The store contains all of the data for records loaded from the server. It is also responsible for creating instances of DS.Model that wrap the individual data for a record, so that they can be bound to in your Handlebars templates.

Define your application's store like this:

import DS from 'ember-data';

export default DS.Store.extend({
});

Most Ember.js applications will only have a single DS.Store that is automatically created by their Ember.Application.

You can retrieve models from the store in several ways. To retrieve a record for a specific id, use DS.Store's findRecord() method:

store.findRecord('person', 123).then(function (person) {
});

By default, the store will talk to your backend using a standard REST mechanism. You can customize how the store talks to your backend by specifying a custom adapter:

import DS from 'ember-data';

export default DS.Adapter.extend({
});

You can learn more about writing a custom adapter by reading the DS.Adapter documentation.

Store createRecord() vs. push() vs. pushPayload()

The store provides multiple ways to create new record objects. They have some subtle differences in their use which are detailed below:

createRecord is used for creating new records on the client side. This will return a new record in the created.uncommitted state. In order to persist this record to the backend you will need to call record.save().

push is used to notify Ember Data's store of new or updated records that exist in the backend. This will return a record in the loaded.saved state. The primary use-case for store#push is to notify Ember Data about record updates (full or partial) that happen outside of the normal adapter methods (for example SSE or Web Sockets).

pushPayload is a convenience wrapper for store#push that will deserialize payloads if the Serializer implements a pushPayload method.

Note: When creating a new record using any of the above methods Ember Data will update DS.RecordArrays such as those returned by store#peekAll(), store#findAll() or store#filter(). This means any data bindings or computed properties that depend on the RecordArray will automatically be synced to include the new or updated record values.

Returns the normalized (dasherized) modelName. This method should be used whenever receiving a modelName in a public method.

If the adapter updates attributes the record will notify the store to update its membership in any filters. To avoid thrashing, this method is invoked only once per run loop per record.

When a record is destroyed, this un-indexes it and removes it from any record arrays so it can be GCed.

This method is called by findRecord if it discovers that a particular type/id pair hasn't been loaded yet to kick off a request to the adapter.

If possible, this method asks the adapter to generate an ID for a newly created record.

This internal method is used by push.

Returns a map of IDs to client IDs for a given modelName.

Returns an instance of the adapter for a given type. For example, adapterFor('person') will return an instance of App.PersonAdapter.

If no App.PersonAdapter is found, this method will look for an App.ApplicationAdapter (the default adapter for your entire application).

If no App.ApplicationAdapter is found, it will return the value of the defaultAdapter.

Build a brand new record for a given type, ID, and initial data.

Create a new record in the current store. The properties passed to this method are set on the newly created record.

To create a new instance of a Post:

store.createRecord('post', {
  title: "Rails is omakase"
});

To create a new instance of a Post that has a relationship with a User record:

let user = this.store.peekRecord('user', 1);
store.createRecord('post', {
  title: "Rails is omakase",
  user: user
});

For symmetry, a record can be deleted via the store.

Example

let post = store.createRecord('post', {
  title: "Rails is omakase"
});

store.deleteRecord(post);

This method is called once the promise returned by an adapter's createRecord, updateRecord or deleteRecord is resolved.

If the data provides a server-generated ID, it will update the record and the store's indexes.

Takes a type and filter function, and returns a live RecordArray that remains up to date as new records are loaded into the store or created locally.

The filter function takes a materialized record, and returns true if the record should be included in the filter and false if it should not.

Example

store.filter('post', function(post) {
  return post.get('unread');
});

The filter function is called once on all records for the type when it is created, and then once on each newly loaded or created record.

If any of a record's properties change, or if it changes state, the filter function will be invoked again to determine whether it should still be in the array.

Optionally you can pass a query, which is the equivalent of calling query with that same query, to fetch additional records from the server. The results returned by the server could then appear in the filter if they match the filter function.

The query itself is not used to filter records, it's only sent to your server for you to be able to do server-side filtering. The filter function will be applied on the returned results regardless.

Example

store.filter('post', { unread: true }, function(post) {
  return post.get('unread');
}).then(function(unreadPosts) {
  unreadPosts.get('length'); // 5
  let unreadPost = unreadPosts.objectAt(0);
  unreadPost.set('unread', false);
  unreadPosts.get('length'); // 4
});

findAll asks the adapter's findAll method to find the records for the given type, and returns a promise which will resolve with all records of this type present in the store, even if the adapter only returns a subset of them.

import Ember from 'ember';

export default Ember.Route.extend({
  model: function(params) {
    return this.store.findAll('author');
  }
});

When the returned promise resolves depends on the reload behavior, configured via the passed options hash and the result of the adapter's shouldReloadAll method.

