I've been trying my hand at Backbone recently and have a very basic question.
I need to search for different kind of records and the search API returns a JSON response like
{ foo:
[
{ name: 'foo1', url: '/foo1' },
{ name: 'foo2', url: '/foo2' }
],
bar:
[ { name: 'bar1', url: '/bar1' } ],
baz:
[ { name: 'baz1', url: '/baz1' } ]
}
I have a backbone model for Foo, Bar and Baz. A collection which on fetch should hit my server and get me the search results. I attempted something like
window.searchEntities = Backbone.Collection.extend({
url: '/get_search_results'
model: function(attrs, options) {
//Typecast the JSON to Foo, Bar and Baz Models
});
});
However, I do not know how to parse the results returned by server so that my collection holds models Foo, Bar and Baz? Or should I tweak the results returned by server so that its easier to handle this with Backbone?
As I'm seeing your JSON is not returning 3 different Models but 3 different Collection due the 3 of them are containing Arrays.
I think you should start from the beginning, If I understood weel you want to return a bunch of Models of different type, let's say:
[
{
type: "Foo",
name: "foo1",
url: "/foo1"
},
{
type: "Foo",
name: "foo2",
url: "/foo2"
},
{
type: "Bar",
name: "bar1",
url: "/bar1"
},
{
type: "Baz",
name: "baz1",
url: "/baz1"
},
]
I see a Collection there, and also different Models of different types.
Now let's see the SearchCollection, I don't think you can play with the model property like you show in your example, so let's say all the Models has a common Model parent Result:
window.SearchEntities = Backbone.Collection.extend({
url: '/get_search_results'
model: Result
});
From here we can do it simple and don't create sub-classes of Result if there is not a real need for it:
window.Result = Backbone.Model.extend({
initialize: function(){
this.url = this.get( "url" );
}
});
And you're done:
var searchEntities = new window.SearchEntities();
searchEntities.fetch();
// ... wait for the fetch ends
searchEntities.at(0).get( "type" ); //=> "Foo"
Still I don't feel confortable by two reasons:
I don't see clear why you want to play with the Result.url.
Where are the ids of your Models? which are very important for Backbone.
Related
Using ExtJS 6 one can have a store bind to the model and use the methods sync to save or load to load data.
I imagine that if a data is removed from store, upon calling sync the data will be removed from database too.
In my use case, I have different URLs and mandatory Ajax query fields for each action of create/update, load and delete data.
I have only seen examples showing load or save to storage, how can I declare the load, save and delete using Ajax in the same model?
Another doubt I have is that stores themselves can have a proxy, so they can perform those operations too, at least the load operation that I have seen in use. What's the difference between having these on the model or store? What's the best practice?
Example model from Sencha docs (is this only for read?):
Ext.define('MyApp.model.Base', {
extend: 'Ext.data.Model',
fields: [{
name: 'id',
type: 'int'
}],
schema: {
namespace: 'MyApp.model', // generate auto entityName
proxy: { // Ext.util.ObjectTemplate
type: 'ajax',
url: '{entityName}.json',
reader: {
type: 'json',
rootProperty: '{entityName:lowercase}'
}
}
}
});
Another example I found on https://examples.sencha.com/extjs/6.0.1/examples/classic/writer/writer.html using the proxy config, this seems more like what I would need as it specifies a URL for each operation:
var store = Ext.create('Ext.data.Store', {
model: 'Writer.Person',
autoLoad: true,
autoSync: true,
proxy: {
type: 'ajax',
api: {
read: 'app.php/users/view',
create: 'app.php/users/create',
update: 'app.php/users/update',
destroy: 'app.php/users/destroy'
},
reader: {
type: 'json',
successProperty: 'success',
root: 'data',
messageProperty: 'message'
},
writer: {
type: 'json',
writeAllFields: false,
root: 'data'
},
listeners: {
exception: function(proxy, response, operation){
Ext.MessageBox.show({
title: 'REMOTE EXCEPTION',
msg: operation.getError(),
icon: Ext.MessageBox.ERROR,
buttons: Ext.Msg.OK
});
}
}
},
listeners: {
write: function(proxy, operation){
if (operation.action == 'destroy') {
main.child('#form').setActiveRecord(null);
}
Ext.example.msg(operation.action, operation.getResultSet().message);
}
}
});
I believe I can have something like this in my case (this is just an example not tested!):
Ext.define('My.Person.Model', {
proxy: {
type: 'ajax',
api: {
read: 'http://myapiserver/getuser',
create: 'http://myapiserver/upsertuser',
update: 'http://myapiserver/upsertuser',
destroy: 'http://myapiserver/removeuser'
},
reader: {
type: 'json',
successProperty: 'success',
root: 'data',
messageProperty: 'message'
},
writer: {
type: 'json',
writeAllFields: false,
root: 'data'
},
// How can I have the parameters for each one?
extraParams : {
isuserUnderage : ' '
, query : '%'
}
}
});
I have no idea how to do this, specially specifying parameters for each type of Ajax request (read, create, update, destroy), I can have an upsert request that will send all fields, but the remove request will require only the ID, the get request can have optional fields for filtering, like filtering persons by name.
