Skip to content

Latest commit

 

History

History
968 lines (754 loc) · 21.7 KB

TRANSITION.md

File metadata and controls

968 lines (754 loc) · 21.7 KB
Raw

Ember Data 1.0 has a number of changes from Ember Data 0.13. The changes are pretty significant, so you should assume that you will need to dedicate some time to the upgrade.

Elimination of Model.find and Model.createRecord

This change future-proofs the API for modules and makes it less error-prone to clean up state between tests.

Ember Data 0.13:

App.PostsRoute = Ember.Route.extend({
  model: function() {
    return App.Post.find()
  }
});

App.PostRoute = Ember.Route.extend({
  model: function(params) {
    return App.Post.find(params.post_id)
  }
});

Ember Data 1.0.beta.1:

App.PostsRoute = Ember.Route.extend({
  model: function() {
    return this.store.find('post');
  }
});

App.PostRoute = Ember.Route.extend({
  model: function(params) {
    return this.store.find('post', params.post_id);
  }
});

Ember Data 0.13:

App.NewPostRoute = Ember.Route.extend({
  model: function() {
    return App.Post.createRecord();
  }
});

Ember Data 1.0.beta.1:

App.NewPostRoute = Ember.Route.extend({
  model: function() {
    return this.store.createRecord('post');
  }
});

Note that the store is automatically injected into all routes and controllers. If you were looking up records in views or components, you may wish to inject the store into those as well:

// inject the store into all components
App.inject('component', 'store', 'store:main');

In general, looking up models directly in a component is an anti-pattern, and you should prefer to pass in any model you need in the template that included the component.

{{my-popup person=user}}

All Finders Return Promises

In Ember Data 0.13, finders returned empty models or record arrays that would get filled in with the data once the server responded.

In Ember Data 1.0.beta.1, finders always return promises.

Ember Data 0.13:

var person = App.Post.find(1);
person.one('didLoad', function() {
  // do something with the post
});

Ember Data 1.0.beta.1:

this.store.find('post', 1).then(function(post) {
  // do something with post
});

If you need to wait for several find operations to return, you can use RSVP.all to join them together.

Ember Data 0.13:

var person1 = App.Person.find(1);
var person2 = App.Person.find(2);

var counter = 2;

person1.one('didLoad', bothLoaded);
person2.one('didLoad', bothLoaded);

function bothLoaded() {
  if (--counter === 0) {
    // work with person1 and person2
  }
}

Ember Data 1.0.beta.1:

var promise1 = this.store.find('person', 1);
var promise2 = this.store.find('person', 2);

Ember.RSVP.all([ promise1, promise2 ]).then(function(people) {
  // people is an array of Person records
  // work with people here
});

Note that the promises returned from finders have special support for Ember data-binding. They can be used in templates and templates will automatically update once the promise has resolved.

Transaction is Gone: Save Individual Records

You no longer need to push records into a transaction in order to save them.

Ember Data 0.13:

App.NewPostRoute = Ember.Route.extend({
  model: function() {
    return App.Post.createRecord();
  },

  setupController: function(controller, model) {
    var transaction = this.store.transaction()
    controller.set('transaction', this.store.transaction());
    transaction.add(model);
  },

  actions: {
    save: function() {
      this.controllerFor('newPost').get('transaction').commit();
    }
  }
});

Ember Data 1.0.beta.1:

App.NewPostRoute = Ember.Route.extend({
  model: function() {
    return this.store.createRecord('post');
  }

  actions: {
    save: function() {
      this.modelFor('newPost').save();
    }
  }
});

You also no longer need transactions to save relationships independently of the model it belongs to, e.g. comments of a post.