Reloading

If { reload: true } is passed or adapter.shouldReloadAll evaluates to true, then the returned promise resolves once the adapter returns data, regardless if there are already records in the store:

store.push({
  data: {
    id: 'first',
    type: 'author'
  }
});

// adapter#findAll resolves with
// [
//   {
//     id: 'second',
//     type: 'author'
//   }
// ]
store.findAll('author', { reload: true }).then(function(authors) {
  authors.getEach("id"); // ['first', 'second']
});

If no reload is indicated via the abovementioned ways, then the promise immediately resolves with all the records currently loaded in the store.

Background Reloading

Optionally, if adapter.shouldBackgroundReloadAll evaluates to true, then a background reload is started. Once this resolves, the array with which the promise resolves, is updated automatically so it contains all the records in the store:

// app/adapters/application.js
export default DS.Adapter.extend({
  shouldReloadAll(store, snapshotsArray) {
    return false;
  },

  shouldBackgroundReloadAll(store, snapshotsArray) {
    return true;
  }
});

// ...

store.push({
  data: {
    id: 'first',
    type: 'author'
  }
});

let allAuthors;
store.findAll('author').then(function(authors) {
  authors.getEach('id'); // ['first']

  allAuthors = authors;
});

// later, once adapter#findAll resolved with
// [
//   {
//     id: 'second',
//     type: 'author'
//   }
// ]

allAuthors.getEach('id'); // ['first', 'second']

If you would like to force or prevent background reloading, you can set a boolean value for backgroundReload in the options object for findAll.

import Ember from 'ember';

export default Ember.Route.extend({
  model: function() {
    return this.store.findAll('post', { backgroundReload: false });
  }
});

If you pass an object on the adapterOptions property of the options argument it will be passed to you adapter via the snapshotRecordArray

import Ember from 'ember';

export default Ember.Route.extend({
  model: function(params) {
    return this.store.findAll('post', {
      adapterOptions: { subscribe: false }
    });
  }
});

import MyCustomAdapter from './custom-adapter';

export default MyCustomAdapter.extend({
  findAll: function(store, type, sinceToken, snapshotRecordArray) {
    if (snapshotRecordArray.adapterOptions.subscribe) {
      // ...
    }
    // ...
  }
});

See peekAll to get an array of current records in the store, without waiting until a reload is finished.

Retrieving Related Model Records

If you use an adapter such as Ember's default JSONAPIAdapter that supports the JSON API specification and if your server endpoint supports the use of an 'include' query parameter, you can use findAll() to automatically retrieve additional records related to those requested by supplying an include parameter in the options object.

For example, given a post model that has a hasMany relationship with a comment model, when we retrieve all of the post records we can have the server also return all of the posts' comments in the same request:

import Ember from 'ember';

export default Ember.Route.extend({
  model: function() {
   return this.store.findAll('post', {include: 'comments'});
  }
});

Multiple relationships can be requested using an include parameter consisting of a comma-separated list (without white-space) while nested relationships can be specified using a dot-separated sequence of relationship names. So to request both the posts' comments and the authors of those comments the request would look like this:

import Ember from 'ember';

export default Ember.Route.extend({
  model: function() {
   return this.store.findAll('post', {include: 'comments,comments.author'});
  }
});

See query to only get a subset of records from the server.

This method makes a series of requests to the adapter's find method and returns a promise that resolves once they are all loaded.

If a relationship was originally populated by the adapter as a link (as opposed to a list of IDs), this method is called when the relationship is fetched.

The link (which is usually a URL) is passed through unchanged, so the adapter can make whatever request it wants.

The usual use-case is for the server to register a URL as a link, and then use that URL in the future to make a request for the relationship.

This method returns a record for a given type and id combination.

The findRecord method will always resolve its promise with the same object for a given type and id.

The findRecord method will always return a promise that will be resolved with the record.

Example

import Ember from 'ember';

export default Ember.Route.extend({
  model: function(params) {
    return this.store.findRecord('post', params.post_id);
  }
});

If the record is not yet available, the store will ask the adapter's find method to find the necessary data. If the record is already present in the store, it depends on the reload behavior when the returned promise resolves.