Example to be more clear of the problem.
Example data:
[
{
"id": "1",
"name": "Fred",
"age": 21,
"sex": "m"
},
{
"id": "2",
"name": "Susan",
"age": 12,
"sex": "f"
},
{
"id": "3",
"name": "Marcus",
"age": 22,
"sex": "m"
},
{
"id": "4",
"name": "Alex",
"age": 32,
"sex": "m"
}
]
Endpoints example:
Endpoints have parameters, these are mandatory, this means that calling an enpoint without a parameter will cause a server error, also passing a parameter that is not specified will cause a server error! If a parameter is not necessary one can pass a string with a single whitespace .
To read:
Endpoint: http://myapiserver/getuser?query={query}
Name is a filter by name, for example http://myapiserver/getuser?query=fred will bring users with name that has the string fred.
To write, we usually have an upsert, so it works for both insert and update:
Endpoint: http://myapiserver/upsertuser?id={id}&name={name}&age={age}&sex={sex}
So to update we can pass the ID: http://myapiserver/upsertuser?id=1&name=Frederick&age=21&sex=m and to insert we pass an empty string for ID: http://myapiserver/upsertuser?id= &name=Maurice&age=41&sex=m
To remove:
Endpoint: http://myapiserver/removeuser?id={id}
Example: http://myapiserver/removeuser?id=1, removes person with ID 1.
Because you say it's mandatory to use GETs with query params, I would encourage you to rethink your tech stack because the RESTful verbs really make it more clear what your action is, and you remove the actual action from your URL routes. However, I know sometimes this is totally out of our control, so I'll try my best here... I have to say, I've never experienced something like this, so I don't know if what I'm showing here is a best practice.
I can't show a true implementation because Sencha Fiddle is a simple sandbox, not meant for actual server-side implementations. I'm also assuming that you're using the classic toolkit, but if you need it in modern, it's a fairly easy port that you can do.
I prefer the proxy inside of the model for several reasons... if I need to use this model in several different stores throughout my app, then each store will inherit the same proxy. If I want to use the same model, but I don't want its proxy, I can simply override it when defining the store. Also, if the proxy doesn't exist on the model, then the framework assumes what your URL should be, which doesn't work when I want to use models individually.
I think I've come up with what you're asking for in this Fiddle. Really the core of what you want is in GETUser.js.
// We need to create our own proxy that will handle this for us
Ext.define('AjaxGet', {
extend: 'Ext.data.proxy.Ajax',
alias: 'proxy.ajaxGet',
// Per your requirement, we want to send individual requests
batchActions: false,
createOperation: function (action, config) {
// This means we're doing an action against one of our records
if (config && config.records) {
if (action === 'destroy') {
config.params = config.records[0].getDeleteParams();
} else if (action === 'create' || action === 'update') {
config.params = config.records[0].getUpsertParams();
}
}
return this.callParent(arguments);
}
});
// This is the desired, "GET" User model that uses GETs and query params for all actions
Ext.define('GETUser', {
extend: 'Ext.data.Model',
idProperty: 'Id',
fields: [{
name: 'Name',
type: 'string'
}, {
name: 'Id',
type: 'int'
}, {
name: 'Age',
type: 'int'
}, {
name: 'Sex',
type: 'string'
}],
proxy: {
type: 'ajaxGet',
api: {
read: 'Users',
create: 'upsertuser',
update: 'upsertuser',
destroy: 'removeuser'
},
actionMethods: {
create: 'GET',
update: 'GET',
destroy: 'GET'
}
},
getUpsertParams: function () {
const data = this.getData();
// Means this record hasn't been saved, so we're in the CREATE state
if (this.phantom) {
// We don't want to send the ID with what the framework sets as the ID
data.Id = undefined;
}
return data;
},
getDeleteParams: function () {
return {
Id: this.get('Id')
};
}
});
So what I ended up doing was creating a custom proxy that overrides the createOperation method to check which operation we're doing... based on that operation, we use the methods in the model to retrieve the params we want to send to the API. You need actionMethods in the proxy because otherwise, they default to POSTs.