Ember Data 0.13:

App.PostController = Ember.ObjectController.extend({
  saveComment: function(){
    var transaction = this.get('store').transaction();
    var  comment = {
      userId: this.get('userId'),
      post: this.get('model'),
      comment: this.get('comment'),
      created_at: Date.create().format(Date.ISO8601_DATETIME)
    };
    if (transaction) {
      this.set('transaction', transaction);
      transaction.createRecord(App.Comment, comment);
    } else {
      console.log('store.transaction() returned null');
    }

    this.get('transaction').commit();
    this.set('comment', '');
  }
});

Ember Data 1.0.beta.1:

App.PostController = Ember.ObjectController.extend({
  saveComment: function(){
    var  comment = {
      userId: this.get('userId'),
      post: this.get('model'),
      comment: this.get('comment'),
      createdAt: Date.create().format(Date.ISO8601_DATETIME)
    };
    var comment = this.store.createRecord('comment', comment);
    comment.save();
    this.set('comment', '');
  }
});

If you want to batch up a bunch of records to save and save them all at once, you can just put them in an Array and call .invoke('save') when you're ready.

We plan to support batch saving with a single HTTP request through a dedicated API in the future.

Save Returns a Promise

Calling save on a record returns a promise that will be resolved when the server returns successfully or rejected if the server returns with an error code.

Ember Data 0.13:

person.save();

person.one('didCommit', function() {
  // work with saved person
  // some bugs existed regarding newly created records and ID timing
});

person.one('didError', function() {
  // work with errored person
});

Ember Data 1.0.beta.1:

person.save().then(function() {
  // work with saved person
  // newly created records are guaranteed to have IDs assigned
}, function() {
  // work with person that failed to save
});

This also means that retrying should be simple in Ember Data 1.0.beta.1:

person.save().then(null, function() {
  return person.save();
}).then(function() {
  // person was retried once
});

You can of course implement whatever retry strategy you want (incremental backoff, try N times, etc.) on top of the basic promise primitive.

function retry(promise, retryCallback, nTimes) {
  // if the promise fails,
  return promise.then(null, function(reason) {
    // if we haven't hit the retry limit
    if (nTimes-- > 0) {
      // retry again with the result of calling the retry callback
      // and the new retry limit
      return retry(retryCallback(), retryCallback, nTimes);
    }

    // otherwise, if we hit the retry limit, rethrow the error
    throw reason;
  });
}

// try to save the person up to 5 times 
retry(person.save(), function() {
  return person.save();
}, 5);

Because all async operations in Ember Data 1.0.beta.1 are promises, you can combine them together using normal promise primitives.

this.store.find('person').then(function(people) {
  people.forEach(function(person) {
    person.set('isPaidUp', true);
  });

  return Ember.RSVP.all(people.invoke('save'));
}).then(function(people) {
  // people we successfully saved
});

In Ember 0.13, the semantics for how to determine that a find had finished loading were unclear and confusing.

Adapters

The adapter API has undergone a significant change.

Existing adapters will likely need to be rebuilt. The good news is that the new adapter API is somewhat simpler.

To register your own adapter in place of the default RESTAdapter

Ember Data 0.13:

App.Store = DS.Store.extend({
  adapter: DS.MyRESTAdapter.create()
});

Ember Data 1.0.beta.1:

App.ApplicationAdapter = DS.MyRESTAdapter;

Promises

Adapter hooks no longer call directly into the store to notify the store that the backend has returned. Instead, all adapter hooks return a promise that the store will automatically handle as appropriate for the type of query that kicked off the request.

Ember Data 0.13:

App.MyAdapter = DS.Adapter.extend({
  find: function(store, type, id) {
    $.getJSON("/" + this.pluralize(type) + "/" + id, function(payload) {
      store.push(payload);
    }
  }
});

Ember Data 1.0.beta.1:

App.MyAdapter = DS.Adapter.extend({
  find: function(store, type, id) {
    return $.getJSON("/" + this.pluralize(type) + "/" + id);
  }
});

The short version is that adapter hooks don't need to know how to update the store, records, or record arrays once the server has returned a payload. They just need to return a promise, and Ember Data will take care of interpreting it correctly.

