Customizing Serializers

When requesting a record, the JSONAPISerializer expects your server to return a JSON representation of the record that conforms to the following conventions.

The JSONAPISerializer expects the backend to return a JSON:API Document that follows the JSON:API specification and the conventions of the examples found in the JSON:API spec. This means all type names should be pluralized and attribute and relationship names should be dash-cased. For example, if you request a record from /people/123, the response should look like this:

{
  "data": {
    "type": "people",
    "id": "123",
    "attributes": {
      "given-name": "Jeff",
      "family-name": "Atwood"
    }
  }
}

A response that contains multiple records may have an array in its data property.

{
  "data": [{
    "type": "people",
    "id": "123",
    "attributes": {
      "given-name": "Jeff",
      "family-name": "Atwood"
    }
  }, {
    "type": "people",
    "id": "124",
    "attributes": {
      "given-name": "Yehuda",
      "family-name": "Katz
    }
  }]
}

Data that is not a part of the primary request but includes linked relationships should be placed in an array under the included key. For example, if you request /articles/1 and the backend also returned any comments associated with that person the response should look like this:

{
  "data": {
    "type": "articles",
    "id": "1",
    "attributes": {
      "title": "JSON:API paints my bikeshed!"
    },
    "links": {
      "self": "http://example.com/articles/1"
    },
    "relationships": {
      "comments": {
        "data": [
          { "type": "comments", "id": "5" },
          { "type": "comments", "id": "12" }
        ]
      }
    }
  },
  "included": [{
    "type": "comments",
    "id": "5",
    "attributes": {
      "body": "First!"
    },
    "links": {
      "self": "http://example.com/comments/5"
    }
  }, {
    "type": "comments",
    "id": "12",
    "attributes": {
      "body": "I like XML better"
    },
    "links": {
      "self": "http://example.com/comments/12"
    }
  }]
}

Ember Data uses the JSONAPISerializer by default, but you can override this default by defining a custom serializer. There are two ways to define a custom serializer. First, you can define a custom serializer for your entire application by defining an "application" serializer.

import JSONAPISerializer from '@ember-data/serializer/json-api';

export default class ApplicationSerializer extends JSONAPISerializer {
}

You can also define a serializer for a specific model. For example, if you had a post model you could also define a post serializer:

import JSONAPISerializer from '@ember-data/serializer/json-api';

export default class PostSerializer extends JSONAPISerializer {
}

To change the format of the data that is sent to the backend store, you can use the serialize() hook. Let's say that we have this JSON:API response from Ember Data:

{
  "data": {
    "id": "1",
    "type": "product",
    "attributes": {
      "name": "My Product",
      "amount": 100,
      "currency": "SEK"
    }
  }
}

But our server expects data in this format:

{
  "data": {
    "id": "1",
    "type": "product",
    "attributes": {
      "name": "My Product",
      "cost": {
        "amount": 100,
        "currency": "SEK"
      }
    }
  }
}

Here's how you can change the data:

import JSONAPISerializer from '@ember-data/serializer/json-api';

export default class ApplicationSerializer extends JSONAPISerializer {
  serialize(snapshot, options) {
    let json = super.serialize(...arguments);

    json.data.attributes.cost = {
      amount: json.data.attributes.amount,
      currency: json.data.attributes.currency
    };

    delete json.data.attributes.amount;
    delete json.data.attributes.currency;

    return json;
  }
}

Similarly, if your backend store provides data in a format other than JSON:API, you can use the normalizeResponse() hook. Using the same example as above, if the server provides data that looks like:

{
  "data": {
    "id": "1",
    "type": "product",
    "attributes": {
      "name": "My Product",
      "cost": {
        "amount": 100,
        "currency": "SEK"
      }
    }
  }
}

And so we need to change it to look like:

{
  "data": {
    "id": "1",
    "type": "product",
    "attributes": {
      "name": "My Product",
      "amount": 100,
      "currency": "SEK"
    }
  }
}

Here's how we could do it:

import JSONAPISerializer from '@ember-data/serializer/json-api';

export default class ApplicationSerializer extends JSONAPISerializer {
  normalizeResponse(store, primaryModelClass, payload, id, requestType) {
    payload.data.attributes.amount = payload.data.attributes.cost.amount;
    payload.data.attributes.currency = payload.data.attributes.cost.currency;

    delete payload.data.attributes.cost;

    return super.normalizeResponse(...arguments);
  }
}

To normalize only a single model, you can use the normalize() hook similarly.

