Skip to content

DeriveODM is a reactive ODM - Object Document Mapper - framework, a "wrapper" around MongoDB, that removes all the hassle of data-persistence by handling it transparently in the background, in a DRY manner.

License

Notifications You must be signed in to change notification settings

yuval-a/deriveODM

Repository files navigation

Summary | Reference | Comparison to Mongoose | Demonstration

Introduction

DeriveODM (formerly DeriveJS) lets you manipulate and create Javascript data objects, while automatically and transparently persisting and updating them on a database (such as MongoDB), in the background, without any additional hassle or code.

It wraps your data classes and objects with Javascript Proxies, "tapping-in" to native operations such as creating instances (using the normal new operator), and updating property values (using the normal assignment operator =), and then handling passing database calls and operations to the database in the background, while leveraging MongoDB's bulk operations capabilities, orchestrating them in a smart, efficient way, to save unnecessary calls to the DB engine, and running bulk operations in fixed (settable) intervals. The background engine is mostly handled transparently by a module called SyncManager.

Note: this is a complete technical reference, if you'd like to read a less verbose introduction, you can read this article on Hashnode

Table of Contents

The Rationale Behind DeriveJS, Using an Analogy

If you are familiar with a front-end UI framework such as ReactJS, you know that whenever a change is made to the state object - React will automatically know to issue a re-render of the component - this is known as a "pull" methodology, where as in other similar frameworks, you might need to explicitly call a render() method (this is a "push" methodology, in that context). In a similar analogy to the way React works - when using Derive - you are not required to call an explicit save() method to have your data be saved and persisted on a DB - it's enough that you make some change to an exisiting data object, or create a new one - and Derive will already know to handle that data's persistence.

To sum-up: DeriveJS is a reactive ODM (Object Document Mapper), that lets you deal with data, in a DRY way, without having to take care of all the hassle of the logistics of database persistency.

Comparison to Mongoose

If you used or are using Mongoose and considering moving to Derive, or would like to see a comparison between the two, you can go over this document.

Reference

For a complete reference of all available methods, functions and objects available in Derive - see this document.

How to use

It only takes a few easy steps:

Define a data model
const User = Model({
    _email$: "",
    _name: "",
    role:"editor", // set a default role for a User object
    age: null,
    password: null,
    setPassword(pass) {
        // hash the plain-text password ("hashit" is just an example function for your preffered hashing function)
        var passwordHash = hashit(pass);
        this.password = passwordHash;
    }
},"User");

The first time you define it, a Users collection is defined on the database, with an _email unique index and a _name index (you can also alter the properties, change indexes later).

Create an instance
let user = new User ("[email protected]","Someone Somebody");

There will now be a new document in the Users collection, having "[email protected]" as the _email and "Someone Somebody" as the _name.

Manipulate properties
user.age = 30;

The document will now have the value 30 set to its age property.

Call instance functions
user.setPassword("plaintextpassword");
Derive a data model

One of the powerful features of Derive, is the ability to "extend" data models, while having the "sub-models" share the same collection as their "super-models" - this, together with the "Default Criteria" feature can enable meaningful "data inheritance".

const Admin = 
   User.derive({
   	role_: "admin"
   });

let admin = new Admin("[email protected]", "Admin Name");

With DeriveJS you can create and manipulate a large amount of data objects, and know that they will be persisted in the database, efficiently and in a short time.

Although the methodology behind the framework is mostly that of "send and forget" regarding data persistence - DeriveJS also exposes callback functions that allows getting notified exactly when specific objects are actually saved on the database, or exactly when specific properties have been actually updated, for the occasions when you need to know it for certain operations.

Getting Started

Install via npm:
npm install derivejs
or clone the git repository.

You should also setup or have access to a MongoDB server, and have it running.

To get started, require DeriveJS, and then call the Model module. The module is a Promise that resolves with a Model function. By the time the Promise is resolved, the module has finished its initializations, and is connected to the MongoDB server. When you call the module, you can pass an options object argument to it. The options can contain these key:value arguments:

  • dbUrl: the MongoDB server connection url, as a string. Default: "mongodb://localhost:27017/".
  • dbName: the name of the database, as a string. Default: "deriveDB".
  • debugMode: A boolean. If set to true - will display some real-time internal DB-related SyncManager information - such as when it is locked for operation (before running bulk database operations), or unlocked, when running bulk inserts/updates, and more... Default: true (!)
  • defaultMethodsLog: when set to true - the default class methods for database/data events (_inserted(), _isDuplicate(), _error() and changed()) will run a relevant console.log, see Database persistence callbacks for more information about these methods.
  • dbOptions: You can use this to override the default MongoDB driver connection options. Note that these are the options passed by default:
native_parser:true, 
forceServerObjectId:true,
// New in 3.X Mongo engine
useUnifiedTopology: true,
ignoreUndefined: true

Here is an example of how to initialize the module:

const derive = require('derivejs');
derive.Model
({
    dbUrl: "mongodb://localhost:27017/",
    dbName: "mydatabase",
})
.then(
    Model=> {
		// `Model` is a function here.
    }
);

NEW: starting with version 1.6.0 you may use the Connect module function (derive.Connect) instead of Model (derive.Model), which is exactly the same as Model, and is used as an alias (both reference the same module) - and was added to avoid confusion between the resolved Model function (the function you use to define data models) and the module Model function (the function which is used to connect to the database and resolve with the Model function), so the second line above could be:

derive.Connect

as well.

Defining a Data Model

Once the promise resolves successfully, you have access to the Model function, which can be used to define a "Data Model", by passing an object literal as an argument, describing the data properties and their default values (as well as some instance methods, when needed). The Model function returns a class "representing" that data model, and its functionality (as mentioned, that class is a special "proxied" class that comes with some built-in stuff in it, to handle database persistence, and offer some static and default instance methods which will be described below).

Let's create a data model to represent a "spaceship":

var Spaceship = Model({
    _name: "",
    _TYPE: "",
    crew: []
}, "Spaceship");

The property names specify the expected property key names for this data model. The values are default values - values that will be assigned to the properties upon initial creation of the data object (and data document), you can put any values you like, including null, empty strings, strings, numbers, objects (with keys and values) and so on. Unlike many other ODMs - Derive does not enforce any strict data types, preserving the spirit of dynamic data typing.

