DRY models for client and server
npm install invisibleInvisible is a JavaScript library that leverages
browserify to achieve the Holy Grail of web programming:
model reuse in the client and the server.
Install with npm:
```
npm install invisible
Wire up Invisible into your express (Works only with Express 4.x) app:
`javascript
var express = require("express");
var path = require("path");
var invisible = require("invisible");
var app = express();
app.use(invisible.router({
rootFolder: path.join(__dirname, 'models')
}));
app.listen(3000);
`
To make your models available everywhere, define them and call Invisible.createModel.
`javascript
// models/person.js
var Invisible = require("invisible");
var crypto = require("crypto");
var _s = require("underscore.string");
function Person(firstName, lastName, email){
this.firstName = firstName;
this.lastName = lastName;
this.email = email;
}
Person.prototype.fullName = function(){
return this.firstName + ' ' + this.lastName;
}
Person.prototype.getAvatarUrl = function(){
cleanMail = _s.trim(this.email).toLowerCase();
hash = crypto.createHash("md5").update(cleanMail).digest("hex");
return "http://www.gravatar.com/avatar/" + hash;
}
module.exports = Invisible.createModel("Person", Person);
`
Now your models will be available under the Invisible namespace. Require as usual in the server:
`javascript`
var Invisible = require("invisible")
var john = new Invisible.models.Person("John", "Doe", "john.doe@mail.com");
john.fullName(); //John Doe
In the client, just add the invisible script:
`html`
Invisible.js uses browserify to expose you server defined
models in the browser, so you can use any broserify-able library to implement them. Note that this
integration is seamless, no need to build a bundle, Invisible.js does that for you on the fly.
In addition to making your models available everywhere, Invisible extends them with methods to handle
persistence. In the server this means interacting with MongoDB and in the client making requests to an
auto-generated REST API, that subsequently performs the same DB action.
javascript
jane.save(function(err, result){
if (err){
console.log("something went wrong");
} else {
console.log("Jane's id is " + jane._id);
console.log("which equals " + result._id);
}
})
`
The first time the save method is called, it creates the model in the database and sets its _id attribute.
Subsequent calls update the model. Validations are called upon saving, see the validations section for details.Note that a full Invisible model instance is passed to the callback, and the calling instance is also updated
when the process is done.
$3
`javascript
jane.delete(function(err, result){
if (err){
console.log("something went wrong");
} else {
console.log("Jane is no more.");
}
})
`
$3
`javascript
Invisible.Person.query(function(err, results){
if (err){
console.log("something went wrong");
} else {
console.log("Saved persons are:");
for (var i = 0; i < results.length; i++){
console.log(results[i].fullName());
}
}
});Invisible.Person.query({firstName: "Jane"}, function(err, results){
if (err){
console.log("something went wrong");
} else {
console.log("Persons named Jane are:");
for (var i = 0; i < results.length; i++){
console.log(results[i].fullName());
}
}
});
Invisible.Person.query({}, {sort: "lastName", limit: 10}, function(err, results){
if (err){
console.log("something went wrong");
} else {
console.log("First 10 persons are:");
for (var i = 0; i < results.length; i++){
console.log(results[i].fullName());
}
}
});
`Queries the database for existent models. The first two optional arguments correspond to the
Query object
and Query options
in the MongoDB Node.JS driver. The one difference is that when using ids you can pass strings,
that will be converted to ObjectID when necessary.
$3
`javascript
Invisible.Person.findById(jane._id, function(err, model){
if (err){
console.log("something went wrong");
} else {
console.log("Jane's name is " + model.firstName);
console.log("But we knew that!");
}
})
`Looks into the database for a model with the specified
_id value. As in the query method, you can pass
a string id instead of an ObjectID instance.Validations
Invisible.js integrates with revalidator to handle model validations.
A
validate method is added to each model which looks for a defined validation schema, and is executed each time
a model is saved, both in the client and the server. For example:`javascript
function Person(email){
this.email = email;
}Person.prototype.validations = {
properties: {
email: {
format: 'email',
required: true
}
}
}
var john = new Person("john.doe@none.com");
john.save(function(err, result){
console.log("All OK here.");
});
john.email = "invalid";
john.save(function(err, result){
console.log(err);
/* Prints: {valid: false, errors: [{attribute: 'format',
property: 'email', message: 'is not a valid email'}]} */
})
`Invisible.js also introduces "method" validations, which allow you to specify a method which should be called
in the validation process. This way asynchronic validations, such as querying the database, can be performed:
`javascript
Person.prototype.validations = {
methods: ['checkUnique']
}Person.prototype.checkUnique = function(cb) {
Invisible.Person.query({email: this.email}, function(err, res){
if (res.length > 0){
cb({valid: false, errors: ["email already registered"]});
} else {
cb({valid: true, errors: []});
}
});
}
`The custom validation method takes a callback and should call it with the return format of revalidator: an object
with a "valid" boolean field and an "errors" list. Note that the method validations are only called if the
properties validations succeed, and stop the validation process upon the first failure.
