ember-orbit
(or "EO") is a library that integrates
orbit.js with
ember.js to provide flexibility and
control in your application's data layer.
-
Access to the full power of Orbit and its ecosystem, including compatible sources, buckets, and coordination strategies.
-
A data schema that's declared through simple model definitions.
-
Stores that wrap Orbit sources and provide access to their underlying data as easy to use records and record arrays. These stores can be forked, edited in isolation, and merged back to the original as a coalesced changeset.
-
Live-updating filtered query results and model relationships.
-
The full power of Orbit's composable query expressions.
-
The ability to connect any number of sources together with Orbit coordination strategies.
-
Orbit's git-like deterministic change tracking and history manipulation capabilities.
EO provides a thin "Emberified" layer over the top of some core
primitives from Orbit, including Store
, Cache
, and Model
classes. Most
common developer interactions with Orbit will be through these classes.
EO does not attempt to wrap every base class from Orbit. For instance, you'll
need to use Orbit's Coordinator
and coordination strategies to define
relationships between Orbit sources. In this way, you can install any Orbit
Source
or Bucket
library and wire them together in your EO application.
Important: It's strongly recommended that you read the Orbit guides at orbitjs.com before using EO, since an understanding of Orbit is vital to making the most of EO.
- Ember.js v3.24 or above
- Ember CLI v3.24 or above
- Node.js v12 or above
EO obeys semver and thus should not be considered to have a stable API until 1.0. Until then, any breaking changes in APIs or Orbit dependencies should be accompanied by a minor version bump of EO.
todomvc-ember-orbit is a simple TodoMVC example that uses EO to illustrate a number of possible configurations and application patterns.
Install EO in your project with:
ember install ember-orbit
The generators for orbit sources and buckets will attempt to install any additional orbit-related dependencies.
EO creates the following directories by default:
-
app/data-buckets
- Factories for creating OrbitBucket
s, which are used to save / load orbit application state. -
app/data-models
- For EOModel
classes, which represent data records. -
app/data-sources
- Factories for creating OrbitSource
s, which represent different sources of data. -
app/data-strategies
- Factories for creating Orbit coordination strategies.
Note that "factories" are simply objects with a create
method that serves to
instantiate an object. The factory interface conforms with the expectations of
Ember's DI system.
EO installs the following services by default:
-
store
- Anember-orbit
Store
to manage querying and updating models. -
dataCoordinator
- An@orbit/coordinator
Coordinator
that manages sources and coordination strategies between them. -
dataSchema
- An@orbit/data
Schema
that represents a schema for models that is shared by thestore
and other sources. -
dataKeyMap
- An@orbit/data
KeyMap
that manages a mapping between keys and local IDs for scenarios in which a server does not accept client-generated IDs.
All the directories and services configured by EO can be customized for your app, as described in the "Customizing EO" section below.
Models are used to access the underlying data in an EO Store
.
They provide a proxy to get and set attributes and relationships. In addition,
models are used to define the schema that's shared by the sources in your
Orbit application.
The easiest way to create a Model
class is with the data-model
generator:
ember g data-model planet
This will create the following module in app/data-models/planet.js
:
import { Model } from 'ember-orbit';
export default class Planet extends Model {}
You can then extend your model to include keys, attributes, and relationships:
import { Model, attr, hasOne, hasMany, key } from 'ember-orbit';
export default class Planet extends Model {
@key() remoteId;
@attr('string') name;
@hasMany('moon', { inverse: 'planet' }) moons;
@hasOne('star') sun;
}
You can create polymorphic relationships by passing in an array of types:
import { Model, attr, hasOne, hasMany } from 'ember-orbit';
export default class PlanetarySystem extends Model {
@attr('string') name;
@hasMany(['moon', 'planet']) bodies;
@hasOne(['star', 'binaryStar']) star;
}
EO's Store
class is a thin wrapper around Orbit's
MemorySource
,
while EO's Cache
class wraps Orbit's
MemoryCache
.
The difference between memory sources and caches is explained extensively in
Orbit's docs.
The essential difference between EO's Store
and Cache
and the underlying
Orbit classes is that EO is model-aware. Unlike plain Orbit, in which results
are returned as static POJOS, every query and update result in EO is translated
into Model
instances, or simply "records". When changes occur to the
underlying Orbit sources and caches, they will be reflected immediately in EO's
records.
Every EO record is connected to a cache, which in turn belongs to a store. When
stores or caches provide results in the form of records, they are always
instantiated by, and belong to, a Cache
. For a given identity (type
/ id
pair), there is only ever one record instance per cache. These are maintained in
what is fittingly called an "identity
map".