Per Type Adapters

Ember Data 0.13:

App.Post = DS.Model.extend({
  // ...
});

App.PostAdapter = DS.RESTAdapter.extend({
  // ...
});

App.Store = DS.Store.extend();

App.Store.registerAdapter(App.Post, App.PostAdapter);

Ember Data 1.0.beta.1:

App.Post = DS.Model.extend({
  // ...
});

App.PostAdapter = DS.RESTAdapter.extend({
  // ...
});

// No store is needed
// Adapters are wired up by name

Per Type Serializers

Ember Data 0.13:

App.Post = DS.Model.extend({
  // ...
});

App.PostSerializer = DS.RESTSerializer.extend({
  // ...
});

App.PostAdapter = DS.RESTAdapter.extend({
  serializer: 'App.PostSerializer',
});

App.Store = DS.Store.extend();

App.Store.registerAdapter(App.Post, App.PostAdapter);

Ember Data 1.0.beta.1:

App.Post = DS.Model.extend({
  // ...
});

App.PostSerializer = DS.RESTSerializer.extend({
  // ...
});

// That's it. Everything is wired up by naming

REST Adapter and Serializer Configuration

There are significant improvements and simplifications to configuring the REST Adapter and Serializer.

The short version is that once an Ajax request has completed, the resulting payload is sent through the following hooks:

  1. The payload is sent to extractSingle if the original request was for a single record (like find/save) or extractArray if the original request was for an Array of records (like findAll/findQuery)
  2. The default behavior of those methods is to pull apart the top-level of the payload into multiple smaller records.
  3. Each of those smaller records is sent to normalize, which can do normalization a record at a time.
  4. Finally, specific types of records can be specially normalized.

Here's how it works.

Imagine you start with a payload like this for request for Post 1.

{
  "id": "1",
  "title": "Rails is omakase",
  "_links": [{
    "user": {
      "href": "/people/dhh"
    }
  }],
  "_embedded": {
    "comments": [{
      "ID_": "1",
      "CMT_BODY": "Rails is unagi"
    }, {
      "ID_": "2",
      "CMT_BODY": "Omakase O_o"
    }]
  }
}

Ember Data 1.0.beta.1 is expecting a payload like this:

{
  "post": {
    "id": 1
    "title": "Rails is omakase",
    "comments": ["1", "2"],
    "_links": {
      "user": "/people/dhh"
    },
  },

  "comments": [{
    "id": "1",
    "body": "Rails is unagi"
  }, {
    "id": "2",
    "body": "Omakase O_o"
  }]
}

The simplest way to approach this would be to override the extractSingle method and do all the munging in one place. This would totally work.

App.PostSerializer = DS.RESTSerializer.extend({
  extractSingle: function(store, type, payload, id, requestType) {
    var post = {}, commentIds = [];

    post.id = payload.id;
    post.title = payload.title;
    post._links = { user: payload._links.mapProperty('user').findProperty('href').href };

    // Leave the original un-normalized comments alone, but put them
    // in the right place in the payload. We'll normalize the comments
    // below in `normalizeHash`
    var comments = payload._embedded.comments.map(function(comment) {
      commentIds.push(comment.ID_);
      return { id: comment.ID_, body: comment.CMT_BODY };
    });

    post.comments = commentIds;
    
    var post_payload = { post: post, comments: comments };

    return this._super(store, type, post_payload, id, requestType);
  }
});

However, Ember Data also allows you to break things into into more sensible chunks. This is especially useful if you have a lot of custom normalization to do on different pieces of the JSON.

App.PostSerializer = DS.RESTSerializer.extend({
  extractSingle: function(store, type, payload, id, requestType) {
    var post = {}, commentIds = [];

    post.id = payload.id;
    post.title = payload.title;
    post._links = { user: payload._links.mapProperty('user').findProperty('href').href };

    // Leave the original un-normalized comments alone, but put them
    // in the right place in the payload. We'll normalize the comments
    // below in `normalizeHash`
    var comments = payload._embedded.comments;
    post.comments = comments.mapProperty('ID_');
    
    var post_payload = { post: post, comments: comments };

    return this._super(store, type, post_payload, id, requestType);
  },

  normalizeHash: {
    comments: function(hash) {
      return { id: hash.ID_, body: hash.CMT_BODY };
    }
  }
});

You can also implement the normalize method to do generic things like camelize all of the keys in a hash. The normalize method gets called once for every sub-hash.

Imagine that we have a payload like this, which is totally fine except for the use of underscores rather than camelization and "_id" instead of "id":

{
  "post": {
    "_id": "1",
    "author_name": "DHH",
    "title": "Rails is omakase",
    "comments": [ "1", "2" ]  
  },
  "comments": [{
    "_id": "1",
    "author_name": "unagisan",
    "like_count": 12,
    "body": "Unagi!"
  }, {
    "_id": "2",
    "author_name": "tlo",
    "like_count": 100,
    "body": "Omakase is delicious"
  }]
}

We don't need to implement extractSingle, because the top-level is already organized perfectly.