Real time events
Invisible.js uses socket.io to emmit an event whenever something changes for a model, and lets you add listener
functions to react to those changes in realtime.
To add realtime features:
`javascript
var server = app.listen(3000);
invisible.addRealtime(server);
`And then in your code:
`javascript
Invisible.Person.on('new', function(model){
console.log(model.fullName() + " has been created");
});Invisible.Person.on('update', function(model){
console.log(model.fullName() + " has been updated");
});
Invisible.Person.on('delete', function(model){
console.log(model.fullName() + " has been deleted");
});
`Authentication
Invisible.js provides a default method to authenticate the requests to the REST API, based on OAuth2's Resource
Owner Password flow. This means than when activating authentication, a route is exposed that exchanges user
credentials for a request token used to sign the rest of the API calls. Unsigned calls will get a 401 response.
To use authentication, you must first define a user model in whatever way you like; the only constraint is that
you must be able to identify a user with a pair of credentials such as username/password. By default the
User
name is assumed for the model, but this can be overriden via the userModel configuration.`javascript
//models/user.js
function User(email){
this.email = email;
}User.prototype.setPassword = function(rawPassword){
this.hashedPassword = someHashFunction(rawPassword);
}
User.prototype.isPassword = function(rawPassword){
return this.hashedPassword == someHashFunction(rawPassword);
}
module.exports = Invisible.createModel("User", User);
`Once the User model is defined, authentication is activated by defining an
authenticate function in the
configuration, that takes the credentials and returns the authenticated user:`javascript
//auth.js
module.exports = function authenticateUser(email, password, done){ Invisible.User.query({email: email}, function(err, users){
if (err) {
return done(err);
}
if (users.length < 1){
return done(null, false);
}
var user = users[0];
if (!user.isPassword(password)) {
return done(null, false);
}
return done(null, user);
});
}
//app.js
//...express and invisible configurations...
auth = require("./auth");
app.use(invisible.router({
rootFolder: path.join(__dirname, 'models'),
authenticate: auth
}));
`Optionally, an
authExpiration configuration can be included to specify the amount of seconds the acess token
can be used before requiring a refresh. The token refresh is managed seamlessly by the client models.In order for the client model to get an access token to sign its requests, a
login function must be called when
the user enters his credentials; note that these are used for the exchange and not stored for further use:`javascript
//...get credentials from login form...
Invisible.login(email, password, function(err){
if(err){
console.log("Invalid credentials!");
} else {
console.log("User logged, requests signed");
}
});
`Once login is successful, the calls to the REST API will be allowed to the client models. A
logout method is also
provided to drop the tokens from being used in further model requests.The only endpoint which does not require a signed request is the POST to the user model, to allow user registration.
Authorization
Once you can identify the user making requests, you'll usually want to establish what he can and can't do with the data.
The models provide hooks to authorize a user to call its methods: allowCreate, allowUpdate, allowFind and allowDelete. All of them take the user instance and should callback telling if the method is allowed to that user:`javascript
function Message(from_id, to_id, text){
this.from_id = from_id;
this.to_id = to_id;
this.text = text;
}Message.prototype.allowCreate = function(user, cb) {
//a user can only create messages sent by him
return cb(null, this.from_id === user._id);
}
Message.prototype.allowUpdate = function(user, cb) {
//a user can only update messages sent by him
return cb(null, this.from_id === user._id;
}
Message.prototype.allowFind = function(user, cb) {
//a user can only get messages sent by him or to him
return cb(null, this.from_id === user._id || this.to_id === user._id;
}
Message.prototype.allowDelete = function(user, cb) {
//a user cannot delete messages
return cb(null, false);
}
module.exports = Invisible.createModel("Message", Message);
`
Similarly, allowEvents can be used to decide if a user is authorized to receive a real time update on a model:
`javascript
Message.prototype.allowEvents = function(user, cb) {
//only sent events when a message is sent to the user
return cb(null, this.to_id === user._id;
}
`Another hook,
baseQuery, is available to restrict what segment of the database the user should have access to.
It also takes the user, and callbacks with a criteria object like the one for the query method. This base criteria,
is and-ed with the criteria used in query to filter out unauthorized data. Following the previous example:`javascript
Message.baseQuery = function(user, cb){
//only expose message sent to or by the user
return cb(null, {$or: [{from: user._id}, {to: user._id}]})
}
`
Now when calling Invisible.Message.query` in the client, only messages sent by and to the logged user will be