Records, including all their attributes and relationships, will stay in sync with the underlying data in their associated cache.
There are three primary methods available to query records:
-
store.query()
- returns a promise that resolves to a static recordset. -
store.cache.query()
- returns a static set of in-memory results immediately. -
store.cache.liveQuery()
- returns a live recordset that will be refreshed whenever the data changes in the cache.
All of these query methods take the same arguments as any other queryable Orbit source - see the Orbit guides for details.
The following liveQuery
immediately returns a live resultset that will stay
updated with the "terrestrial" planets in the store:
let planets = store.cache.liveQuery((qb) =>
qb
.findRecords('planet')
.filter({ attribute: 'classification', value: 'terrestrial' })
);
The EO Store
also supports findRecord
and findRecords
methods. These
methods are async and call query
internally:
// find all records of a type
let planets = await store.findRecords('planet');
// find a specific record by type and id
let planet = await store.findRecord('planet', 'abc123');
These methods are also available on the EO Cache
, but are synchronous:
// find all records of a type
let planets = store.cache.findRecords('planet');
// find a specific record by type and id
let planet = store.cache.findRecord('planet', 'abc123');
There are two primary approaches to update data in EO:
-
Directly via async methods on the main
Store
. Direct updates flow immediately into Orbit's request flow, where they can trigger side effects, such as remote server requests. -
In an isolated "forked"
Store
, usually via sync methods on its associatedCache
and/orModel
instances. These changes remain in this fork until they are merged back to a base store.
The Store
exposes several async methods to update data:
addRecord
- adds a single record.updateRecord
- updates the fields of a single record.updateRecordFields
- for updating the fields of a single record, with a first argument that provides the identity of the record.removeRecord
- removes a single record.update
- the most flexible and powerful method, which can perform one or more operations in a single request.
Here are some examples of each:
// add a new record (returned as a Model instance)
let planet = await store.addRecord({ type: 'planet', id: '1', name: 'Earth' });
console.log(planet.name); // Earth
// update one or more fields of the record
await store.updateRecord({ type: 'planet', id: '1', name: 'Mother Earth' });
console.log(planet.name); // Mother Earth
// remove the record
await store.removeRecord({ type: 'planet', id: '1' });
// or alternatively: await store.removeRecord(planet);
// add more planets in a single `Transform`
let [mars, venus] = await store.update((t) => [
t.addRecord({ type: 'planet', name: 'Mars' }),
t.addRecord({ type: 'planet', name: 'Venus' })
]);
EO stores can be forked and merged, just as described in the Orbit guides.
Once you have forked a store, you can proceed to make synchronous changes to the
fork's associated Cache
and/or Model
instances. These changes will be
tracked and can then be merged back to the base store.
Here's an example:
// (async) start by adding two planets and a moon to the store
await store.update(t => [
t.addRecord(earth),
t.addRecord(venus),
t.addRecord(theMoon)
]);
// (async) query the planets in the store
let planets = await store.query(q => q.findRecords('planet').sort('name')));
console.log('original planets', planets);
// (sync) fork the store
forkedStore = store.fork();
let forkedCache = forkedStore.cache;
// (sync) add a planet and moons to the fork's cache
forkedCache.update(t => [
t.addRecord(jupiter),
t.addRecord(io),
t.addRecord(europa)
]);
// (sync) query the planets in the forked cache
planets = forkedCache.query(q => q.findRecords('planet').sort('name')));
console.log('planets in fork', planets);
// (async) merge the forked store back into the original store
await store.merge(forkedStore);
// (async) query the planets in the original store
planets = await store.query(q => q.findRecords('planet').sort('name')));
console.log('merged planets', planets);
Some notes about forking / merging:
-
Once a store has been forked, the original and forked stores’ data can diverge independently.
-
Merging a fork will coalesce any changes made to the forked cache into a single new transform, and then update the original store.
-
A store fork can simply be abandoned without cost. Just remember to free any references to the JS objects themselves.
Important - One additional concern to be aware of is that EO will generate new records for each store. Care should be taken to not mix records between stores, since the underlying data in each store can diverge. If you need to access a record in a store's fork, just query the forked store or cache for that record.
The Cache
exposes sync versions of the Store
's async update methods:
addRecord
- for adding a single record.updateRecord
- for updating the fields of a single record.removeRecord
- for removing a single record.update
- the most flexible and powerful method, which can perform one or more operations in a single request.
By default, only forked caches are able to be updated directly. This provides
protection against data loss, since changes to caches do not participate in
Orbit's data flows. An exception is
made for forks because the changes are tracked and applied back to stores via
merge
.
If you want to override these protections and update a non-forked cache, you can
set cache.allowUpdates = true
, but know that those updates won't leave the
cache.
