wildcard the path
        '/@HTTP/secure/another/route/test':{actions:['put','post']}//NB - actions confirm to http verbs
    };
 
    happnInstance.services.security.upsertGroup(testGroup, {}, function(e, group){
 
        //our group has been upserted with the right permissions
 
        //this is how we add custom routes to the service, these routes are both available to users who belong to the 'TEST GROUP' group or the _ADMIN user (who has permissions to all routes)
 
        happnInstance.connect.use('/secure/route/test', function(req, res, next){
 
            res.setHeader('Content-Type', 'application/json');
            res.end(JSON.stringify({"secure":"value"}));
 
        });
 
        happnInstance.connect.use('/secure/another/route/test', function(req, res, next){
 
            res.setHeader('Content-Type', 'application/json');
            res.end(JSON.stringify({"secure":"value"}));
 
        });
 
 
    });
 
});
 

logging in with a secure client gives us access to a token that can be used, either by embedding the token in a cookie called happn_token or a query string parameter called happn_token, if the login has happened on the browser, the happn_token is autmatically set by default

 
//logging in with the _ADMIN user, who has permission to all web routes
 
var happn = require('happn');
happn.client.create({config:{username:'_ADMIN', password:'testPWD'}, secure:true},function(e, instance) {
 
    //the token can be derived from instance.session.token now
 
    //here is an example of an http request using the token:
 
    var http = require('http');
 
    var options = {
        host: '127.0.0.1',
      	port:55000,
      	path:'/secure/route/test'
    }
 
    if (use_query_string)
      	options.path += '?happn_token=' + instance.session.token;
    else
    	options.headers = {'Cookie': ['happn_token=' + instance.session.token]}
 
    http.request(options, function(response){
 
    	//response.statusCode should be 200;
 
    }).end();
 
 
});
 
 

HTTPS SERVER

happn can also run in https mode, the config has a section called transport

 
//cert and key defined in config
 
var config = {
  	transport:{
    	mode:'https',
    	cert: '-----BEGIN CERTIFICATE-----\n[CERT ETC...]\n-----END CERTIFICATE-----',
    	key: '-----BEGIN RSA PRIVATE KEY-----\n[KEY ETC...]\n-----END RSA PRIVATE KEY-----'
  	}
}
 
// or cert and key file paths defined in config
// IF BOTH OF THESE FILES DONT EXIST, THEY ARE AUTOMATICALLY CREATED AS SELF SIGNED
 
var config = {
    transport:{
    	mode:'https',
    	certPath:'home/my_cert.pem',
    	keyPath:'home/my_key.rsa'
    }
}
 
// or have the system create a cert and key for you, in the home directory of the user that started the happn process - called .happn-https-cert and .happn-https-key
 
var config = {
    transport:{
    	mode:'https'
    }
}
 
var happn = require('../lib/index')
var service = happn.service;
var happnInstance; //this will be your server instance
 
//create the service here - now in https mode - running over the default port 55000
 
service.create(config ...
 

HTTPS CLIENT

NB - the client must now be initialized with a protocol of https, and if it is the node based client and the cert and key file was self signed, the allowSelfSignedCerts option must be set to true

 
var happn = require('happn');
 
happn.client.create({config:{protocol:'https', allowSelfSignedCerts:true}},function(e, instance) {
...
 

PAYLOAD ENCRYPTION

if the server is running in secure mode, it can also be configured to encrypt payloads between it and socket clients, this means that the client must include a keypair as part of its credentials on logging in, to see payload encryption in action plase go to the following test

PUBSUB MIDDLEWARE

incoming and outgoing packets delivery can be intercepted on the server side, this is how payload encryption works, to add a custom middleware you need to add it to the pubsub service's configuration, a middleware must adhere to a specific interface, as demonstrated below:

 
 
var testMiddleware = {
 
  incomingCount:0,
  outgoingCount:0,
 
  incoming:function(packet, next){
    //modify incoming packet here
    packet.modified = true;
    this.incomingCount++;
    next();
  },
 
  outgoing:function(packet, next){
    //modify outgoing packet here
    packet.modified = true;
    this.outgoingCount++;
    next();
  }
};
 
var happn_service = happn.service;
var test_client = happn.client;
 
var testConfig = {
  secure: true,
  port:44445,
  services:{
    pubsub:{
      config:{
        transformMiddleware:[{instance:testMiddleware}]//middelware added in the order it is required to run in
                                                      // either as an instance or as a path {path:'my-middleware-module'}
                                                      // path style middlewares are instantiated using require and new
      }
    }
  }
};
 
service.create(testConfig,
 
  function (e, happnInst) {
    if (e)
      return callback(e);
 
    serviceInst = happnInst;
 
    happn_client.create({
      config: {
        port:44445,
        username: '_ADMIN',
        password: 'happn'
      },
      info:{
        from:'startup'
      }
    }, function (e, instance) {
 
      if (e) return callback(e);
 
      clientInst = instance;
 
      //the login of the client generated traffic
      expect(testMiddleware.incomingCount > 0).to.be(true);
      expect(testMiddleware.outgoingCount > 0).to.be(true);
 
      clientInst.disconnect(function(){
        serviceInst.stop(callback);
      });
    });
  }
);
 

TESTING WITH KARMA

testing payload encryption on the browser: gulp --gulpfile test/test-browser/gulp-01.js

OTHER PLACES WHERE HAPPN IS USED:

HAPPNER - an experimental application engine that uses happn for its nervous system, see: www.github.com/happner/happner