Reloading

The reload behavior is configured either via the passed options hash or the result of the adapter's shouldReloadRecord.

If { reload: true } is passed or adapter.shouldReloadRecord evaluates to true, then the returned promise resolves once the adapter returns data, regardless if the requested record is already in the store:

store.push({
  data: {
    id: 1,
    type: 'post',
    revision: 1
  }
});

// adapter#findRecord resolves with
// [
//   {
//     id: 1,
//     type: 'post',
//     revision: 2
//   }
// ]
store.findRecord('post', 1, { reload: true }).then(function(post) {
  post.get("revision"); // 2
});

If no reload is indicated via the abovementioned ways, then the promise immediately resolves with the cached version in the store.

Background Reloading

Optionally, if adapter.shouldBackgroundReloadRecord evaluates to true, then a background reload is started, which updates the records' data, once it is available:

// app/adapters/post.js
import ApplicationAdapter from "./application";

export default ApplicationAdapter.extend({
  shouldReloadRecord(store, snapshot) {
    return false;
  },

  shouldBackgroundReloadRecord(store, snapshot) {
    return true;
  }
});

// ...

store.push({
  data: {
    id: 1,
    type: 'post',
    revision: 1
  }
});

let blogPost = store.findRecord('post', 1).then(function(post) {
  post.get('revision'); // 1
});

// later, once adapter#findRecord resolved with
// [
//   {
//     id: 1,
//     type: 'post',
//     revision: 2
//   }
// ]

blogPost.get('revision'); // 2

If you would like to force or prevent background reloading, you can set a boolean value for backgroundReload in the options object for findRecord.

import Ember from 'ember';

export default Ember.Route.extend({
  model: function(params) {
    return this.store.findRecord('post', params.post_id, { backgroundReload: false });
  }
});

If you pass an object on the adapterOptions property of the options argument it will be passed to you adapter via the snapshot

import Ember from 'ember';

export default Ember.Route.extend({
  model: function(params) {
    return this.store.findRecord('post', params.post_id, {
      adapterOptions: { subscribe: false }
    });
  }
});

import MyCustomAdapter from './custom-adapter';

export default MyCustomAdapter.extend({
  findRecord: function(store, type, id, snapshot) {
    if (snapshot.adapterOptions.subscribe) {
      // ...
    }
    // ...
  }
});

See peekRecord to get the cached version of a record.

Retrieving Related Model Records

If you use an adapter such as Ember's default JSONAPIAdapter that supports the JSON API specification and if your server endpoint supports the use of an 'include' query parameter, you can use findRecord() to automatically retrieve additional records related to the one you request by supplying an include parameter in the options object.

For example, given a post model that has a hasMany relationship with a comment model, when we retrieve a specific post we can have the server also return that post's comments in the same request:

import Ember from 'ember';

export default Ember.Route.extend({
  model: function(params) {
   return this.store.findRecord('post', params.post_id, {include: 'comments'});
  }
});

In this case, the post's comments would then be available in your template as model.comments.

Multiple relationships can be requested using an include parameter consisting of a comma-separated list (without white-space) while nested relationships can be specified using a dot-separated sequence of relationship names. So to request both the post's comments and the authors of those comments the request would look like this:

import Ember from 'ember';

export default Ember.Route.extend({
  model: function(params) {
   return this.store.findRecord('post', params.post_id, {include: 'comments,comments.author'});
  }
});

This method is called at the end of the run loop, and flushes any records passed into scheduleSave

Get the reference for the specified record.

Example

let userRef = store.getReference('user', 1);

// check if the user is loaded
let isLoaded = userRef.value() !== null;

// get the record of the reference (null if not yet available)
let user = userRef.value();

// get the identifier of the reference
if (userRef.remoteType() === "id") {
let id = userRef.id();
}

// load user (via store.find)
userRef.load().then(...)

// or trigger a reload
userRef.reload().then(...)

// provide data for reference
userRef.push({ id: 1, username: "@user" }).then(function(user) {
userRef.value() === user;
});

This method returns true if a record for a given modelName and id is already loaded in the store. Use this function to know beforehand if a findRecord() will result in a request or that it will be a cache hit.

Example

store.hasRecordForId('post', 1); // false
store.findRecord('post', 1).then(function() {
   store.hasRecordForId('post', 1); // true
 });

Returns the model class for the particular modelName.