A note about objects as values: If specifying an empty object ({}) as a value, you will not be able to set any new properties to it (as they will not be part of the "schema"), however, it is perfectly acceptable to assign a totally new object with properties and values as a value to any defined property. If you assign an object, with properties and values as a default value to a property in the schema specification - you may change the values of those properties, but trying to set a new property to that object will yield an error ("Trying to set unknown property).

Now Spaceship is a class you can create new object instances of. The second argument for the Model function is a name, that will be used for both the class name, and the collection name (where an s will be added to, to signify "plural" form; So the collection name will be Spaceships in this case). There is an additional optional argument, that can set the "sync interval" duration, the amount of time between each interval where the SyncManager class instance runs the database bulk operations stacked since the last sync. The default is 0, and the call is made using setImmediate to avoid "clogging" Node's Event Loop (Note: prior to version 2.0.0, the default was 1000 ms). (The Model function also have two additional arguments _syncer, and _proxy, used internally and that shouldn't be used).

Notice that some properties were defined with an underscore, and some are uppercase. This is intentional, and meaningful. These are called "Modifiers" and are explained in the next section.

Going deeper

Modifiers

You can use some special characters in the property names (called "Modifiers"), that will define certain characteristics of those properties:

  • _ (start) - Index: When used as the first character of a property, will mark it as an index. Other than being defined as an index inside MongoDB, an index is also always setabble from the model class constructor. Notes: The order of the indexes defined, matters - as this will also be the order of the arguments in the constructor. Furthermore - the order affects which index is considered as the "Main Index". Also note, that the property name does include the underscore character. Example: _name;

  • $ (last) - Unique Index: When used as the last character of an index property, that index will be set as a unique index (using the same value for unique indexes will yield an error). Example: _name$, will define a unique index called _name (notice - the $ char at the end of the property name will be removed, and will not be defined as part of the name).

  • ALL_UPPERCASE - Read Only: When a property name is defined with all capital letters, it will be marked as read-only. If you try to set the value of a read-only property (using the = operator) - you will get an error message. Note, that if you also define a read-only property as an index, like in the above example - that property will still be settable via the constructor arguments (but not from anywhere else).

  • $ (start) - Meta Property Putting the Dollar sign as the first character of a property name - will define it as a "meta" property (aka a "secret" property). A meta property will not be considered as part of the data structure of the model - it and its value will not be persisted on the database. If you iterate over the values of the data instance - it will not appear (it won't be enumerable). But you may still get and set its value locally. This is useful for saving some additional information that you only need locally, and does not require persistence on the database server. These can also be used to reference callback functions, as demonstrated later-on.
    There are also, in-fact, five "built-in" meta properties, four are automatically created for each object: $_ModelInstance which is always equal to true, and is used internally when setting a DeriveJS object value to an instance of another DeriveJS object (in which case it will be saved as a DBRef object), $DefaultCriteria (which is explained later), $_BARE which contains the "raw" (unproxified) document. Setting the values of the $_BARE object - will not affect the database, $_dbEvents - an EventEmitter for listening and responding to DB persistence events. The fifth property is $Listen, which is not created automatically but can be defined as an array of property names that you want their value-changes to be "listened" to (as explained in "Listening for changes").

  • _ (end) - Default Criteria: Using a _ character as the last character in the end of a property name, will add the property and its value to the $DefaultCriteria object, (note, the last _ will be omitted from the property name). If using both last _ and last $ (i.e. setting both a unique index and a default-criteria value, make sure the _ is the last character, and the $ is one before it).

Having obtained our data class, we can create object instances of it:

var ship = new Spaceship('The Beyond'); 

Each new instance - will have an identical data record (a "document") in a MongoDB database, in a Spaceships collection, which will always be synced with the changes you make to the "local" object. Once you create a new instance, that instance will also have an auto-generated _id value (of type ObjectID) associated with it.
(Warning: You should not alter the _id property yourself, nor should you try to give it your own value, doing so may lead to data inconsistency, and may affect data integrity).

After a while you should see a message on the console: The Beyond inserted, that message comes from an instance method that all model instances have by default, and can be overridden by subclasses, or by redefining in the model definition:

_inserted() {
    console.log (this[MainIndex]+" inserted");
}

(MainIndex is an internal variable that holds the "primary" index of the collection, in this case _name). The method is called as soon as the data object is persisted on the MongoDB server.

Notice how we passed a _name for the new instance via the first argument of the constructor. We can do that, since we defined it as an index. We can also pass a value for the _TYPE index as the second argument. Let's create a different ship, and define it as "TYPE A":

var shipA = new Spaceship('The Beyonder','A');

If we now run the Mongo console, and run a "find all" query on the Spaceships collection (db.Spaceships.find({}), we will see our two Spaceship data objects saved on the server:

{ "_id" : ObjectId("5a063879cc5cac16b82d05f0"), "_name" : "The Beyond", "_TYPE" : "", "crew" : [ ] }
{ "_id" : ObjectId("5a063879cc5cac16b82d05f1"), "_name" : "The Beyonder", "_TYPE" : "A", "crew" : [ ] }

Your _id values will vary, of-course.

Built-in Methods: Callbacks and Hooks

NEW VERSION UPDATE:

  • Version 3 made the use of Change Streams optional, by adding an additional last boolean argument to all data retrieval functions, set to false by default. Using ChangeStreams and assigning event listeners to data object instances creates strong references to the instances in memory, thereby preventing them from being garbage collected, this can amount to excess memory usage and can lead to memory leaks - therfore this feature is turned off by default, and can be optionally turned on, for specific data objects. Consider this when using this feature, and use it sparingly. Prior (2+) versions always used ChangeStreams by default for all data objects.
  • Added the instance watch(on) method, that lets you turn on or off ChangeStream support.
  • Turning on the use of Change Streams on data objects means that when you work with a local document, even when the data changes in the DB from some other external source - this change will be detected and will be automatically reflected in your version of the document as well.
    NOTE: Change Streams are only supported for Replica Sets. If you use a single DB instance, then they will not be used, and changes will only be triggered by the SyncManager itself - this means that in this case - changes to the DB from another external source - will not be immediately reflected in your local documents (you would need to call the .get() method to retrieve a "fresh copy"). Note that the Mongo team recommends to always use replica sets in production environments.
    Change Streams are also only supported in WiredTiger storage engine (which is the default for Mongo).
  • New in version 2+: all Derive objects now have a $_dbEvents meta property which is an EventEmitter, that you can use to listen for DB persistence events for specific instances. See Database Persistence Events ($_dbEvents) for more information.
  • Version 2+ update: the $_updated method is now deprecated in favour of a new different syntax for defining callbacks for specific DB updates. See "Assignment with$callback" Syntax (Update Callbacks) for more information.

