model

Datastore-agnostic ORM in JavaScript

Model

Model is a datastore-agnostic ORM in JavaScript. It serves as the model-component for the Geddy MVC Web framework for NodeJS.

Model currently implements adapters for:

  • Postgres
  • MySQL
  • SQLite
  • Riak
  • MongoDB
  • LevelDB
  • In-memory
  • Filesystem

Apache License, Version 2

Model requires version 0.8.x of Node.js or higher. If you want to run the tests, or work on Model, you'll want the Jake JavaScript build-tool.

Installing with NPM

NPM install model

Here's a minimal example which uses the LevelDB adapter for a Foo model:

var model = require('model');
 
// Setup the blueprint of the model. This is where you can 
// setup the model properties, defaults, and validations 
var Foo = function () {
  this.setAdapter('level', {
    db: './data'
  });
 
  // Define the whitelisted properties on the model. 
  // Properties not listed wont be saved 
  this.defineProperties({
    name: { type: 'string', required: true },
    description: { type: 'text' },
    enabled: { type: 'boolean' },
    archived: { type: 'boolean' },
  });
};
 
// This registers the model with the model package so 
// things like associations can work 
Foo = model.register('Foo', Foo);
 
// Now we export Foo to create a reusable model module 
module.exports = Foo;

You can then use it like the following example:

var Foo = require('./foo.js');
Foo.first(1, function (errmodel) {
  // Check if there was an error with the DB 
  if (err) throw new Error('Uh oh, something broke');
 
  // If there was no error, but no model was found it must be missing 
  if (!err && !model) throw new Error('Foo not found');
 
  // Update the model's name property 
  model.name = "New name!";
 
  // Once we're done updating properties we can call save on the model. 
  // Save will send the current model data to the DB you specified 
  model.save(function (errupdatedModel) {
    if (err) throw new Error('Could not save the model');
    console.log("The model was updated!");
  });
});
 

Model uses a pretty simple syntax for defining a model. (It should look familiar to anyone who has used an ORM like ActiveRecord, DataMapper, Django's models, or SQLAlchemy.)

var User = function () {
  this.property('login', 'string', {required: true});
  this.property('password', 'string', {required: true});
  this.property('lastName', 'string');
  this.property('firstName', 'string');
 
  this.validatesPresent('login');
  this.validatesFormat('login', /[a-z]+/, {message: 'Subdivisions!'});
  this.validatesLength('login', {min: 3});
  this.validatesConfirmed('password', 'confirmPassword');
  this.validatesWithFunction('password', function (s) {
      if (password.indexOf('login')) {
          // Returning a string indicates that the validation has failed. 
          // The string will be used as the message. 
          return "Password must not contain your username";
      }
      return true;
  });
 
  this.someMethod = function () {
    // Do some stuff 
  };
};
 
// prepares the model and adds it as a property on `model` 
User = model.register('User', User);

Although you can set a default adapter, you may want to override that default on a per model basis. To do that simply call this.setAdapter(name, config) just like you would with createAdapter, like so for a MongoDB adapter:

var model = require('model');
 
var Foo = function () {
  this.setAdapter('mongo', {
    "hostname":"localhost",
    "port":27017,
    "username":"",
    "password":"",
    "dbName":"mydatabase"
  });
 
  this.defineProperties({
    name: { type: 'string', required: true }
  });
};
 
Foo = model.register('Foo', Foo);
 
module.exports = Foo;

Properties can be defined using the property method, which takes a name, a type, and some options.

Alternatively, you can use the defineProperties method to lay out your model's properties with a single method call:

var User = function () {
  this.defineProperties({
    login: {type: 'string', required: true}
  , password: {type: 'string', required: true}
  , lastName: {type: 'string'}
  , firstName: {type: 'string'}
  });
}

Model supports the following datatypes:

  • string
  • text
  • number
  • int
  • boolean
  • date
  • datetime
  • time
  • object

The object data type can take a JSON string or an object that will serialize to JSON.

There is no currency or decimal datatype. For currencies it is recommended to use an int representing the smallest domination (such as cents), like the Stripe API does.

Custom instance methods can be attached to the prototype of a model. Be careful not to use a name that conflicts with a property, and avoid overriding a method supplied by model like save (instead use a [lifecycle event).