The class of a model might be useful if you want to get a list of all the relationship names of the model, see relationshipNames for example.

normalize converts a json payload into the normalized form that push expects.

Example

socket.on('message', function(message) {
  let modelName = message.model;
  let data = message.data;
  store.push(store.normalize(modelName, data));
});

This method returns a filtered array that contains all of the known records for a given type in the store.

Note that because it's just a filter, the result will contain any locally created records of the type, however, it will not make a request to the backend to retrieve additional records. If you would like to request all the records from the backend please use store.findAll.

Also note that multiple calls to peekAll for a given type will always return the same RecordArray.

Example

let localPosts = store.peekAll('post');

Get a record by a given type and ID without triggering a fetch.

This method will synchronously return the record if it is available in the store, otherwise it will return null. A record is available if it has been fetched earlier, or pushed manually into the store.

Note: This is an synchronous method and does not return a promise.

let post = store.peekRecord('post', 1);

post.get('id'); // 1

Push some data for a given type into the store.

This method expects normalized JSON API document. This means you have to follow JSON API specification with few minor adjustments: - record's type should always be in singular, dasherized form - members (properties) should be camelCased

Your primary data should be wrapped inside data property:

store.push({
  data: {
    // primary data for single record of type `Person`
    id: '1',
    type: 'person',
    attributes: {
      firstName: 'Daniel',
      lastName: 'Kmak'
    }
  }
});

Demo.

data property can also hold an array (of records):

store.push({
  data: [
    // an array of records
    {
      id: '1',
      type: 'person',
      attributes: {
        firstName: 'Daniel',
        lastName: 'Kmak'
      }
    },
    {
      id: '2',
      type: 'person',
      attributes: {
        firstName: 'Tom',
        lastName: 'Dale'
      }
    }
  ]
});

Demo.

There are some typical properties for JSONAPI payload: * id - mandatory, unique record's key * type - mandatory string which matches model's dasherized name in singular form * attributes - object which holds data for record attributes - DS.attr's declared in model * relationships - object which must contain any of the following properties under each relationships' respective key (example path is relationships.achievements.data): - links - data - place for primary data - meta - object which contains meta-information about relationship

For this model:

import DS from 'ember-data';

export default DS.Model.extend({
  firstName: DS.attr('string'),
  lastName: DS.attr('string'),

  children: DS.hasMany('person')
});

To represent the children as IDs:

{
  data: {
    id: '1',
    type: 'person',
    attributes: {
      firstName: 'Tom',
      lastName: 'Dale'
    },
    relationships: {
      children: {
        data: [
          {
            id: '2',
            type: 'person'
          },
          {
            id: '3',
            type: 'person'
          },
          {
            id: '4',
            type: 'person'
          }
        ]
      }
    }
  }
}

Demo.

To represent the children relationship as a URL:

{
  data: {
    id: '1',
    type: 'person',
    attributes: {
      firstName: 'Tom',
      lastName: 'Dale'
    },
    relationships: {
      children: {
        links: {
          related: '/people/1/children'
        }
      }
    }
  }
}

If you're streaming data or implementing an adapter, make sure that you have converted the incoming data into this form. The store's normalize method is a convenience helper for converting a json payload into the form Ember Data expects.

store.push(store.normalize('person', data));

This method can be used both to push in brand new records, as well as to update existing records.

Push some raw data into the store.

This method can be used both to push in brand new records, as well as to update existing records. You can push in more than one type of object at once. All objects should be in the format expected by the serializer.

import DS from 'ember-data';

export default DS.ActiveModelSerializer;

let pushData = {
  posts: [
    { id: 1, post_title: "Great post", comment_ids: [2] }
  ],
  comments: [
    { id: 2, comment_body: "Insightful comment" }
  ]
}

store.pushPayload(pushData);

By default, the data will be deserialized using a default serializer (the application serializer if it exists).

Alternatively, pushPayload will accept a model type which will determine which serializer will process the payload.

import DS from 'ember-data';

export default DS.ActiveModelSerializer;

import DS from 'ember-data';

export default DS.JSONSerializer;

store.pushPayload('comment', pushData); // Will use the application serializer
store.pushPayload('post', pushData); // Will use the post serializer

This method delegates a query to the adapter. This is the one place where adapter-level semantics are exposed to the application.

Exposing queries this way seems preferable to creating an abstract query language for all server-side queries, and then require all adapters to implement them.