I've been over the docs quite a few times, but this aspect still isn't clear to me. It's entirely possible that I'm thinking backbone-relational does something that it doesn't.
I'm looking for the way to define relationships based on key to avoid all the boilerplate fetching nonsense.
Take the canonical Artists and Albums example:
An artist has many albums as defined by album.artist_id
/api/artist/62351 might return
{
id: 62351,
name:"Jimi Hendrix"
}
similarly /api/album?artist_id=62351 might return
[
{
id:5678,
name: "Are You Experienced?"
artist_id: 62351
},
{
id: 4321,
name: "Axis: Bold as love",
artist_id: 62351
}
]
How might I define Artist and Album relationships such that
var hendrixInstance = new Artist({id:62351});
hendrixInstance.get('albums');
would fetch and return a collection of albums based on the album foreign_key artist_id? It must just be some permutation of key/keySource/keyDestination that I've yet to try, or be a problem that backbone-relational isn't trying to solve, but my doc groking has failed and I think a concise answer to this on SO might help future Googlers.
var Artist = Backbone.RelationalModel.extend({
urlRoot: '/api/artist',
relations:[{
key: 'albums', //Docs say this is the foreign key name, but in practice it doesn't appear that way. Need keySource/Destination?
type: Backbone.HasMany,
reverseRelation: {
key: 'artist',
type: Backbone.HasOne
}
}]
});
var Album = Backbone.RelationalModel.extend({
urlRoot: '/api/album'
});
Bonus points to an example model that references its self adjacency list style with parent_id
So, #xzhang's method above kept me iterating on this problem. First off, I'd love to be proven wrong on this, but I haven't found a way that backbone-relational handles this problem without additional custom code. Since this in my mind is an incredibly basic example of a OneToMany relationship, I'm still holding out hope that I'm just not getting something obvious.
Here's what I ended up doing to handle the situation. Unfortunately it still does not automatically fetch from the server when someobject.fetch('somerelationship') is called, which is what I really want. The parse function won't be necessary for most people, but it's required for the api I'm calling.
First I set up a base collection from which to extend:
var BaseCollection = Backbone.Collection.extend({
initialize: function(models, options) {
if (_.isObject(options.relation)) {
this.url = '/api/'
+ options.relation.keySource
+ '?search.'+options.relation.reverseRelation.keySource
+ '=' + options.foreignId;
}
},
parse: function(res) { return res.success ? res.list : res },
});
Then a reusable helper function (could probably be rolled into BaseCollection) to assist with creating relationships
function collectionOptions(instance) {
return {"relation":this, "foreignId":instance.get('id') };
};
And finally, those relationships are told to use BaseCollection as their CollectionType, and the collectionOptions() helper is assigned to set collectionOptions.
var Form = BaseModel.extend({
urlRoot: '/api/form',
relations:[
{
key: 'fills',
keySource: 'fill',
relatedModel: Fill,
type: Backbone.HasMany,
collectionOptions: collectionOptions,
collectionType: BaseCollection,
reverseRelation: {
key: 'form',
keySource: 'form_id',
type: Backbone.HasOne
}
},{
key: 'children',
keySource: 'form',
relatedModel: 'Form',
type: Backbone.HasMany,
collectionOptions: collectionOptions,
collectionType: BaseCollection,
reverseRelation: {
key: 'parent',
keySource: 'parent_id',
type: Backbone.HasOne
}
}
]
});
This allows me to avoid changing the server side API to return a list of ids and then individually fetch those ids. Instead I can just:
var form = new Form({id:1});
form.get('children').fetch();
form.toJSON(); //now has {id:1, ..., ..., children:[child,child,child,...]}
An extension to autoFetch children on the first call to .get('children') would be just the ticket, but I haven't discovered how to do that without modifying backbone-relational itself.
I am facing the exactly problem (Backbone-relational hasmany best practices), after 2 days research and look into the source code, I don't think key/keySource/keyDestination will do the work (correct me if I am wrong).
So I end up with create my own relation type, so far works fine. This may not a good solution, but hope can help you.
var LazyMany = Backbone.HasMany.extend({
setRelated: function (related) {
var relation = this.options
, instance = this.instance
;
if (related && !_.result(related, 'url')) {
related.url = relation.relatedModel.prototype.urlRoot +
'?' + relation.reverseRelation.key + '=' + instance.id;
}
return LazyMany.__super__.setRelated.apply(this, arguments);
}
});
Then in your model:
var Album = Backbone.RelationalModel.extend({
urlRoot: '/api/album/'
});
var Artist = Backbone.RelationalModel.extend({
urlRoot: '/api/artist/',
relations:[{
key: 'albums',
type: LazyMany,
includeInJSON: false,
relatedModel: Album,
reverseRelation: {
key: 'artist',
// I didn't test this, I don't have artist_id, artist is "id" in my app
keySource: 'artist_id',
keyDestination: 'artist_id',
includeInJSON: 'id'
}
}]
});
So if you don't define a collectionType or your collection don't have a url field, LazyMany will create a collection with url: /api/album/?artist=62351.