New Model Instance Lifecycle

When creating a new model object instance - first the, "internal" constructor of the model class is called. This constructor is defined within the Model module, and is the same for all model classes (having different initialization properties, of-course). First, if the constructor was called as a result of retrieving a data object (document) from the database - the object values will be populated accordingly. Then, this constructor function is basically in-charge of the following things, in this order:

  1. "Proxifiying" all (non-meta) property values of the object (so they can all be automatically persisted to the DB upon change).
  2. Starting (running) the associated SyncManager class (if it's not already running) - which will "gather" DB-related operations and bulk-run them in fixed intervals for the associated DB collection).
  3. If this is a new data object - create a local data (MongoDB) document first - this is the point where the _created function of the data class is called (if defined) - and is the "last chance" to make new changes before the data object will be persisted on the DB (of-course, you can also update any of its properties as you like later-on, and they'll be persisted as well).
  4. The SyncManager puts an insert operation in the queue of the next bulk DB operations to be run.

It's important to note, that if you extend a Model class, and define both a constructor on the extended class and a _created method - the _created method will be called before the extended class constructor finishes - as it needs to call the base class ("super") constructor first, which will call _created if defined. So, for example with the following:

var Spaceship = Model({
   _name: "",
}, "Spaceship");

class Ship extends Spaceship {
   constructor (_name) {
      super(_name);
      console.log (_name+" called from constructor");
   }
   _created() {
      console.log (this._name+" called from _created");
   }
}

var ship = new Ship("a new ship");

You will first see the console.log message from the _created and then the message from the (extended class) constructor. After a short while, when the document is inserted on the DB - you will see messages logged from the _inserted method (if a console.log was defined there and debugMode was defined with true when initializing the Model module).

Database Persistence Callbacks

Derive has several predefined callbacks defined on the Model class level, are available to all Model instances, and can be overriden in a child class, or in the model definition itself. These method names start with an underscore (_).

_inserted()

If you need to know exactly when an object is actually persisted in the database every instance has a built-in instance method: _inserted() which is called as soon as it's inserted in the DB. You can override that method - either by extending a class - or defining the method directly inside the model definition -- and put some "post-persistence" code there, if you need.

class Ship extends Spaceship {
    _inserted() {
        console.log (this._name+" created, with id: "+this._id); 
    }
}
var ship = new Ship("The Created");

will yield: The Created created, with id: 5a063f842ef67924f4e0f9bb (with a different id of-course).

"Assignment with $callback" Syntax (Update Callbacks)

This is the new way in version 2 and up to assign function callbacks for specific property updates to a data object. With this syntax, instead of directly assigning a value to a property of a data object, you instead assign it an object with two properties:

$value

The actual value you want to assign.

$callback

A function that will be called once the property of the equalivent document in the DB is actually updated.

Example:
Feisty.captain = {
    $value: Wort,
    $callback: ()=> {
        console.log ("Wort was updated as the captain of the Feisty");
    }
}

The value of $value will be assigned to the property, and the function in $callback will be called once that property is updated with that value on the DB.

_isDuplicate()

Called when the MongoDB server yields a "duplicate key value" error, and contains by default:
console.log (this[MainIndex]+" has a duplicate key value!");

_error(msg)

Called whenever there is a data-related error for this object, and contains by default:
console.log ("Error in "+this[MainIndex]+": "+msg);

Database Persistence Events ($_dbEvents)

NOTES:

Each Derive data object has a built-in $_dbEvents meta property, which is an EventEmitter object, that you can use to listen for DB persistence events and changes, by calling the on or once methods to listen for events in specific instances, and attach handler functions (see Node Events documentation for more information about EventEmitter).

These are the events available via $_dbEvents:

inserted Event

Called once a MongoDB document for this instance was inserted to the DB collection. The callback function receives two arguments:

id

The _id of the inserted document.

insertedObject

This is the same relevant Derive data object instance that was created.

Example
(new PhotonTorpedos()).$_dbEvents.once("inserted", (id, torpedos)=> {
    // `torpedos` is the PhotonTorpedos instance.
});
updated Event

Called once a MongoDB document's property is updated on the DB. The callback function receives three arguments:

id

The _id of the updated document.

updatedFields

An object where the keys are updated property names, and the values are the new updated values.

updatedDocument

The Derive object instance.

Example
BoldlyGo.$_dbEvents.on("updated", (id, updatedFields)=> {
    console.log ("BoldlyGo updated properties: ");
    console.dir (updatedFields, {depth:null});
});

The following are additional ways to implement DB persistence callbacks. They were the recommended ways for previous versions of Derive.
For versions 2 and up, the recommended way is to subscribe to DB events, or use "assignment with $callback" syntax.

Setting A Callback Function to a Meta Property

You can define a meta property on the model, to hold a callback function.

var Spaceship = Model({
    _name: "",
    _TYPE: "",
    crew: [],

    $createdCallback: undefined

}, "Spaceship");

Then extend the class, and allow passing a callback function via the constructor. Call the callback from the overridden _inserted function:

class Ship extends Spaceship {
    constructor(_name, _TYPE, callback) {
        super(_name, _TYPE);
        this.$createdCallback = callback;
    }
    _inserted() {
        this.$createdCallback.call(this);
    }
}

var ship = new Ship("shipA","", function() {
    console.log ("shipA created!");
});

Note how when overriding a constructor in a child class - you need to specify the indexes as arguments before adding new ones, and of-course, you need to call the parent constructor via super.

watch(on)

watch is a function all data object instances have, calling it can toggle on/off the ChangeStream support for the data instance. When ChangeStream support is on - the object instance will be aware of updates and changes to the document in the database from external sources, reflecting those in the local data object instance, as well as triggering an 'updated' event on $_dbEvents. Note that turning ChangeStream on, will assign an event listener to the ChangeStream - that will keep a strong reference in memory to the data object instance, thereby making it unavailable for garbage collection - this means that update events will trigger even if the data object instance is set to null. If you use watch to turn on ChangeStream support and then no longer need to data object instance, you should make sure you turn ChangeStream off first, by calling watch with false - this will make the instance garbage collectible once there are no more references to it.