For more hooks to customize the serializer with, see the Ember Data serializer API documentation.

In order to keep track of unique records in the store Ember Data expects every record to have an id property in the payload. Ids should be unique for every unique record of a specific type. If your backend uses a key other than id you can use the serializer's primaryKey property to correctly transform the id property to id when serializing and deserializing data.

import JSONAPISerializer from '@ember-data/serializer/json-api';

export default class ApplicationSerializer extends JSONAPISerializer {
  primaryKey = '_id';
}

In Ember Data the convention is to camelize attribute names on a model. For example:

import Model, { attr } from '@ember-data/model';

export default class PersonModel extends Model {
  @attr('string') givenName;
  @attr('string') familyName;
  @attr('boolean') isPersonOfTheYear;
}

However, the JSONAPISerializer expects attributes to be dasherized in the document payload returned by your server:

{
  "data": {
    "id": "44",
    "type": "people",
    "attributes": {
      "given-name": "Zaphod",
      "family-name": "Beeblebrox",
      "is-person-of-the-year": true
    }
  }
}

If the attributes returned by your server use a different convention you can use the serializer's keyForAttribute() method to convert an attribute name in your model to a key in your JSON payload. For example, if your backend returned attributes that are under_scored instead of dash-cased you could override the keyForAttribute method like this.

import { underscore } from '@ember/string';
import JSONAPISerializer from '@ember-data/serializer/json-api';

export default class ApplicationSerializer extends JSONAPISerializer {
  keyForAttribute(attr) {
    return underscore(attr);
  }
}

Irregular keys can be mapped with a custom serializer. The attrs object can be used to declare a simple mapping between property names on Model records and payload keys in the serialized JSON object representing the record. An object with the property key can also be used to designate the attribute's key on the response payload.

If the JSON for person has a key of familyNameOfPerson, and the desired attribute name is simply familyName, then create a custom Serializer for the model and override the attrs property.

import Model, { attr } from '@ember-data/model';

export default class PersonModel extends Model {
  @attr('string') familyName;
}

import JSONAPISerializer from '@ember-data/serializer/json-api';

export default class PersonSerializer extends JSONAPISerializer {
  attrs = {
    familyName: 'familyNameOfPerson'
  };
}

References to other records should be done by ID. For example, if you have a model with a hasMany relationship:

import Model, { hasMany } from '@ember-data/model';

export default class PostModel extends Model {
  @hasMany('comment', { async: true }) comments;
}

The JSON should encode the relationship as an array of IDs and types:

{
  "data": {
    "type": "posts",
    "id": "1",
    "relationships": {
      "comments": {
        "data": [
          { "type": "comments", "id": "1" },
          { "type": "comments", "id": "2" },
          { "type": "comments", "id": "3" }
        ]
      }
    }
  }
}

Comments for a post can be loaded by post.get('comments'). The JSON:API adapter will send 3 GET requests to /comments/1/, /comments/2/ and /comments/3/.

Any belongsTo relationships in the JSON representation should be the dasherized version of the property's name. For example, if you have a model:

import Model, { belongsTo } from '@ember-data/model';

export default class CommentModel extends Model {
  @belongsTo('post') originalPost
}

The JSON should encode the relationship as an ID to another record:

{
  "data": {
    "type": "comment",
    "id": "1",
    "relationships": {
      "original-post": {
        "data": { "type": "post", "id": "5" },
      }
    }
  }
}

If needed these naming conventions can be overwritten by implementing the keyForRelationship() method.

import JSONAPISerializer from '@ember-data/serializer/json-api';

export default class ApplicationSerializer extends JSONAPISerializer {
  keyForRelationship(key, relationship) {
    return key + 'Ids';
  }
}

In some circumstances, the built-in attribute types of string, number, boolean, and date may be inadequate. For example, a server may return a non-standard date format.