Then you just need fetch the collection: artist.get('albums').fetch().
Hope this can help, and I am still looking for better solutions.
I have two ember models with a relationship like this
App.Foo = DS.Model.extend
bar: DS.belongsTo("App.Bar", embedded: true)
App.Bar = DS.Model.extend
primaryKey: "blah"
blah: DS.attr "string
If I create and save a new record like this:
foo = App.store.createRecord App.Foo
foo.set "bar", App.Bar.createRecord(blah: "blahblah")
App.store.commit()
I see 2 post requests to the server:
URL: /foos
Payload: {"foo":{"bar":null}}
and
URL: /bars
Payload: {"bar":{"blah":"blahblah"}}
The association is embedded so I would like to see:
URL: /foos
Payload: {"foo":{"bar":{"blah":"blahblah"}}}
Can I achieve this with the ember-data REST adapter or do I need to write my own code to do this?
I am observing the same behavior in my application.
Setting 'embedded' to true only helps you get data as embedded, but while you post it separate requests will be generated.
You have write your in code if you want to achieve it in one request.
This is not a direct answer to your question, but for what it's worth I've found that "fighting" Ember's design by trying to save multiple models at once led me down a terrible path.
It is much easier and more reliable to add the required attributes for a new Bar to Foo and then create and return the new objects on the server, like so:
App.Foo = DS.Model.extend({
blah: DS.attr('string'),
barName: DS.attr('string'),
bar: DS.belongsTo('bar')
})
this.store.createRecord('foo', { blah: "stuff", barName: "foo" }).save()
# => POST /foos
Request:
{ foo: { blah: "stuff", bar_name: "boo" } }
Response:
{ foo: { id: 1, blah: "stuff", bar: { id: 1, name: "boo" } } }
Yes, you end up with an unused attribute on Foo, but you save yourself a lot of code.
Let's suppose I have two simple fixture files, one for the user(1) and one for the messages(2).
The Backbone Model for the messages is the following (3).
If I load the "Message Fixture", I would like to have also the related info regarding the user as specified in Message Model.
What is the proper way to active this goal in a spec view (4) by using jasmine test suite?
Please see the comments in (4) for more details.
(1)
// User Fixture
beforeEach(function () {
this.fixtures = _.extend(this.fixtures || {}, {
Users: {
valid: {
status: 'OK',
version: '1.0',
response: {
users: [
{
id: 1,
name: 'olivier'
},
{
id: 2,
name: 'pierre',
},
{
id: 3,
name: 'george'
}
]
}
}
}
});
});
(2)
// Message Fixture
beforeEach(function () {
this.fixtures = _.extend(this.fixtures || {}, {
Messages: {
valid: {
status: 'OK',
version: '1.0',
response: {
messages: [
{
sender_id: 1,
recipient_id: 2,
id: 1,
message: "Est inventore aliquam ipsa"
},
{
sender_id: 3,
recipient_id: 2,
id: 2,
message: "Et omnis quo perspiciatis qui"
}
]
}
}
}
});
});
(3)
// Message model
MessageModel = Backbone.RelationalModel.extend({
relations: [
{
type: Backbone.HasOne,
key: 'recipient_user',
keySource: 'recipient_id',
keyDestination: 'recipient_user',
relatedModel: UserModel
},
{
type: Backbone.HasOne,
key: 'sender_user',
keySource: 'sender_id',
keyDestination: 'sender_user',
relatedModel: UserModel
}
]
});
(4)
// Spec View
describe('MyView Spec', function () {
describe('when fetching model from server', function () {
beforeEach(function () {
this.fixture = this.fixtures.Messages.valid;
this.fixtureResponse = this.fixture.response.messages[0];
this.server = sinon.fakeServer.create();
this.server.respondWith(
'GET',
// some url
JSON.stringify(this.fixtureResponse)
);
});
it('should the recipient_user be defined', function () {
this.model.fetch();
this.server.respond();
// this.fixtureResponse.recipient_user is not defined
// as expected by the relation defined in (3)
expect(this.fixtureResponse.recipient_user).toBeDefined();
});
});
});
});
Take a look at this series of tutorials http://tinnedfruit.com/2011/03/03/testing-backbone-apps-with-jasmine-sinon.html
This is the specific part about Model testing.