Listening For Local Changes

The fourth built-in method can be used when you want to listen for value-changes on certain properties of your object, (Note: this will trigger on local changes to the properties, regardless to their state in the equavilent documents in the database collection)

changed(property, newValue, oldValue)

The changed method contains this code by default:
console.log (this[MainIndex]+":",property,"changed from",oldValue,"to",newValue);

$Listen

To register a property for the listener, put its name (as a string) inside an array defined as the $Listen meta-property (e.g. $Listen: [ "property" ,"otherproperty", "objectprop.prop"].

Unique Indexes

Now, we decide that we want the _name property index to be unique:

var Spaceship = Model({
    _name$: "",
    _TYPE: "",
    crew: [],
},"Spaceship");

Once the engine modifies the _name index to make it unique, if there are records with duplicate _name values - Mongo will throw an error, and the unique index will not be defined. You will need to take care of the duplicates yourself for it to successfully be defined. You can either issue a relevant .remove command from the mongo console, or you can also use the static clear() method on the Spaceship class. The clear method can accept a "find query" filter as an argument (e.g. {_name:"The Beyond"}). So, for example if you have several ships named "The Beyond" and you want to define a unique constraint on _name and create a new, single "The Beyond" ship, then you can use:

await Model({},"Spaceship").clear({_name:"The Beyond"});
var Spaceship = Model({
    _name$: "",
    _TYPE: "",
    crew: [],
}, "Spaceship");

var ship = new Spaceship ("The Beyond");

The first line is just for getting a reference to the Spaceships collection, to call clear to remove all "The Beyond" ships, then we can define the model with the new constraint, and create a new unique ship.

At this point we have the _name index defined as a unique index -- there can't be more than one object with the same _name value. Let's see what happens when we try to create two Spaceships with the same name:

var ship1 = new Spaceship ("The Boldly Go","A");
var ship2 = new Spaceship ("The Boldly Go","B");

We will get a message on the console: The Boldly Go has a duplicate key value!

One of the records was successfully inserted to the database, the other was detected as having a duplicate _name and was rejected. The message is coming from the _isDuplicate() instance method, that all Model instances have. Its body is defined with:

_isDuplicate() {
    console.log (this[mainIndex]+" has a duplicate key value!");
}

You may override this method in your class extension or model definition, and define your own logic to take care of duplicates. If we look inside the database, we'll see that only the ship with _TYPE A is saved on the database.

It is possible to remove a unique constraint from an index - simply define it without the $ sign - and the engine will know how to redefine it. You may also "disable" an index - redefine it as a "normal" property - to do so, just define it without the leading _. Suppose we want to unindex the _TYPE property, we can define our model as:

var Spaceship = Model({
    _name$: "",
    TYPE: "",
    crew: [],
}, "Spaceship" );

However - any past documents in our database collection, will still have the old _TYPE property in them.

The remodel method

If you'd like to retroactively rename all the old _TYPE properties in existing documents to TYPE, you can use the static remodel() method. That function gets an object of options, as an argument - which you can use to switch on certain types of the method's operations. Currently there are two:

deep: true,
renameIndexes: true

With renameIndexes:true - the method will rename properties in documents in the collection that starts with a _ and are defined in the model without it. Similarly it will rename properties without a leading _ that appear in the model definition with it (for situations where you decide to define an index for a non-index property). So, to rename all _TYPE properties to TYPE in our collection, we could have used: Spaceship.remodel( { renameIndexes: true } );

The deep operation allows you to retroactively define new properties on existing documents on a collection. If you add a new property to your model definition (for example, you add a captain property with a default value of an empty string) - then older documents won't have it (which may or may not be what you want) - using deep:true will add the new property with its default value to all existing documents in the collection.

To read more about indexes and how they are managed in DeriveJS -- see "Indexes and how they are handled in the Mongo server".

You will probably want to have remodel called only once after you make changes to your models, therefore it might be a good idea to call it in response to a certain command-line argument when running your Node app. For example:

// process.argv is an array, where the first item contains 'node', the second item contains the script file name, and the rest of the items are command-line arguments
if (process.argv[2] == "--remodel") {
    Spaceship.remodel({ deep:true, renameIndexes: true});
}

Going Further - Extending and Deriving Models

One of Derive's powerful features is the ability to create "sub-models" associated with the same collection as a "super-model", using the derive method, which, together with the $DefaultCriteria modifier, can be used to create meaningful data "inheritance":

// Define Animal data model
const Animal = Model({
   // The underscore denotes this property as an index
   _name:"", 
   type:"", 
}, "Animal");

/* 
 * Define a "derived" Dog data model (both Dog and Animal will be under the Animals collection), 
 * and assign "Dog" as a default value for `type` for all `Dog` models.
 * We use an underscore as the last character to add `type: "Dog"` - to the "Default Criteria" - 
 * this means that all data retrieval methods will automatically add `type:"Dog"` to their find queries.
 */
const Dog = Animal.derive({ type_:"Dog" });

// The new data object and document (Ubu) will also automatically have `type: "Dog"` associated with it.
let ubu = new Dog("Ubu");

// Get all Dogs
Dog.getAll().then(dogs=> {
   // Got all dogs here.
});

Going back to our "Spaceships":

Let's define a "sub-type" of Spaceship, we'll call it a Battleship, and we'll also add a property to hold its weapons. If we do it like this (note, this is the WRONG way):

class Battleship extends Spaceship {
    constructor (_name,_TYPE) {
        super(_name,_TYPE);
            this.weapons = [];
        }
    }
}

And then create a new instance:

var bship = new Battleship("The Destroyer");

Then we will get an error message on the console: Error in The Destroyer: Trying to set unknown property: weapons (property value is left unchanged). Since weapons is not a known property of the model, as it wasn't defined as part of its structure. That error message, by the way, is displayed via the _error(msg) instance method that all instances have, and is called whenever there is an error related to that instance. The new object will be inserted into the database, it just won't have any weapons property defined on it.