Class methods can be attached to the constructor of a model. As with instance methods, take care not to override existing class methods, such as all, first, and count.

Here is an example of declaring class and instance methods:

var User = function () {
    this.property('email', 'string');
    this.property('sendUpdates', 'boolean');
 
    // setting a property on `this` in the constructor creates an instance method 
    this.sendConfirmation = function (callback) {
        // send confirmation message 
    }
}
 
// setting a property on the prototype creates an instance method 
User.prototype.sendNewsletter = function (callback) {
    // send newsletter 
}
 
// setting a property on the constructor creates a class method 
User.sendNewsletters = function (callback) {
    // send newsletter to users that opted into updates 
}
 
User = model.register('User', User);

An adapter allows a model to communicate with the database in a common way across any supported databases. For example, this allows you to easily switch from an in memory database for testing to something like MongoDB just by changing a single line in your model config.

Some adapters require you to install a 3rd party module from NPM. Below is a list of the adapters that require an install and how to install it. The --save flag is optional on all of these but will put it in your package.json file for you:

  • Postgres: npm install pg --save
  • MySQL: npm install mysql --save
  • SQLite: npm install sqlite3 --save
  • MongoDB: npm install mongodb --save
  • LevelDB: npm install level --save

The in-memory, filesystem, and Riak adapters work out of the box and don't need any additional libraries.

Use model.createAdapter(name, config) to initialize an adapter and connect to the database.

NOTE: The config parameter for each adapter depends on the module used. As an example, postgres uses database for the database name whereas MongoDB uses dbName. Model doesn't try to standardize the config for each adapter. Instead it just passes the config you give it in createAdapter to the NPM module. To check the config setup for your adapter go to the module's official site and look at their docs.

var adapter = model.createAdapter('postgres', {
, host: 'localhost'
, username: 'user'
, password: 'password'
, database: 'mydb'
});

Use the defaultAdapter property on model.config to set a default adapter for all models that don't manually specify .setAdapter in the model definition.

model.config.defaultAdapter = model.createAdapter('postgres', {
  host: 'localhost',
  username: 'user',
  password: 'password',
  database: 'mydb'
});

If you want to hook into when the adapter connects to the database you can hook into the connect method which will fire when the connection is either successful or it fails.

myAdapter.connect(function (err) {
  if (err) throw new Error('Error: ' + err);
  console.log('Database connection successful');
}

Same as adapter.connect, but when a disconnect happens.

myAdapter.disconnect(function (err) {
  if (err) throw new Error('Error: ' + err);
  console.log('Database disconnected successfully');
}

Same as adapter.connect, but in event form and only when it is successful.

myAdapter.addListener('connect', function () {
  console.log('Database connection successful');
});

Same as adapter.disconnect, but in event form and only when it is successful.

myAdapter.addListener('disconnect', function () {
  console.log('Database disconnected successfully');
});

Fires whenever there is any error in the database connection during a connect or disconnect attempt.

myAdapter.addListener('error', function (err) {
  throw new Error('Error: ' + err);
});

Creating an instance of one of these models is easy:

var params = {
  login: 'alex'
, password: 'lerxst'
, lastName: 'Lifeson'
, firstName: 'Alex'
};
var user = User.create(params);

Validations provide a nice API for making sure your data items are in a good state, and creates friendly error messages to provide feedback to the user. When an item is "valid" it means that its data meet all the criteria you've set for it. You can specify that certain fields have to be present, have to be certain length, or meet any other specific criteria you want to set.