Ember Data can have new JSON transforms registered for use as attributes:

import Transform from '@ember-data/serializer/transform';
import EmberObject from '@ember/object';

export default class CoordinatePointTransform extends Transform {
  serialize(value) {
    return [value.get('x'), value.get('y')];
  }
  deserialize(value) {
    return EmberObject.create({ x: value[0], y: value[1] });
  }
}

import Model, { attr } from '@ember-data/model';

export default class Cursor extends Model {
  @attr('coordinate-point') position;
}

When coordinatePoint is received from the API, it is expected to be an array:

{
  cursor: {
    position: [4,9]
  }
}

But once loaded on a model instance, it will behave as an object:

let cursor = store.findRecord('cursor', 1);
cursor.get('position.x'); //=> 4
cursor.get('position.y'); //=> 9

If position is modified and saved, it will pass through the serialize function in the transform and again be presented as an array in JSON.

Not all APIs follow the conventions that the JSONAPISerializer uses with a data namespace and sideloaded relationship records. Some legacy APIs may return a simple JSON payload that is just the requested resource or an array of serialized records. The JSONSerializer is a serializer that ships with Ember Data that can be used alongside the RESTAdapter to serialize these simpler APIs.

To use it in your application you will need to define a serializer:application that extends the JSONSerializer.

import JSONSerializer from '@ember-data/serializer/json';

export default class ApplicationSerializer extends JSONSerializer {
  // ...
}

For requests that are only expected to return 1 record (e.g. store.findRecord('post', 1)) the JSONSerializer expects the response to be a JSON object that looks similar to this:

{
  "id": "1",
  "title": "Rails is omakase",
  "tag": "rails",
  "comments": ["1", "2"]
}

For requests that are only expected to return 0 or more records (e.g. store.findAll('post') or store.query('post', { filter: { status: 'draft' } })) the JSONSerializer expects the response to be a JSON array that looks similar to this:

[{
  "id": "1",
  "title": "Rails is omakase",
  "tag": "rails",
  "comments": ["1", "2"]
}, {
  "id": "2",
  "title": "I'm Running to Reform the W3C's Tag",
  "tag": "w3c",
  "comments": ["3"]
}]

The JSONAPISerializer is built on top of the JSONSerializer so they share many of the same hooks for customizing the behavior of the serialization process. Be sure to check out the API docs for a full list of methods and properties.

Although Ember Data encourages you to sideload your relationships, sometimes when working with legacy APIs you may discover you need to deal with JSON that contains relationships embedded inside other records. The EmbeddedRecordsMixin is meant to help with this problem.

To set up embedded records, include the mixin when extending a serializer then define and configure embedded relationships.

For example, if your post model contained an embedded author record that looks similar to this:

{
  "id": "1",
  "title": "Rails is omakase",
  "tag": "rails",
  "authors": [
    {
      "id": "2",
      "name": "Steve"
    }
  ]
}

You would define your relationship like this:

import JSONSerializer from '@ember-data/serializer/json';
import { EmbeddedRecordsMixin } from '@ember-data/serializer/rest';

export default class PostSerializer extends JSONSerializer.extend(EmbeddedRecordsMixin) {
  attrs = {
    authors: {
      serialize: 'records',
      deserialize: 'records'
    }
  };
}

If you find yourself needing to both serialize and deserialize the embedded relationship you can use the shorthand option of { embedded: 'always' }. The example above could therefore be expressed as such:

import JSONSerializer from '@ember-data/serializer/json';
import { EmbeddedRecordsMixin } from '@ember-data/serializer/rest';

export default class PostSerializer extends JSONSerializer.extend(EmbeddedRecordsMixin) {
  attrs = {
    authors: { embedded: 'always' }
  };
}

The serialize and deserialize keys support 3 values:

• records is used to signal that the entire record is expected
• ids is used to signal that only the id of the record is expected
• false is used to signal that the record is not expected

For example you may find that you want to read an embedded record when extracting a JSON payload but only include the relationship's id when serializing the record. This is possible by using the serialize: 'ids' option. You can also opt out of serializing a relationship by setting serialize: false.