To add new properties to an already existing model - you should use the static derive(modelExtension) method on the Model class. That method use an object defining the additional properties and values that should be added to the parent-model (the "super-model" :) ),

So, the more "right" way to extend the Spaceship to a Battleship in this case is (and let's also set the _TYPE to "Battleship" as well):

class Battleship extends Spaceship
.derive ( { weapons: [] } ) 
{
    constructor (_name) {
        super(_name,"Battleship");
    }
}

var bship = new Battleship("The Destroyer");

The derive method returns a new Model Class (not a subclass of the original Model class) - that shares the same collection as its "parent" data model (so both will be part of the Spaceships collection, and both will use the same existing database synchronization engine (SyncManager).

It is also possible to "override" properties and values in the derived model, and so we can simply "override" the _TYPE property in the derived model, and set its default value to "Battleship". There won't even be a need for a subclass in this case:

var Battleship = Spaceship.derive ({ 
    _TYPE:"Battleship", 
    weapons: [] 
});

var bship = new Battleship("The New Destroyer");

Note: you may not set an existing index as a unique (by adding a $) in a derived model - doing so will have no effect on the index. You may define new indexes - and only objects instances of the derived class will have them. Indexes defined within a derived model - are always defined with the sparse:true property on the Mongo DB.

You can read more about indexes and how they are managed in deriveJS here

Retrieving data from a database

You will often want to "restore" existing database objects from data collections and populate your local ones with the persisted data. Each Model class have various different static methods used to achieve this; most of these are wrappers around certain Mongo find queries, which will make the process easier and more intuitive.

There are several methods that can be used to retrieve data from the database, here is a brief explanation for each:

NOTE: Starting from version 3, all data retrieval functions have an additional boolean collectionWatch argument as their last argument, with false as the default value. Setting this argument to true will enable ChangeStream support for the retrieved data object(s), will make the data object(s) be aware to updates and changes to the DB document from an external source, and reflect those in the local data object(s). Assigning ChangeStream event listeners to data instances creates strong references in memory to them, thereby preventing them from being garbage collected (even if you set them to null or undefined). This can amass to an excess use of memory and can lead to memory leaks - thereby this feature is turned off by default, and should be used sparingly (for example, if you intend to keep references to data object instances indefinitely). Prior versions (2+) had this feature on for all data object instances by default.

get(which, collectionWatch = false)

Returns one object instance.

getAll(which, sortBy, limit=0, skip=0, collectionWatch = false)

Returns all (with an optional filter query) object instances (in an array). There are 3 more additional optional arguments:

  • sortBy - to specify a different index to sort by, using an object such as {<indexName>: <-1 or 1>} - use -1 for descending order, and 1 for ascending (the default is {MainIndex:-1}).
  • limit - to limit the number of returned results (default is 0, which is unlimited).
  • skip to specify an offset index for retrieved results (default is 0).

getAllCursor(which, sortBy, limit=0, skip=0, collectionWatch = false)

This function is identical in functionality to getAll, only it returns a Cursor. The cursor has identical functionality to the Node's MongoDB driver cursor, with one additional function: getNext - which is identical to the cursor's native next function, only it returns a Promise that resolves with the next document as a Derive data object (or null if there are no more documents to retrieve ).

getAllRead(which, sortBy, limit=0, skip=0, collectionWatch = false)

This function is idential to getAll, only it returns "raw" Mongo documents, and not Derive data objects. As with getAll there are some performance degradation when retrieving a large amount of documents, as a new instance is created for each, if you only need to retrieve data for read only purposes, you may use getAllRead instead. Remember that the returned documents will not be reactive.

map(which, index, returnArray, limit=0, skip=0, collectionWatch = false)

map returns all (with an optional filter query) object instances mapped by an index as an object, or as an array (according to the third boolean argument returnArray). The second argument index lets you set the index name that will be used for the mapping (set as null to use the default MainIndex). The 4th and 5th arguments are limit and skip that allows you to get only some of the documents.

mapRead(which, index, returnArray, limit=0, skip=0)

This function is idential to map, only it returns "raw" Mongo documents, and not Derive data objects. As with map there are some performance degradation when retrieving a large amount of documents, as a new instance is created for each, if you only need to retrieve data for read only purposes, you may use mapRead instead. Remember that the returned documents will not be reactive.

has(which, returnDocument, collectionWatch = false)

Returns a boolean indicating if a database collection contains certain value(s). The second returnDocument is a boolean - if the document exist and this argument is set to true - the document will also be returned (as a Derive data object), otherwise false will be returned (regardless of the value of returnDocument).

count(which)

Returns the total number of documents in the DB collection.

All of these methods can use a which argument, and to understand how to use it, you need to know about MainIndex:

MainIndex

MainIndex is an internal value that each model class has and is determined during its definition process.
The MainIndex will contain the "most important" index for that class/collection.
Its value will be determined according to the following:

  • If there is at least one unique index - the first unique index defined on the model will be determined as the MainIndex.
  • If no unique index is defined, then the first non-unique will be determined as the MainIndex.
  • If no index is defined in the model, the _id property will be determined as the MainIndex.

mainIndex() method

You can use the static mainIndex method to get the name of the main index, as a string. This can be useful, for example, when overriding any of the callback methods, e.g.:

const Spaceship = Model({
    _name: "",
    _TYPE: "",
    _inserted() {
       console.log ("Mention the value of this instance's main index: " + this[Spaceship.mainIndex()]);
    }
}, "Spaceship");

which Argument

When you use the which argument, you have two options - if you pass a primitive value (string, number or boolean) - then the function will look for objects where the MainIndex is that value, if you pass an object - then that object will be used as a query object for Mongo's find.
Note if your MainIndex is an object itself - you need to use the normal find query format (e.g. if your index-object is called ob: {ob: {prop1:value,prop2:value}}).