Each Model
exposes all of its fields, including attributes and relationships,
as properties that stay updated.
Attributes and has-one relationships are also directly editable. For instance:
let jupiter = forkedCache.findRecord('planet', 'jupiter');
let sun = forkedCache.findRecord('star', 'theSun');
console.log(jupiter.name); // 'Jupiter'
// update attribute
jupiter.name = 'Jupiter!';
console.log(jupiter.name); // 'Jupiter!'
// update has-one relationship
jupiter.sun = theSun;
In order to not conflict with user-defined fields, all standard methods on
Model
are prefixed with a $
. The following synchronous methods are
available:
$replaceAttribute
$replaceRelatedRecord
$replaceRelatedRecords
$addToRelatedRecords
$removeFromRelatedRecords
$update
$remove
let jupiter = forkedCache.findRecord('planet', 'jupiter');
let io = forkedCache.findRecord('moon', 'io');
let europa = forkedCache.findRecord('moon', 'europa');
let sun = forkedCache.findRecord('star', 'theSun');
jupiter.$replaceAttribute('name', 'JUPITER!');
jupiter.$addToRelatedRecords('moons', io);
jupiter.$removeFromRelatedRecords('moons', europa);
jupiter.$replaceRelatedRecord('sun', sun);
console.log(jupiter.name); // 'JUPITER!'
console.log(jupiter.moons.includes(io)); // true
console.log(jupiter.moons.includes(europa)); // false
console.log(jupiter.sun.id); // 'theSun'
Behind the scenes, these changes each result in a call to forkedCache.update
.
Of course, this method could also be called directly instead of issuing updates
through the model:
forkedCache.update((t) => [
t.replaceAttribute(jupiter, 'name', 'JUPITER!');
t.addToRelatedRecords(jupiter, 'moons', io);
t.removeFromRelatedRecords(jupiter, 'moons', europa);
t.replaceRelatedRecord(jupiter, 'sun', sun);
]);
The store's contents exist only in memory and will be cleared every time your
app restarts. Let's try adding another source and use the dataCoordinator
service to keep it in sync with the store.
We can use the data-source
generator to create a backup
source:
ember g data-source backup --from=@orbit/indexeddb
This will generate a source factory in app/data-sources/backup.js
:
import SourceClass from '@orbit/indexeddb';
import { applyStandardSourceInjections } from 'ember-orbit';
export default {
create(injections = {}) {
applyStandardSourceInjections(injections);
injections.name = 'backup';
return new SourceClass(injections);
}
};
Note that injections
should include both a Schema
and a KeyMap
, which
are injected by default for every EO application. We're also adding
a name
to uniquely identify the source within the coordinator. You could
optionally specify a namespace
to be used to name the IndexedDB database.
Every source that's defined in app/data-sources
will be discovered
automatically by EO and added to the dataCoordinator
service.
Next let's define some strategies to synchronize data between sources.
There are four different types of coordination strategies that can be generated
by default using the standard data-strategy
generator:
request
sync
event-logging
log-truncation
Let's define a sync strategy to backup changes made to the store into our new
backup
source.
ember g data-strategy store-backup-sync --type=sync
This should create a SyncStrategy
factory in
app/data-strategies/store-backup-sync.js
as follows:
import { SyncStrategy } from '@orbit/coordinator';
export default {
create() {
return new SyncStrategy({
name: 'store-backup-sync',
/**
* The name of the source which will have its `transform` event observed.
*/
source: 'store',
/**
* The name of the source which will be acted upon.
*
* When the source receives the `transform` event, the `sync` method
* will be invoked on the target.
*/
target: 'backup',
/**
* A handler for any errors thrown as a result of invoking `sync` on the
* target.
*/
// catch(e) {},
/**
* A filter function that returns `true` if `sync` should be performed.
*
* `filter` will be invoked in the context of this strategy (and thus will
* have access to both `this.source` and `this.target`).
*/
// filter(...args) {};
/**
* Should resolution of the target's `sync` block the completion of the
* source's `transform`?
*
* Can be specified as a boolean or a function which which will be
* invoked in the context of this strategy (and thus will have access to
* both `this.source` and `this.target`).
*/
blocking: true
});
}
};
You should also consider adding an event logging strategy to log events emitted from your sources to the browser console:
ember g data-strategy event-logging
Sources have another kind of log as well: a transform log, which tracks transforms that are applied. A log truncation strategy will keep the size of transform logs in check. It observes the sources associated with the strategy and truncates their transform logs when a common transform has been applied to them all. Let's add a log truncation strategy as well:
ember g data-strategy log-truncation
Next we'll need to activate our coordinator as part of our app's boot process. The coordinator requires an explicit activation step because the process is async and we may want to allow developers to do work beforehand.