App.PostSerializer = DS.RESTSerializer.extend({
  // This method will be called 3 times: once for the post, and once
  // for each of the comments
  normalize: function(type, hash, property) {
    // property will be "post" for the post and "comments" for the
    // comments (the name in the payload)

    // normalize the `_id`
    var json = { id: hash._id };
    delete hash._id;

    // normalize the underscored properties
    for (var prop in hash) {
      json[prop.camelize()] = hash[prop]; 
    }

    // delegate to any type-specific normalizations
    return this._super(type, json, property);
  }
});

So to sum up, you should:

  • use extractSingle and extractArray when the top-level of your payload is organized differently than Ember Data expects
  • use normalize to normalize sub-hashes if all sub-hashes in the payload can be normalized in the same way.
  • use normalizeHash to normalize specific sub-hashes.
  • make sure to call super if you override extractSingle, extractArray or normalize so the rest of the chain will get called.
  • beta.1 expects comments key now instead of comments_ids. This is likely to be configurable in beta.2.

Underscored Keys, _id and _ids

In 0.13, the REST Adapter automatically camelized incoming keys for you. It also expected belongsTo relationships to be listed under name_id and hasMany relationships to be listed under name_ids.

If your application returns json with underscored attributes and _id or _ids for relation, you can extend ActiveModelSerializer and all will work out of the box.

App.ApplicationSerializer = DS.ActiveModelSerializer.extend({});

Note: DS.ActiveModelSerializer is not to be confused with the ActiveModelSerializer gem that is part of Rails API project. A conventional Rails API project with produce underscored output and the DS.ActiveModelSerializer will perform the expected normalization behavior such as camelizing property keys in your JSON.

Underscored API Endpoints

In 0.13 the REST Adapter automatically generated API endpoints for multi word models with an underscore ('/blog_posts'), begining with Beta 1 the REST Adapter will generate endpoints camelized. To go back to underscored endpoints you can define the pathForType method in your ApplicationAdapter.

App.ApplicationAdapter = DS.RESTAdapter.extend({
  pathForType: function(type) {
    var underscored = Ember.String.underscore(type)
    return Ember.String.pluralize(underscored);
  }
});

Underscored Root Objects in JSON Responses

In 0.13 the REST Adapter expected the root objects of a JSON response to be underscored for multi word models.

{
  "blog_posts": [{
    "id": 1,
    "title": "A Post"
  }, {
    "id": 2,
    "title": "Another Post"
  }]
}

Beginning in Beta 1 the REST Adapter expects the root objects of a JSON response to be camelized.

{
  "blogPosts": [{
    "id": 1,
    "title": "A Post"
  }, {
    "id": 2,
    "title": "Another Post"
  }]
}

If your server uses underscored root objects you can define the typeForRoot method in your ApplicationSerializer.

App.ApplicationSerializer = DS.RESTSerializer.extend({
  typeForRoot: function(root) {
    var camelized = Ember.String.camelize(root);
    return Ember.String.singularize(camelized);
  }
});

Underscored Root Objects in JSON Requests

In 0.13 the REST Adapter would send underscored versions of multi word model names as the root element. Beginning in Beta 1 the REST Adapter sends camelized root object.

If your server expects underscored root objects you can define the serializeIntoHash method in your ApplicationSerializer.

App.ApplicationSerializer = DS.RESTSerializer.extend({
  serializeIntoHash: function(data, type, record, options) {
    var root = Ember.String.decamelize(type.typeKey);
    data[root] = this.serialize(record, options);
  }
});

Underscored JSON for Saving

In 0.13 the REST Adapter would send underscored JSON for create/update requests. Beginning in Beta 1 the REST Adapter sends camelized JSON.

If your server expects underscored JSON you can define the serializeAttribute method in your ApplicationSerializer.