Selective Fields (Projection)

If you'd like to return data objects containing only some of the fields from the model, you may pass an object in which having a fields property containing an array with a list of fields to include in the returned data object.
Notes:
This can be used with any of the data retrieval functions (get function variations, map functions, and join functions).
If you use this, you have to specify your search criteria using the normal Mongo's find syntax.
Only fields you specify in the array in the fields property will be included in the data object instance.
If you include fields that doesn't exist on a data document - the data object will not have that field.
_id field is always returned. You cannot disable this.

Let's see some examples, we assume the models from the previous examples are already defined.

To get our "The Beyond" Spaceship from the database into a local object:

var thebeyond;
Spaceship.get("The Beyond")
.then (ship=> {
	thebeyond = ship;
	// thebeyond contains the object from the DB
});

Or, using the async/await way:

async ()=> {
    var thebeyond = await Spaceship.get("The Beyond");
    // thebeyond contains the object from the DB
});

Put all of our Spaceship objects from the DB into an array:

var spaceships = await Spaceship.getAll();

Get all "_TYPE C" spaceships into an array:

var spaceships = await Spaceship.getAll({_TYPE:"C"});

map - Additional Information

The map method has additional two optional arguments (other than the first which): index: to specify a different index to be used as the key for mapping the objects (other than MainIndex), and returnArray: a boolean specifying if to return the result as an "associative array" of object instances mapped to key indexes or as an object with object instances mapped to index keys.

var spaceships = await Spaceship.map();

spaceships will then be an object, where each key is a _name value, and each value is the equivalent Spaceship object. You can then, for example, reference The Beyond ship from it:

var thebeyond = spaceships["The Beyond"];

map is obviously meant to be used with a unique index. If it's used with a non-unique one - and several objects exist on the database with duplicate values for an index - then some of the objects will be "overriden" when calling map.

So with:

var spaceships = await Spaceship.map({},"_TYPE");

If we have one Spaceship with _TYPE set to C, and several others with _TYPE as en empty string - then map will return only two ships: One of the empty string ones, and the C one.

We can use the which argument to further filter-out the objects, for example - return all _TYPE C ships (they will still be mapped to their _names):

var spaceships = await Spaceship.map({_TYPE:"C"});

The last method has can be used to determine if a certain object exist on the database. For example:

var hasTheBeyond = await Spaceship.has("The Beyond");

Will return true if a Spaceship object with its _name set to "The Beyond" exist on the DB. The method also has a second argument - returnDocument, a boolean that if set to true will also return the object if it exist on the DB (or false if it doesn't).

A Word of Caution for When Using the get Functions

Upon retrieving the objects from the database - their constructor functions will be called for each object. Therefore - if you override the constructor and have any code that affects or changes the data there - it will run - that is usually not desired when retrieving data object, so you should make sure you call the get functions from a (usually "higher") class that runs a constructor that does not change the data (like the default constructor).

$DefaultCriteria

All model classes have a "static" (meta) property - $DefaultCriteria, this is an object containing key-value pairs that will be added by default to database queries, when calling one of the get methods, using a primitive value as an argument. This is useful when creating derived or extended classes, and wanting to include just the "sub-classes" in query results. It will be more clear with an example:

Returning to our Spaceship example, we define a Spaceship super-model, as before:

var Spaceship = Model({
    _name$: "",
    TYPE: "",
    crew: [],
}, "Spaceship");

Then we define a sub-model of Spaceship: Battleship

var Battleship = Spaceship.derive ({ 
    _TYPE:"Battleship", 
    weapons: [] 
});

If we call getAll on Battleship:

var await battleships = Battleship.getAll();

then we'll get all Spaceship objects in the entire Spaceship collection into battleships, while we probably only want to get back the "derived" Battleship objects. When we want to differentiate the Battleship objects from the others in the collection - we can add certain key:value pairs into the $DefaultCriteria; to do so, we add an underscore (_) character to the end of the property name. So, we can add the _TYPE in Battleship:

var Battleship = Spaceship.derive ({ 
    _TYPE_:"Battleship",
    weapons: [] 
});

When we do that, then _TYPE:"Battleship" will be added to the find query when we call get methods from Battleship, and so, for example, calling getAll:

var await battleships = Battleship.getAll();

Will now return just the Spaceship objects that also has their _TYPE set to "Battleship".

Note: You shouldn't access $DefaultCriteria directly. In case you do, make sure you set its properties, and not override it completely.

Using Model Instances as Values in Other Models

Let's define a new model: CrewMember. The crew array of our Spaceships will contain CrewMember objects:

var CrewMember = Model({
    _name: "",
    rank: "",
    role: ""
}, "CrewMember");

We can also add a function to the Spaceship model, to handle adding new CrewMembers to the crew array, (and let's also add a captain property)

var Spaceship = Model({
    _name$: "",
    TYPE: "",
    captain: null,
    crew: [],
    addCrew: function(crewMember) {
        this.crew.push(crewMember);
    }
},"Spaceship");

Next, we create a new CrewMember - captain Ricard:

var ricard = new CrewMember("Ricard");
ricard.rank = "captain";
ricard.role = "captain";

We get "The Beyond" spaceship, add ricard to its crew, and also set it as its captain:

var thebeyond = await Spaceship.get("The Beyond");
thebeyond.addCrew(ricard);
thebeyond.captain = ricard;

Now if we look at The Beyond in the database, we can see:

> db.Spaceships.find({_name:"The Beyond"})
{ "_id" : ObjectId("5a0f42663432962910d45ab7"), "_name" : "The Beyond", "_TYPE" : "", "crew" : [ DBRef("CrewMembers", ObjectId("5a1c9a63b3c2df42587b6a90")) ], "captain" : DBRef("CrewMembers", ObjectId("5a1c9a63b3c2df42587b6a90")) }

The CrewMember object was saved as a DBRef object. A DBRef is a "reference" to an item from another collection (in this case the CrewMembers collection). It contains the ObjectId of the item, and the collection name. To "dereference" the object, you can use the get function by passing it the DBRef itself. So, if we want to get back captain Ricard, from The Beyond's crew array, we can use:

var thebeyond = await Spaceship.get("The Beyond");
var ricard = await CrewMember.get(thebeyond._captain);

Indexes and How They are Handled in the Mongo Server

Indexes in collections are saved in 4 different compound indexes (specified by their index name:)

  • "nonUnique": an index containing all non unique (non-sparse) indexes
  • "unique": an index containing all unique (non-sparse) indexes
  • "sparse_nonUnique": an index containing all sparse non-unique indexes -- non-unique indexes defined in derived models.
  • "sparse_unique": an index containing all sparse unique indexes -- unique indexes defined in derived models.