Here's a list of supported validation methods:

  • validatesPresent – ensures the property exists
  • validatesLength – ensures the minimum, maximum, or exact length
  • validatesFormat – validates using a passed-in regex
  • validatesConfirmed – validates a match against another named parameter (useful for passwords)
  • validatesAbsent – ensures the property does not exist
  • validatesWithFunction – uses an arbitrary function to validate

Here are some simple examples of each validation method:

var User = function () {
  this.property('login', 'string', {required: true});
  this.property('password', 'string', {required: true});
  this.property('lastName', 'string');
  this.property('firstName', 'string');
 
  this.validatesPresent('login');
  this.validatesFormat('login', /[a-z]+/, {message: 'Subdivisions!'});
  this.validatesLength('login', {min: 3});
  this.validatesConfirmed('password', 'confirmPassword');
  this.validatesWithFunction('password', function (s) {
      // Something that returns true or false 
      return s.length > 0;
  });
  this.validatesAbsent('unconfirmed');
  this.validatesWithFunction('password', function (valuemodel) {
      // if the function returns false it will use a standard message 
      if (typeof value != typeof "") {
         return false;
      }
 
      // if it returns a string the string will be used as an error message 
      if  (value.length <= 3) {
         return "Your password must be at least 4 characters long ";
      }
 
      // return true if the validation passed 
      return true;
  });
 
  // methods for instances can be defined in the constructor like this 
  this.someMethod = function () {
    // Do some stuff 
  };
};

You can specify a custom error message for when a validation fails using the 'message' option:

var Zerb = function () {
  this.property('name', 'string');
  this.validatesLength('name', {is: 3, message: 'Try again, gotta be 3!'});
};

You can decide when you want validations to run by passing the 'on' option.

var User = function () {
  this.property('name', 'string', {required: false});
  this.property('password', 'string', {required: false});
 
  this.validatesLength('name', {min: 3, on: ['create', 'update']});
  this.validatesPresent('password', {on: 'create'});
  this.validatesConfirmed('password', 'confirmPassword', {on: 'create'});
};
 
// Name validation will pass, but password will fail 
myUser = User.create({name: 'aaa'});
 

The default behavior is for validation on both 'create' and 'update':

  • create – validates on MyModelDefinition.create
  • update – validates on myModelInstance.updateProperties

You can also define custom validation scenarios other than create and update. (There is a builtin custom 'reify' scenario which it uses when instantiating items out of your datastore. This happens when the first and all query methods are called.)

// Force validation with the `reify` scenario, ignore the too-short name property 
myUser = User.create({name: 'aa'}, {scenario: 'reify'});
 
// You can also specify a scenario with these methods: 
// Enforce 'create' validations on a fetch -- may result in invalid instances 
User.first(query, {scenario: 'create'}, cb);
// Do some special validations you need for credit-card payment 
User.updateProperties(newAttrs, {scenario: 'creditCardPayment'});

Any validation errors show up inside an errors property on the instance, keyed by field name. Instances have an isValid method that returns a Boolean indicating whether the instance is valid.

// Leaving out the required password field 
var params = {
  login: 'alex'
};
var user = User.create(params);
 
// Prints 'false' 
console.log(user.isValid());
// Prints 'Field "password" is required' 
console.log(user.errors.password);

After creating the instance, call the save method on the instance. This method takes a callback in the familiar (err, data) format for Node.

if (user.isValid()) {
  user.save(function (errdata) {
    if (err) {
      throw err;
    }
    console.log('New item saved!');
  });
}

Use the updateProperties method to update the values of the properties on an instance with the appropriate validations. Then call save on the instance.

user.updateProperties({
  login: 'alerxst'
});
if (user.isValid()) {
  user.save(function (errdata) {
    if (err) {
      throw err;
    }
    console.log('Item updated!');
  });
}

Use the remove method to remove one ore multiple instances.

model.User.remove(user.id, function(err) {
  if (err) {
    throw err;
  }
  console.log('Item removed!');
});

You can also pass a query to the remove method instead of an id.

Both the base model constructors and model instances are EventEmitters. They emit events during the create/update/remove lifecycle of model instances. In all cases, the plain-named event is fired after the event in question, the 'before'-prefixed event, of course happens before.