If you do something like this:

store.query('person', { page: 1 });

The call made to the server, using a Rails backend, will look something like this:

Started GET "/api/v1/person?page=1"
Processing by Api::V1::PersonsController#index as HTML
Parameters: { "page"=>"1" }

If you do something like this:

store.query('person', { ids: [1, 2, 3] });

The call to the server, using a Rails backend, will look something like this:

Started GET "/api/v1/person?ids%5B%5D=1&ids%5B%5D=2&ids%5B%5D=3"
Processing by Api::V1::PersonsController#index as HTML
Parameters: { "ids" => ["1", "2", "3"] }

This method returns a promise, which is resolved with an AdapterPopulatedRecordArray once the server returns.

This method makes a request for one record, where the id is not known beforehand (if the id is known, use findRecord instead).

This method can be used when it is certain that the server will return a single object for the primary data.

Let's assume our API provides an endpoint for the currently logged in user via:

// GET /api/current_user
{
  user: {
    id: 1234,
    username: 'admin'
  }
}

Since the specific id of the user is not known beforehand, we can use queryRecord to get the user:

store.queryRecord('user', {}).then(function(user) {
  let username = user.get('username');
  console.log(`Currently logged in as ${username}`);
});

The request is made through the adapters' queryRecord:

import DS from "ember-data";

export default DS.Adapter.extend({
  queryRecord(modelName, query) {
    return Ember.$.getJSON("/api/current_user");
  }
});

Note: the primary use case for store.queryRecord is when a single record is queried and the id is not known beforehand. In all other cases store.query and using the first item of the array is likely the preferred way:

// GET /users?username=unique
{
  data: [{
    id: 1234,
    type: 'user',
    attributes: {
      username: "unique"
    }
  }]
}

store.query('user', { username: 'unique' }).then(function(users) {
  return users.get('firstObject');
}).then(function(user) {
  let id = user.get('id');
});

This method returns a promise, which resolves with the found record.

If the adapter returns no data for the primary data of the payload, then queryRecord resolves with null:

// GET /users?username=unique
{
  data: null
}

store.queryRecord('user', { username: 'unique' }).then(function(user) {
  console.log(user); // null
});

Returns id record for a given type and ID. If one isn't already loaded, it builds a new record and leaves it in the empty state.

This method has been deprecated and is an alias for store.hasRecordForId, which should be used instead.

This method is called once the promise returned by an adapter's createRecord, updateRecord or deleteRecord is rejected (with anything other than a DS.InvalidError).

This method is called once the promise returned by an adapter's createRecord, updateRecord or deleteRecord is rejected with a DS.InvalidError.

This method is called by the record's reload method.

This method calls the adapter's find method, which returns a promise. When that promise resolves, reloadRecord will resolve the promise returned by the record's reload.

This method is called by record.save, and gets passed a resolver for the promise that record.save returns.

It schedules saving to happen at the end of the run loop.

Returns a JSON representation of the record using a custom type-specific serializer, if one exists.

The available options are:

• includeId: true if the record's ID should be included in the JSON representation

Returns an instance of the serializer for a given type. For example, serializerFor('person') will return an instance of App.PersonSerializer.

If no App.PersonSerializer is found, this method will look for an App.ApplicationSerializer (the default serializer for your entire application).

if no App.ApplicationSerializer is found, it will attempt to get the defaultSerializer from the PersonAdapter (adapterFor('person')).

If a serializer cannot be found on the adapter, it will fall back to an instance of DS.JSONSerializer.

This method unloads all records in the store. It schedules unloading to happen during the next run loop.

Optionally you can pass a type which unload all records for a given type.

store.unloadAll();
store.unloadAll('post');

For symmetry, a record can be unloaded via the store. Only non-dirty records can be unloaded.

Example

store.findRecord('post', 1).then(function(post) {
  store.unloadRecord(post);
});

When an adapter's createRecord, updateRecord or deleteRecord resolves with data, this method extracts the ID from the supplied data.

The default adapter to use to communicate to a backend server or other persistence layer. This will be overridden by an application adapter if present.

If you want to specify app/adapters/custom.js as a string, do:

import DS from 'ember-data';

export default DS.Store.extend({
  adapter: 'custom',
});

This property returns the adapter, after resolving a possible string key.

If the supplied adapter was a class, or a String property path resolved to a class, this property will instantiate the class.

This property is cacheable, so the same instance of a specified adapter class should be used for the lifetime of the store.