Don't know if will solve your problem, but may contain precious info.
this.fixtureResponse is the source data for the model, but when the model is actually created it makes a copy of that data to an internal property. So, when Backbone Relational resolves the relation, it shouldn't change the source data object.
Did you tried with expect(this.model.get('recipient_user')).toBeDefined()?
Backbone-Relational provides the ability to either create a related model from nested entities within JSON retrieved via the model's fetch or to lazily load related models using fetchRelated.
You're providing Backbone-Relational with the message model data but no way to retrieve the user model data. You could add another response returning the appropriate related user data and call fetchRelated on the message model.
Alternatively inline the user data into the message response and the user model will be created automatically and added as a relation on the message model.
I would like to make a relation between two models User and Task using backbone-relational.
The relation between the two models is the following:
taskModel.creator_id = userModel.id
// TaskModel
var TaskModel = Backbone.RelationalModel.extend({
relations: [
{
type: Backbone.HasOne,
key: 'creator',
keySource: 'creator_id',
relatedModel: Users
}
],
// some code
});
// Task collection
var TaskCollection = Backbone.Collection.extend({
model: TaskModel,
// some code
});
// User Model
var User = Backbone.RelationalModel.extend({
// some code
});
Actually the problem is in the collection.models, please see the attached images:
Please check this jsfiddle: http://jsfiddle.net/2bsE9/5/
var user = new User(),
task = new Task(),
tasks = new Tasks();
task.fetch();
user.fetch();
tasks.fetch();
console.log(user.attributes, task.attributes, tasks.models);
P.S.:
Actually I am using requireJs to get the UserModel, so I cannot include quotes in relatedModel value.
define([
'models/user',
'backbone',
'relationalModel'
], function (User) {
"use strict";
var Task = Backbone.RelationalModel.extend({
relations: [
{
type: Backbone.HasOne,
key: 'creator',
keySource: 'creator_id',
relatedModel: User
}
],
});
);
Edit 2:
http://jsfiddle.net/2bsE9/13/
I updated the jsfiddle to reflect the changes I suggested below. As long as you are calling toJSON on your task, what gets to the server is a json object with the creator_id property set to the actual id of the user. The keyDestination here is redundant as the documentation states it is set automatically if you use keySource.
Edit:
https://github.com/PaulUithol/Backbone-relational#keysource
https://github.com/PaulUithol/Backbone-relational#keydestination
https://github.com/PaulUithol/Backbone-relational#includeinjson
The combination of the three above might solve your issue.
var Task = Backbone.RelationalModel.extend({
relations: [
{
type: Backbone.HasOne,
// The User object can be accessed under the property 'creator'
key: 'creator',
// The User object will be fetched using the value supplied under the property 'creator_id'
keySource: 'creator_id',
// The User object will be serialized to the property 'creator_id'
keyDestination: 'creator_id',
// Only the '_id' property of the User object will be serialized
includeInJSON: Backbone.Model.prototype.idAttribute,
relatedModel: User
}
],
});
The documentation also states that the property specified by keySource or keyDestination should not be used by your code. The property cannot be accessed as an attribute.
Please try this and comment if that fixes your issue.
Btw, here is a nice blog post that uses backbone-relational end to end.
http://antoviaque.org/docs/tutorials/backbone-relational-tutorial/
Edit
Updated jsfiddle
The problem is that Backbone-Relational explicitly deletes the keySource to 'prevent leaky abstractions'. It has a hardcoded call to unset on the attribute, in Backbone-Relational:
// Explicitly clear 'keySource', to prevent a leaky abstraction if 'keySource' differs from 'key'.
if ( this.key !== this.keySource ) {
this.instance.unset( this.keySource, { silent: true } );
}
You will need to overwrite the unset method in your Task model:
var Task = Backbone.RelationalModel.extend({
urlRoot: ' ',
relations: [
{
type: Backbone.HasOne,
key: 'creator',
relatedModel: User,
keySource: 'creator_id'
}
],
unset: function(attr, options) {
if (attr == 'creator_id') {
return false;
}
// Original unset from Backbone.Model:
(options || (options = {})).unset = true;
return this.set(attr, null, options);
},
sync: function (method, model, options) {
options.success({
id: 1,
name: 'barTask',
creator_id: 1
});
}
});
Obvious problems with this approach are that you will need to modify your code if either Backbone changes its Backbone.Model.unset method or Backbone-Relational changes its keySource behavior.