The constructor for a model emits the following events:

  • beforeCreate
  • create
  • beforeValidate
  • validate
  • beforeUpdateProperties
  • updateProperties
  • beforeSave (new instances, single and bulk)
  • save (new instances, single and bulk)
  • beforeUpdate (existing single instances, bulk updates)
  • update (existing single instances, bulk updates)
  • beforeRemove
  • remove

Model-item instances emit these events:

  • beforeUpdateProperties
  • updateProperties
  • beforeSave
  • save
  • beforeUpdate
  • update

Model-item instances also have the following lifecycle methods:

  • afterCreate
  • beforeValidate
  • afterValidate
  • beforeUpdateProperties
  • afterUpdateProperties
  • beforeSave
  • afterSave
  • beforeUpdate
  • afterUpdate

If these methods are defined, they will be called at the appropriate time. You can hook into this lifecycle to do things like set default values for your items:

var User = function () {
  this.property('name', 'string', {required: false});
 
  // Make sure there's a name in the params before validating 
  this.beforeValidate = function (params) {
    params.name = params.name || 'Zerp Derp';
  };
};

Model has support for paired associations including hasMany/belongsTo and hasOne/belongsTo. For example, if you had a User model with a single Profile, and potentially many Accounts:

var User = function () {
  this.property('login', 'string', {required: true});
  this.property('password', 'string', {required: true});
  this.property('confirmPassword', 'string', {required: true});
 
  this.hasOne('Profile');
  this.hasMany('Accounts');
};
 
var Profile = function() {
  // properties 
 
  this.belongsTo('Account');
};
 
var Account = function() {
  // properties 
 
  this.belongsTo('User');
};
 
// Names in association methods must match the names that models are registered with. 
// The name given to hasMany is pluralized. 
User = model.register('User', User);
Profile = model.register('Profile', Profile);
Account = model.register('Account', Account);

Add the hasOne relationship by calling 'set' plus the name of the owned model in singular (in this case setProfile). Retrieve the associated item by using 'get' plus the name of the owned model in singular (in this case getProfile). Here's an example:

var user = User.create({
  login: 'asdf'
, password: 'zerb'
, confirmPassword: 'zerb'
});
user.save(function (errdata) {
  var profile;
  if (err) {
    throw err;
  }
  profile = Profile.create({});
  user.setProfile(profile);
  user.save(function (errdata) {
    if (err) {
      throw err;
    }
    user.getProfile(function (errdata) {
      if (err) {
        throw err;
      }
      console.log(profile.id ' is the same as ' + data.id);
    });
  });
});

Set up the hasMany relationship by calling 'add' plus the name of the owned model in singular (in this case addAccount). Retrieve the associated items with a call to 'get' plus the name of the owned model in plural (in this case getAccounts). An example:

var user = User.create({
  login: 'asdf'
, password: 'zerb'
, confirmPassword: 'zerb'
});
user.save(function (errdata) {
  if (err) {
    throw err;
  }
  user.addAccount(Account.create({}));
  user.addAccount(Account.create({}));
  user.save(function (errdata) {
    if (err) {
      throw err;
    }
    user.getAccounts(function (errdata) {
      if (err) {
        throw err;
      }
      console.log('This number should be 2: ' + data.length);
    });
  });
});

A belongsTo relationship is created similarly to a hasOne: by calling 'set' plus the name of the owner model in singular (in this case setAuthor). Retrieve the associated item by using 'get' plus the name of the owner model in singular (in this case getAuthor). Here's an example:

var book = Book.create({
  title: 'How to Eat an Entire Ham'
, description: 'Such a poignant book. I cried.'
});
book.save(function (errdata) {
  if (err) {
    throw err;
  }
  book.setAuthor(Author.create({
    familyName: 'Neeble'
  , givenName: 'Leonard'
  }));
  book.save(function (errdata) {
    if (err) {
      throw err;
    }
    book.getAuthor(function (errdata) {
      if (err) {
        throw err;
      }
      console.log('This name should be "Neeble": ' + data.familyName);
    });
  });
});

A similar API is used for removing associations, with the word 'remove' plus the name of the owned model:

User.first({login: 'asdf'}, function (erruser) {
  if (err) {
    throw err;
  }
  // Fetch accounts 
  user.getAccounts(function (erraccounts) {
    var originalCount;
    if (err) {
      throw err;
    }
    originalCount = accounts.length;
 
    // Remove the first account 
    user.removeAccount(accounts[0]);
    // Save the user 
    user.save(function (err) {
      if (err) {
        throw err;
      }
      // Fetch accounts again 
      user.getAccounts(function (erraccounts) {
        if (err) {
          throw err;
        }
        Console.log(accounts.length + ' should be one less than ' +
            originalCount);
      });
    });
  });
});

Note that this does not remove the associated item itself -- only the association linking it to the owner object.