Collections may have only some or none of these indexes defined, depending on indexes defined on the model. There will also be the _id index defined, as usual.

Putting it All Together:

The following is a complete demonstration, expanding on the Spaceship idea and models.
The following code examples can also be found in this github repository.

Defining Our Models

First we define all of our data models, in a separate Models.js file:

module.exports = new Promise( (resolve,reject)=> {
    var Models = {};

    const derive = require('derivejs');

    derive.Model
    ({
        dbUrl: "mongodb://localhost:27017/",
        dbName: "spaceshipyard",
        debugMode: false
    })
    .then(
        async Model=> {

            Models.Weapon = Model({
                _TYPE:"",
                _DAMAGE:-1,
                armed: false,
                arm: function() {
                    this.armed = true;
                },
                unarm: function() {
                    this.armed = false;
                },
                fire: function(target) {
                    if (!this.armed) {
                        console.log ("Weapon is not armed!");
                        return;
                    }
                    if (target.shields.up) {
                        target.shields.percent -= this._DAMAGE;
                    }
                    else {
                        target.integrityHull -= this._DAMAGE;
                    }
                },
                _inserted: function() {
                    if (this.$ready) this.$ready.call(this);
                },
                // For a weapon-ready callback
                $ready: null
            }, "Weapon");
            
            Models.PhotonTorpedos = class extends Models.Weapon
            .derive({
                _TYPE_: "Photon Torpedos",
                _DAMAGE: 20
            }) {
                constructor(readyCallback) {
                    super ();
                    this.$ready = readyCallback;
                }
            };
            
            Models.CrewMember = Model({
                _name: "",
                rank: "",
                role: ""
            }, "CrewMember");
            
            Models.Spaceship = Model({
                _name: "",
                TYPE: "",
                shields: {
                    up: false,
                    percent: 100
                },
                integrityHull: 100,
                crew: [],
                addCrew: function (crewMember) {
                    this.crew.push(crewMember);
                },
                raiseShields: function() {
                    this.shields.up = true;
                    console.log (this._name+": shields are up");
                },
                lowerShields: function() {
                    this.shields.up = false;
                    console.log (this._name+": shields are down");
                },
                captain: "",
                // Listen to changes on this properties
                $Listen: [ "shields.percent", "integrityHull" ]
            }, "Spaceship");
            
            Models.Cruiser = Models.Spaceship
            .derive({
                TYPE_: "Cruiser"
            });

            Models.Battleship = Models.Spaceship
            .derive({
                TYPE_: "Battleship",
                weapons: [],
                attack: function(target, weaponIndex) {
                    Models.Weapon.get(this.weapons[weaponIndex])
                    .then(w=> {
                        if (!w.armed) w.arm();
                        w.fire(target);
                    });
                }
            });

            resolve (Models);
        }
    )
});

Writing Our App

The app code will be in an app.js file:

require ("./Models")
.then(async Models=> {

    // Save convient references to our data models
    const PhotonTorpedos  = Models.PhotonTorpedos;
    const CrewMember = Models.CrewMember;
    const Cruiser = Models.Cruiser;
    const Battleship = Models.Battleship;

    // Will contain data instances
    var BoldlyGo, Feisty,
        Ricard, Wort,
        torpedos;

    function clearAll() {
        return Promise.all([
            CrewMember.clear(),
            Cruiser.clear(),
            Battleship.clear()
        ]);
    }
    async function init() {
        console.log ("Creating Boldly Go Cruiser");
        BoldlyGo = new Cruiser ("The Boldly Go");
        console.log ("Creating Feisty Battleship");
        Feisty = new Battleship ("The Feisty");

        console.log ("Creating Ricard Crew Member");
        Ricard = new CrewMember("Ricard");
        Ricard.role = "captain";
        Ricard.rank = "captain";

        console.log ("Adding Ricard to Boldly Go")
        BoldlyGo.addCrew (Ricard);
        BoldlyGo.captain = Ricard;

        console.log ("Creating Wort Crew Member");
        Wort = new CrewMember("Wort");
        Wort.role = "captain";
        Wort.rank = "commander";

        console.log ("Adding Wort To Feisty");
        Feisty.addCrew(Wort);
        Feisty.captain = Wort;

    }

    function restore() {
        return Promise.all([
            Cruiser.get("The Boldly Go"),
            Battleship.get("The Feisty")
        ]);
    }

    function battle() {
        console.log ("Starting battle");
        BoldlyGo.raiseShields();
        Feisty.attack(BoldlyGo, 0);
    }

    console.log ("Clearing all...");
    await clearAll();
    init();
    console.log ("Adding Photon Torpedos to Feisty");
    // Wait until PhotonTorpedos are added to Feisty, then run a battle
    Feisty.weapons.push( new PhotonTorpedos(function() {
        console.log ("Photon Torpedos ready")
        battle();
    }) );

    /* 
    // You can also use Spaceship.map, but this is done here for the sake of example:
    restore()
    .then(ships=> {
        BoldlyGo = ships[0];
        Feisty   = ships[1];
        BoldlyGo.lowerShields();
        Feisty.attack(BoldlyGo, 0);
    });
    */
    
})
.catch (err=> {
    console.log ("Error initializing models: ",err);
});

If we run the app we can see all the data creation messages appearing asynchronously, along with the battle messages:

Starting battle
The Boldly Go: shields are up
The Boldly Go: shields.percent changed from 100 to 80 

The Boldly Go has raised its shields, and was then attacked by the Feisty's Photon Torpedos, lowering its shields from 100 percent to 80 percent. These last message is from the built-in changed method, as we registered shields.percent for listening.

