Ember Data IndexedDB Adapter

Store your Ember application data offline with IndexedDB. Compatible with Ember Data 1.0.0-beta7. Fully tested.

Usage

Download the latest distribution or use bower to install ember-indexeddb-adapter.

App.ApplicationSerializer = DS.IndexedDBSerializer.extend();
App.ApplicationAdapter = DS.IndexedDBAdapter.extend({
  databaseName: 'some_database_name',
  version: 1,
  migrations: function() {
    this.addModel('person');
    this.addModel('phone');
  }
});

Define your models relationships with {async: false}. All relationships are retrieved automatically in a query. Just do the regular:

App.Person = DS.Model.extend({
  name: DS.attr('string'),
  phones: DS.hasMany('phone') // no {async: true} calls here.
});

Note: use a string as parameter for this.addModel('person').

Version and Migrations

Different from localStorage, IndexedDB requires you to define the Object Stores you'll use (think of them like database table). By defining the models to be used inside the adapter's migrations function, it'll update the schema whenever needed.

Remember that whenever you want to update this schema, you need to increment the version number (integer). Only so will IndexedDB commit the changes.

KeyPath and autoIncrement

By default, the keyPath (id field) is id and autoIncrement is set to false. You can change these doing this:

// inside migrations function
this.addModel('person', {keyPath: 'id', autoIncrement: true});

Important: if you define autoIncrement: true, we won't use Ember's provided ID, but simply throw away it with record.serialize({includeId: false}). If you don't know what you're doing, just leave it as false so Ember Data can take care of the ID for you.

SmartSearch feature

SmartSearch (disabled by default) allows you to make more meaningful queries, ones that you would probably perform against a REST api. With it, you will be able to do the following queries.

Dates

If you have a field with a date transform, the adapter will try to match the records that match a given date.

App.Person = DS.Model.extend({
  name: DS.attr('string'),
  createdAt: DS.attr('date')
});

// Date matches
store.findQuery('person', {createdAt: "today"})
store.findQuery('person', {createdAt: "yesterday"})
store.findQuery('person', {createdAt: "32 days ago"})

Full-text search

Whenever you use a search key in your query, the adapter will try to match it against all fields. Consider the database has the following:

[ { id: 1, name: "Rambo",    cool: false },
  { id: 2, name: "Braddock", cool: true  } ]

The following will match different results:

// returns only the first record
store.findQuery('person', {search: /rambo|braddock/i, cool: false})
// returns only the first record
store.findQuery('person', {search: "rmb", cool: false})
// returns both records
store.findQuery('person', {search: "ao"})
// returns only the last record
store.findQuery('person', {search: "ao", cool: true})

The adapter will try matching the search value against every record using AND logic. In some cases above, although name was matched, cool didn't match.

You can change the fields that are searched reopening the method that defines if a field is to be searched:

DS.IndexedDBAdapter.reopen({
  findQuerySearchCriteria: function(fieldName, type) {
    if (type.toString() == "App.User" && fieldName == "name") {
      return false;
    } else {
      return true;
    }
  }
});

With the redefinition above, the search will try to match every field, except App.User's name.

Enabling SmartSearch

If you want to enable smartSearch, write this:

DS.IndexedDBAdapter.reopen({
  smartSearch: true
});

Building from source

In the root folder, type in your terminal rake build. Make sure you have Ruby installed. The file is generated in dist/.

Areas for improvement

The following are areas that needs some improvements. We'd love if you could send a PR for one of them.

smarter transactions: we open a transaction, get some stuff, then close it and move on. However, if we're going to load a relationship next, it happens in a new transaction. It should all happen in one transaction.
indexes: as of now, there is no migrations API for creating indexes.
search: there's no way to do a fuzzy search with IndexedDB (besides the basic string match). Right now, we go through all records matching them, but later we'll want to create an objectStore to cache all strings with the original record ID. That way we'd be able to cut down the amount of data we put in memory.

Tests

First, install depdendencies with bower: bower install.

Run rackup in your terminal (make sure you have Ruby and the rack gem installed). Then visit http://localhost:9292 in your browser.

Please, disregard Travis CI for now because PhantomJS (1.9.3) doesn't support IndexedDB. IndexedDBShim.js doesn't work on it either, so I'm running tests only in the browser for now. Once version 2 is here, just use phantomjs tests/runner.js tests/index.html in your terminal.

kurko/ember-indexeddb-adapter