'Through' associations allow a model to be associated with another through a third model. A good example would be a Team linked to Players through Memberships.

var Player = function () {
  this.property('familyName', 'string', {required: true});
  this.property('givenName', 'string', {required: true});
  this.property('jerseyNumber', 'string', {required: true});
 
  this.hasMany('Memberships');
  this.hasMany('Teams', {through: 'Memberships'});
};
 
var Team = function () {
  this.property('name', 'string', {required: true});
 
  this.hasMany('Memberships');
  this.hasMany('Players', {through: 'Memberships'});
};
 
var Membership = function () {
  this.belongsTo('User');
  this.belongsTo('Team');
};

The API for this is the same as with normal associations, using the set/add and get, with the appropriate association name (not the model name). For example, in the case of the Team adding Players, you'd use addPlayer and getPlayer.

Sometimes you need mutliple associations to the same type of model (e.g., I have lots of Friends and Relatives who are all Users). You can accomplish this in Model using named associations:

var User = function () {
  this.property('familyName', 'string', {required: true});
  this.property('givenName', 'string', {required: true});
 
  this.hasMany('Kids', {model: 'Users'});
};

The API for this is the same as with normal associations, using the set/add and get, with the appropriate association name (not the model name). For example, in the case of Kids, you'd use addKid and getKids.

If one of your named associations of a model is 'through' another model, such as a join table, it is necessary that the association's name is the same for the model declaring the through association as it is for the model who the association is through.

For example, a team may have many players, but may also have many coaches.

var Team = function(){
  this.hasMany('Players');
  this.hasMany('Coaches', {through: 'TeamCoaches', model: 'Players'});
};
var TeamCoaches = function(){
  this.belongsTo('CoachedTeam', {model: 'Team'});
  this.belongsTo('Coach', {model: 'Player'});
}
var Player = function(){
  this.hasMany('Teams');
  this.hasMany('CoachedTeams', {through: 'TeamCoaches', model: 'Team'});
}

Here a Team has many Players, but also has many Coaches, and we have an inverse relationship set up as well so that a Player has many Teams but also has many CoachedTeams.

Model uses a simple API for finding and sorting items. Again, it should look familiar to anyone who has used a similar ORM for looking up records. The only wrinkle with Model is that the API is (as you might expect for a NodeJS library) asynchronous.

Methods for querying are static methods on each model constructor.

Use the first method to find a single item. You can pass it an id, or a set of query parameters in the form of an object-literal. In the case of a query, it will return the first item that matches, according to whatever sort you've specified.

var user;
User.first({login: 'alerxst'}, function (errdata) {
  if (err) {
    throw err;
  }
  user = data;
  console.log('Found user');
  console.dir(user);
});

Use the all method to find lots of items. Pass it a set of query parameters in the form of an object-literal, where each key is a field to compare, and the value is either a simple value for comparison (equal to), or another object-literal where the key is the comparison-operator, and the value is the value to use for the comparison.

In SQL adapters, you can pass a callback to the all method if you want the results buffered and returned all at once, or steam the results using events.

Pass your callback function as a final argument. Callbacks use the normal (err, data) pattern. Here's an example:

var users
  , dt;
 
dt = new Date();
dt.setHours(dt.getHours() - 24);
 
// Find all the users created since yesterday 
User.all({createdAt: {gt: dt}, function (errdata) {
  if (err) {
    throw err;
  }
  users = data;
  console.log('Found users');
  console.dir(users);
});

The all method returns an EventedQueryProcessor which emits the normal 'data', 'end', and 'error' events. Each 'data' event will return a single model-item.