In our case, we want to restore our store from the backup source before we
enable the coordinator. Let's do this in our application route's beforeModel
hook (in app/routes/application.js
):
import Route from '@ember/routing/route';
import { inject as service } from '@ember/service';
export default class ApplicationRoute extends Route {
@service dataCoordinator;
@service store;
async beforeModel() {
// Populate the store from backup prior to activating the coordinator
const backup = this.dataCoordinator.getSource('backup');
const records = await backup.query((q) => q.findRecords());
await this.store.sync((t) => records.map((r) => t.addRecord(r)));
await this.dataCoordinator.activate();
}
}
This code first pulls all the records from backup and then syncs them with the main store before activating the coordinator. In this way, the coordination strategy that backs up the store won't be enabled until after the restore is complete.
Data buckets are used by sources and key maps to load and persist state. You will probably want to use a bucket if you plan to support any offline or optimistic UX.
To create a new bucket, run the generator:
ember g data-bucket main
By default this will create a new bucket factory based on @orbit/indexeddb-bucket
.
It will also create an initializer that injects this bucket into all your
sources and key maps.
The types, collections, and services used by EO can all be customized for
your application via settings under the orbit
key in config/environment
:
module.exports = function (environment) {
let ENV = {
// ... other settings here
// Default Orbit settings (any of which can be overridden)
orbit: {
schemaVersion: undefined,
types: {
bucket: 'data-bucket',
model: 'data-model',
source: 'data-source',
strategy: 'data-strategy'
},
collections: {
buckets: 'data-buckets',
models: 'data-models',
sources: 'data-sources',
strategies: 'data-strategies'
},
services: {
store: 'store',
bucket: 'data-bucket',
coordinator: 'data-coordinator',
schema: 'data-schema',
keyMap: 'data-key-map',
normalizer: 'data-normalizer',
validator: 'data-validator'
},
skipStoreService: false,
skipBucketService: false,
skipCoordinatorService: false,
skipSchemaService: false,
skipKeyMapService: false,
skipNormalizerService: false,
skipValidatorService: false
}
};
return ENV;
};
Note that schemaVersion
should be set if you're using any Orbit sources, such
as IndexedDBSource
, that track schema version. By default, Orbit's schema
version will start at 1
. This value should be bumped to a a higher number with
each significant change that requires a schema migration. Migrations themselves
must be handled in each individual source.
Sources and strategies may be conditionally included in your app's coordinator
by customizing the default export of the source / strategy factory. A valid
factory is an object with the interface { create: () => {} }
. If a valid
factory is not the default export for your module, it will be ignored.
For example, the following strategy will be conditionally included for all non-production builds:
// app/data-strategies/event-logging.js
import { EventLoggingStrategy } from '@orbit/coordinator';
import config from 'example/config/environment';
const factory = {
create() {
return new EventLoggingStrategy();
}
};
// Conditionally include this strategy
export default config.environment !== 'production' ? factory : null;
Like Orbit itself, EO enables validators by default in all sources. EO provides
the same set of validators to all sources by building a single data-validator
service that is injected into all sources.
Validators are useful to ensure that your data matches its type expectations and
that operations and query expressions are well formed. Of course, they also add
some extra code and processing, which you may want to eliminate (or perhaps only
for production environments). You can disable validators across all sources by
setting Orbit's skipValidatorService
environment flag to false
in
config/environment
, as described above.
If you want to use validators but extend them to include custom validators, you
can override the standard validator service by generating your own
data-validator
service that passes custom arguments to
buildRecordValidatorFor
.
For instance, in order to provide a custom validator for an address
type:
// app/services/data-validator.js
import { buildRecordValidatorFor } from '@orbit/records';
const validators = {
address: (input) => {
if (typeof input?.country !== 'string') {
return [
{
validator: 'address',
validation: 'country',
description: 'is not a string',
ref: input,
},
];
}
},
};
export default {
create() {
return buildRecordValidatorFor({ validators });
},
};
This custom validator service will be injected into all your orbit sources via
applyStandardSourceInjections
, as described above.
git clone https://github.com/orbitjs/ember-orbit.git
cd ember-orbit
yarn install
yarn test
EO owes a great deal to Ember Data, which has influenced the design of many of EO's interfaces. Many thanks to the Ember Data Core Team, including Yehuda Katz, Tom Dale, and Igor Terzic, for their work.
It is hoped that, by tracking Ember Data's features and interfaces where possible, EO will also be able to contribute back to Ember Data.
Copyright 2014-2021 Cerebris Corporation. MIT License (see LICENSE for details).