If we look at our database in our Spaceship collection, we can see The Boldly Go record with shields.percent as 80.

If we now comment the last section, and comment-out this section:

restore()
.then(ships=> {
    BoldlyGo = ships[0];
    Feisty   = ships[1];
    BoldlyGo.lowerShields();
    Feisty.attack(BoldlyGo, 0);
});

Then run the app again, then we will eventually see these messages:

The Boldly Go: shields are down
The Boldly Go: integrityHull changed from 100 to 80

The last one is from the change method. After we restore our ships data from the database - then The Boldly Go lowers its shields, and attacked by Feisty again - with the shields down - The Boldly Go "integrity hull" suffers a damage of 20 percent.

Advanced Subjects

Join

You can perform a "join" query with the join static method all model classes have: join(which, joinWith, localField, foreignField, joinAs, returnAsModel=false, collectionWatch = false)

  • which is the criteria for the document to retrieve from the "primary" ("local" collection),
  • joinWith is the name of the "secondary" ("foreign") collection (as a string).
  • localField is the name of the field that is equivalent to the foreignField on the secondary collection.
  • joinAs is the name of a property where the "joined" document will be included into. returnAsModel - if set to true, then the function will return an instance of the model (as in when using the get function) - you will most likely not want to set it to true, as the model will have "foreign" fields - and once you try setting or changing them - it will try to persist it to the DB. This function is usually used only for getting "readonly" data, and not data you want to modify or change.

An Example Use Case:

Let's say you have a Posts collection and a User collection, and you want to get the data for a certain post, and join it with the user data of the user who posted it. With the following assumptions:

  • You have a Post model defined, with _email as its primary key, and _authorId with a string id containing the id of the user who posted it.
  • Your Users collection documents have a _userId field with string ids
Post.join("[email protected]","Users","_authorId","_userId","author").then(
    post=> {
        // Now the post object here, will also have an "author" field containing all the data for the user with _authorId/_userId
    }
);

Note: Use join to join with a single document from another collection.

joinAll

  • joinAll lets you join data from two separate collections, and return multiple results. The function accepts the following arguments:
  • which is the criteria for the document to retrieve from the "primary" ("local" collection),
  • joinOpts is an object with different options regarding the join:
    • joinWith: the name of the "secondary" ("foreign") collection (as a string).
    • localField: the name of the field that is equivalent to the foreignField on the secondary collection.
    • foreignField: the name of the field on the joined collection, equivalent to localField
    • joinAs is the name of a property where the "joined" document will be included into.
  • findOpts lets you specify additional "post-find" options, an object that can contain the following:
    • sortBy: to return results sorted by a certain index, use Mongo's format for a sort-object, e.g.: {_date:-1}, this will sort by the _date index in a descending order. To sort in an ascending order, use 1 (positive 1) as the value.
    • skip: lets you skip a number of results,
    • limit: lets you limit the number of results returned.
  • returnAsModel if set to true, then the function will return an instance of the model (as in when using the get function) - see notes about this in the documentation for join, and why you should almost never need to set this to true.
  • collectionWatch - if set to true, will turn on ChangeStream support for returned data objects.

Access the "raw" mongoDB collection object

Although DeriveJS is designed, written, and intended to be in charge of all data persistence operations transparently in the background without direct interference, there might come a rare occasion where you will need access to the collection object, to perform native MongoDB operations "yourself" (something that should generally be avoided, and should rarely happen - if you encounder a native MongoDB operation that DeriveJS doesn't enable - I would appreciate if you contact me via Github and tell me about it). To get access to the MongoDB collection associated with a data model class, you can call the static method collection() of the class, which will return the assosicated NodeJS MongoDB driver collection object.

Mixins

"Mixins" are like "plugins" adding additional functionality to a Model, and can be used on multiple models. To implement a mixin, you use the use static function on a Model class, giving it an object literal with functions only. Those functions will be available to be used by the Model instances. A good example will be a "Logger" mixin - that adds the functionality to a model - to write "log" messages to a "Logs collection. For example we can create this module:

Logger.js

module.exports = ()=>
new Promise(async (resolve, reject)=> {
    let Model = await require('derivejs').Model({
        dbUrl: "mongodb://localhost:27017/",
        dbName: "Logs"
    });

    let Log = Model({
        _logMessage: "",
        _date: null
    },"Log");

    resolve({
        log(msg) { new Log(msg, Date.now()); }
    });
});

Then we can implement the log function, for example on a Spaceship model:

var Spaceship = Model({
    _name: "",
    _TYPE: "",
    crew: []
}, "Spaceship");

let Logger = await require('./Logger')();
Spaceship.use(Logger);

let ship = new Spaceship("The Logger").log("The Logger ship created");

Notice that in this example - we create a *separate connection to the DB, for the Logs collection - but this could be done on the same connection with the rest of the models (however it might be a good practice to separate the connection for things such as logging).

collectionReady() (new in version 1.4.0+)

Whenever a DB operation related to a collection occurs - if that collection doesn't exist yet in the DB, MongoDB implicitly creates it, this can take time (in the area of ~1 second), and thus if you run one of the data getter functions (get, getAll, map etc.) - and the collection was not yet created when reaching that point in your code - you will get an error, and the getter function will fail. To prevent this - the collectionReady static method was added to Model classes - it returns a promise that resolves when the collection is created and ready for any operations. If the collection already exists the function will resolve immediately. Use collectionReady in situations where it's not certain that a collection exist, and you need to run a getter function on it.

A note about scaled systems

When having several nodes that needs to access the same collections on a same database - you should avoid running Derive on each, as this could lead to unwanted race conditions -- instead, it is recommended to create a single Node that communicates directly with the DB using Derive, then have the separate nodes communicate with it via message queue - that describe the types of data operations needs to be done - this way, allowing to keep the single node to DB architecture, while still having several nodes running concurrently, sending data operations.

About

DeriveODM is a reactive ODM - Object Document Mapper - framework, a "wrapper" around MongoDB, that removes all the hassle of data-persistence by handling it transparently in the background, in a DRY manner.

Topics

Resources

License

Stars

Watchers

Forks

Packages

No packages published