import JSONSerializer from '@ember-data/serializer/json';
import { EmbeddedRecordsMixin } from '@ember-data/serializer/rest';

export default class PostSerializer extends JSONSerializer.extend(EmbeddedRecordsMixin) {
  attrs = {
    author: {
      serialize: false,
      deserialize: 'records'
    },
    comments: {
      deserialize: 'records',
      serialize: 'ids'
    }
  };
}

If you do not overwrite attrs for a specific relationship, the EmbeddedRecordsMixin will behave in the following way:

belongsTo: { serialize: 'id', deserialize: 'id' } hasMany { serialize: false, deserialize: 'ids' }

There is an option of not embedding JSON in the serialized payload by using serialize: 'ids'. If you do not want the relationship sent at all, you can use serialize: false.

If you would like to create a custom serializer, it is recommended that you start with the JSONAPISerializer or JSONSerializer and extend one of those to match your needs. However, if your payload is extremely different from one of these serializers you can create your own by extending the Serializer base class.

A serializer has two main roles in Ember Data. First, it is responsible for taking a response from an adapter and serializing it into the normalized JSON format that Ember Data understands. Secondly, it transforms snapshots of records into a payload the adapter will send to the server when creating, updating, or deleting a record.

The normalized JSON format that Ember Data expects is a JSON:API document with a couple of additional restrictions.

First, it is important to make sure that the type name of a record in the normalized JSON object exactly matches the filename of the model defined for this record type. By convention Model names are singular in Ember Data, however, the example type names shown in the JSON:API spec are pluralized. The JSON:API spec itself is agnostic about inflection rules, however, Ember Data's own JSONAPISerializer assumes types are plural and it will automatically singularize the types.

Second, attribute and relationship names in the JSON:API document should exactly match the name and casing of the @attr, @belongsTo and @hasMany, properties defined on the Model.

By convention these property names are camelCase in Ember Data models. As with the type names, this is different from the example attribute and relationship names shown in the JSON:API spec. The examples in the spec use dash-case for attribute and relationship names. However, the spec does not require attribute or relationship names to follow any specific casing convention. If you are using Ember Data's own JSONAPISerializer it will assume the attribute and relationship names from your API are dash-case and automatically transform them to camelCase when it creates the normalized JSON object.

Other than these two restrictions, Ember Data's normalized JSON object follows the JSON:API specification.

Example: given this post model.

import Model, { attr, hasMany } from '@ember-data/model';

export default class Post extends Model {
  @attr('string') title;
  @attr('string') tag;
  @hasMany('comment', { async: false }) comments;
  @hasMany('post') relatedPosts;
}

The normalized JSON object that Ember Data expects a serializer to return looks like this:

{
  data: {
    id: "1",
    type: "post",
    attributes: {
      title: "Rails is omakase",
      tag: "rails",
    },
    relationships: {
      comments: {
        data: [{ id: "1", type: "comment" },
               { id: "2", type: "comment" }],
      },
      relatedPosts: {
        links: {
          related: "/api/v1/posts/1/related-posts/"
        }
      }
    }
}

Note that the type is "post" to match the post model and the relatedPosts relationship in the document matches the relatedPosts: hasMany('post') on the model.

When creating a custom serializer you will need to define a normalizeResponse method to transform the response from the adapter into the normalized JSON object described above.

This method receives the store, the Model class for the request, the payload, the id of the record request (or null if there is no id associated with the request), and the request type (a string with the possible values of: 'findRecord', 'queryRecord', 'findAll', 'findBelongsTo', 'findHasMany', 'findMany', 'query', 'createRecord', 'deleteRecord', and 'updateRecord') as arguments.

A custom serializer will also need to define a normalize method. This method is called by store.normalize(type, payload) and is often used for normalizing requests made outside of Ember Data because they do not fall into the normal CRUD flow that the adapter provides.

Finally a serializer will need to implement a serialize method. Ember Data will provide a record snapshot and an options hash and this method should return an object that the adapter will send to the server when creating, updating or deleting a record.

If none of the built-in Ember Data Serializers work for your backend, be sure to check out some of the community maintained Ember Data Adapters and Serializers. A good place to search for them is Ember Observer.