NOTE: Do not pass a callback to the all method if you're streaming -- passing a callback will cause the results to be buffered internally. If you need something to happen when the stream ends, use the 'end' event.

var users
  , dt
  , processor;
 
dt = new Date();
dt.setHours(dt.getHours() - 24);
 
// Find all the users created since yesterday 
processor = User.all({createdAt: {gt: dt});
processor.on('data', function (user) {
  console.log('Found user');
  console.dir(user);
});
processor.on('error', function (err) {
  console.log('whoops');
  throw err;
});
processor.on('end', function () {
  console.log('No more users');
});

Here are a few more examples of queries you can pass to the all method:

// Where "foo" is 'BAR' and "bar" is not null 
{foo: 'BAR', bar: {ne: null}}
// Where "foo" begins with 'B' 
{foo: {'like': 'B'}}
// Where foo is less than 2112, and bar is 'BAZ' 
{foo: {lt: 2112}, bar: 'BAZ'}

Here is the list of comparison operators currently supported:

  • eql: equal to
  • ne: not equal to
  • gt: greater than
  • lt: less than
  • gte: greater than or equal
  • lte: less than or equal
  • like: like

A simple string-value for a query parameter is the same as 'eql'. {foo: 'bar'} is the same as {foo: {eql: 'bar'}}.

For case-insensitive comparisons, use the 'nocase' option. Set it to true to affect all 'like' or equality comparisons, or use an array of specific keys you want to affect.

// Zoobies whose "foo" begin with 'b', with no case-sensitivity 
Zooby.all({foo: {'like': 'b'}}, {nocase: true}, ...
// Zoobies whose "foo" begin with 'b' and "bar" is 'baz' 
// The "bar" comparison will be case-sensitive, and the "foo" will not 
Zooby.all({or: [{foo: {'like': 'b'}}, {bar: 'baz'}]}, {nocase: ['foo']},

Model supports combining queries with OR and negating queries with NOT.

To perform an 'or' query, use an object-literal with a key of 'or', and an array of query-objects to represent each set of alternative conditions:

// Where "foo" is 'BAR' OR "bar" is 'BAZ' 
{or: [{foo: 'BAR'}, {bar: 'BAZ'}]}
// Where "foo" is not 'BAR' OR "bar" is null OR "baz" is less than 2112 
{or: [{foo {ne: 'BAR'}}, {bar: null}, {baz: {lt: 2112}}]}

To negate a query with 'not', simply use a query-object where 'not' is the key, and the value is the set of conditions to negate:

// Where NOT ("foo" is 'BAR' and "bar" is 'BAZ') 
{not: {foo: 'BAR', bar: 'BAZ'}}
// Where NOT ("foo" is 'BAZ' and "bar" is less than 1001) 
{not: {foo: 'BAZ', bar: {lt: 1001}}}

These OR and NOT queries can be nested and combined:

// Where ("foo" is like 'b' OR "foo" is 'foo') and NOT "foo" is 'baz' 
{or: [{foo: {'like': 'b'}}, {foo: 'foo'}], not: {foo: 'baz'}}

The all API-call for querying accepts an optional options-object after the query-conditions for doing sorting, skipping to particular records (i.e., SQL OFFSET), and limiting the number of results returned.

Set a 'sort' in that options-object to specifiy properties to sort on, and the sort-direction for each one:

var users
// Find all the users who have ever been updated, and sort by 
// creation-date, ascending, then last name, descending 
User.all({updatedAt: {ne: null}}, {sort: {createdAt: 'asc', lastName: 'desc'}},
    function (errdata) {
  if (err) {
    throw err;
  }
  users = data;
  console.log('Updated users');
  console.dir(users);
});

You can use a simplified syntax for specifying the sort. The default sort-direction is ascending ('asc'), so you can specify a property to sort on (or multiple properties as an array) if you want all sorts to be ascending:

// Sort by createdAt, ascending 
{sort: 'createdAt'}
// Sort by createdAt, then updatedAt, then lastName, 
// then firstName -- all ascending 
{sort: ['createdAt', 'updatedAt', 'lastName', 'firstName']}

The 'skip' option allows you to return records beginning at a certain item number. Using 'limit' will return you only the desired number of items in your response. Using these options together allow you to implement pagination.

Remember that both these option assume you have your items sorted in the desired order. If you don't sort your items before using these options, you'll end up with a random subset instead of the items you want.

// Returns items 501-600 
{skip: 500, limit: 100}

You can use the 'includes' option to specify second-order associations that should be eager-loaded in a particular query (avoiding the so-called N + 1 Query Problem). This will also work for 'through' associations.