App.ApplicationSerializer = DS.RESTSerializer.extend({
  serializeAttribute: function(record, json, key, attribute) {
    var attrs = Ember.get(this, 'attrs');
    var value = Ember.get(record, key), type = attribute.type;

    if (type) {
      var transform = this.transformFor(type);
      value = transform.serialize(value);
    }

    // if provided, use the mapping provided by `attrs` in
    // the serializer
    key = attrs && attrs[key] || Ember.String.decamelize(key);

    json[key] = value;
  }
}

Embedded Records

Explicit support for embedded records is gone for now.

You can handle embedded records yourself by implementing extractSingle and reorganizing the payload.

Consider this payload:

{
  "post": {
    "id": "1",
    "title": "Rails is omakase",
    "comments": [{
      "id": "1",
      "body": "I like omakase"
    }, {
      "id": "2",
      "body": "I prefer not to rely on elitist chefs"
    }]
  }
}

You could handle embedded records like this:

App.PostSerializer = DS.RESTSerializer.extend({
  extractSingle: function(store, type, payload, id, requestType) {
    var comments = payload.post.comments,
        commentIds = comments.mapProperty('id');

    payload.comments = comments;
    payload.post.comments = commentIds;

    return this._super.apply(this, arguments);
  }
});

JSON Transforms

Ember Data 0.13:

DS.RESTAdapter.registerTransform('raw', {
  deserialize: function(serialized) {
    return serialized;
  },
  serialize: function(deserialized) {
    return deserialized;
  }
});

Ember Data 1.0.beta.1:

App.RawTransform = DS.Transform.extend({
  deserialize: function(serialized) {
    return serialized;
  },
  serialize: function(deserialized) {
    return deserialized;
  }
});

Host and Namespace Configuration

Ember Data 0.13:

DS.RESTAdapter.reopen({
  url: 'http://www.google.com',
  namespace: 'api/v1'
});

Ember Data 1.0.beta.1:

App.ApplicationAdapter = DS.RESTAdapter.extend({
  host: 'http://www.google.com',
  namespace: 'api/v1'
});

Relationships

Defining Relationships in Models

Defining relationships now uses the 'shorthand' name of your model, not the fully qualified path to the class.

Instead of App.Comment, use comment.

Instead of App.BlogPost, use blogPost.

Ember Data 0.13:

App.BlogPost = DS.Model.extend({
  comments: DS.hasMany("App.Comment")
});

App.Comment = DS.Model.extend({
  post: DS.belongsTo("App.BlogPost")
});

Ember Data 1.0.beta.1:

App.BlogPost = DS.Model.extend({
  comments: DS.hasMany("comment")
});

App.Comment = DS.Model.extend({
  post: DS.belongsTo("blogPost")
});

Polymorphic relationships

Polymorphic types are now serialized with a json key of the model name + "Type"

For example given the polymorphic relationship:

App.Comment = DS.Model.extend({
  message: DS.belongsTo('message', {
    polymorphic: true
  })
});

Ember Data 0.13

{
  "message": 12,
  "message_type": "post"
}

Ember Data 1.0.beta.3:

{
  "message": 12,
  "messageType": "post"
}

You can override this behaviour by defining the serializePolymorphicType method on your serializer.

App.CommentSerializer = DS.RESTSerializer.extend({
  serializePolymorphicType: function(record, json, relationship) {
    var key = relationship.key,
        belongsTo = get(record, key);
    key = this.keyForAttribute ? this.keyForAttribute(key) : key;
    json[key + "_type"] = belongsTo.constructor.typeKey;
  }
});

Defining a Custom Serializer

If you have a custom adapter you will likely need to wire up a custom serializer.

Ember Data 0.13:

DS.DjangoRESTSerializer = DS.RESTSerializer.extend();
DS.DjangoRESTAdapter = DS.RESTAdapter.extend({
  serializer: DS.DjangoRESTSerializer
});

Ember Data 1.0.beta.1:

DS.DjangoRESTSerializer = DS.RESTSerializer.extend();
DS.DjangoRESTAdapter = DS.RESTAdapter.extend({
  defaultSerializer: "DS/djangoREST"
});

Your users will still be able to define custom serializers that extend your default serializer. If they do, they should make sure to subclass your custom serializer and not a built-in Ember Data serializer.