For example, with a Team that hasMany Players through Memberships, you might want to display the roster of player for every team when you display teams in a list. You could do it like so:

var opts = {
  includes: ['players']
, sort: {
    name: 'desc'
  , 'players.familyName': 'desc'
  , 'players.givenName': 'desc'
  }
};
Team.all({}, opts, function (errdata) {
  var teams;
  if (err) {
    throw err;
  }
  teams = data;
  teams.forEach(function (team) {
    console.log(team.name);
    team.players.forEach(function (player) {
      console.log(player.familyName + '' + player.givenName);
    });
  });
});

You can also do an eager load of nested associations. If you wanted to get the sponsors of each player, you can do it like so:

Team.all({}, {includes: {players: 'sponsors'}}, function (errdata) {});

You can also get the investors of the teams like so:

Team.all({}, {includes: [{players: 'sponsors'}, 'investors']}, function (errdata) {});

Or get the investors' spouses as well:

Team.all({}, {includes: {players: 'sponsors', investors: 'spouse'}, function (errdata) {});

While there is no hard limit on nesting associations, queries like this search for friends of friends of friends are likely to have poor performance:

Person.all({}, {includes: {friends: {friends: 'friends'}}, function (errdata) {});

You can also query on nested associations. This query will return teams with players sponsored by Daffy Duck:

Team.all({'players.sponsors.name': 'Daffy Duck'}, {includes: {players: 'sponsors'}}, function (errdata) {});

Notice that it's possible to sort the eager-loaded associations in the above queries. Just pass the association-names + properties in the 'sort' property.

In the first example, the 'name' property of the sort refers to the team-names. The other two, 'players.familyName' and 'players.givenName', refer to the loaded associations. This will result in a list where the teams are initially sorted by name, and the contents of their 'players' list have the players sorted by given name, then first name.

You can sort on nested attributes by specifying the association name:

{sort: 'players.sponsors.id'}

Due to limitations in SQL, please take note of the following when using eager loading:

  • Querying on associations is only possible when including the associated model

If you query on an association, you must include the relationship, or the query will fail.

// Good 
Team.all({'players.sponsors.name': 'Daffy Duck'}
        , {includes: {players: 'sponsors'}}
        , function (errdata) {});
 
// Bad 
Team.all({'players.sponsors.name': 'Daffy Duck'}
        , {includes: 'players'}
        , function (errdata) {});
  • Querying on associations is not possible when there is a limit clause

This is a limitation of the current implementation. An exception will be thrown when queries like this are attempted.

// Bad 
Team.all({'players.sponsors.name': 'Daffy Duck'}
        , {includes: {players: 'sponsors'}, limit: 5}
        , function (errdata) {});
 
// Bad too, since .first is an implicit "limit: 1" 
Team.first({'players.sponsors.name': 'Daffy Duck'}
        , {includes: {players: 'sponsors'}, limit: 5}
        , function (errdata) {});
  • Streaming is not possible when sorting on a nested association before the top level id
// Streaming API will work, the sort clause will be modified to ['id', 'players.name'] 
Team.all({'players.sponsors.name': 'Daffy Duck'}
        , {includes: {players: 'sponsors'}, sort: ['players.name']});
 
// Streaming API will still work, but results will only be sent at the end of the query 
Team.all({'players.sponsors.name': 'Daffy Duck'}
        , {includes: {players: 'sponsors'}, sort: ['players.name', 'id']});

The eagerly fetched association will be in a property on the top-level item with the same name as the association (e.g., Players will be in players).

If you have an item, and you're not certain whether an association is already loaded, you can check for the existence of this property before doing a per-item fetch:

if (!someTeam.players) {
  someTeam.getPlayers(function (errdata) {
    console.dir(data);
  });
}

Run the tests with jake test. Run only unit tests with jake test[unit].

The integration tests require the appropriate database and supporting library. (For example, running the Postgres tests require a running Postgres server, and the 'pg' module NPM-installed in your model project directory.) To install the needed modules, just run npm install in the root model directory.

To run the tests on a specific adapter, use jake test[mongo], jake test[postgres], or jake test[memory].

Configure adapter options by creating a test/db.json file. See test/db.sample.json for available options.


Model JavaScript ORM copyright 2112 mde@fleegix.org.