0.0.6-alpha-3 • Published 9 years ago

ember-data-complex v0.0.6-alpha-3

Weekly downloads
1
License
MIT
Repository
github
Last release
9 years ago

Ember Data Complex

Build Status

An extension of / set of patterns for the very popular Ember Data to help build data models that require complex adapter / serializer strategy.

For when you need a model with attributes that come from differing adapters.

All the tests pass and I'm using it in my own projects fine... but this is alpha software

Example

Suppose you're building a locally distributed warehouse application that tracks trucks. In addition to storing trucks on your rails database, you're also real-time sharing them across your app's users via firebase, and, on top of that, you're interested in long term archival storage via amazon s3.

In other words, you need a truck database model that has great real-time data-concurrency, long-term query-able storage, and large file bucket storage. Perhaps if you're Google, you can build such a database, but for the rest of us, we'll need to use 3 different storage options and compose them together in a manageable way on the front end.

Ember Data Complex exposes a pattern and some helpers functions to help you manage your complex data model:

First, write whatever serializers and adapters you'll need

# serializers/fires/truck.coffee
FireTruckSerializer = DS.FirebaseSerializer.extend
  ...

# adapters/fires/truck.coffee
FireTruckAdapter = DS.FirebaseAdapter.extend
  ...

# serializers/rails/truck.coffee
RailsTruckSerializer = DS.ActiveModelSerializer.extend
  ...

# adapters/rails/truck.coffee
RailsTruckAdapter = DS.ActiveModelAdapter.extend
  ...

# serializers/s3/truck.coffee
S3TruckSerializer = DS.AmazonS3Serializer.extend
  ...

# adapters/s3/truck.coffee
S3TruckAdapter = DS.AmazonS3Adapter.extend
  ...

Next, declare your aristocrat master model:

# models/truck.coffee
Truck = DSC.ModelComplex.extend
  fireId: DS.attr "string"
  railsId: DS.attr "string"
  s3Id: DS.attr "string"

  fire: DSC.belongsTo2 "fires/truck", foriegnKey: "fireId", promiseField: "firePromise"
  rails: DSC.belongsTo2 "rails/truck", foriegnKey: "railsId", promiseField: "railsPromise"
  s3: DSC.belongsTo2 "s3/truck", foriegnKey: "s3id", promiseField: "s3Promise"

Here, DSC.Macros.through is just a macro over store.find; it works a lot like DS.belongsTo except it returns a computed property instead of a relationship. This is important because in general rails/truck and s3/truck will different adapters than truck, so a regular DS relationship won't work here.

Incidentally, this means DSC.belongsTo2 is always async.

Next, you'll need to declare your peasant slave models; these are just regular DS.Models.

# models/fires/truck.coffee
Truck = DS.Model.extend
  latitude: DS.attr "number"
  longitude: DS.attr "number"
  licensePlate: DS.attr "string"
  driver: DS.attr "string"

# models/rails/truck.coffee
Truck = DS.Model.extend
  licensePlate: DS.attr "string"
  driver: DS.attr "string"

# models/s3/truck.coffee
Truck = DS.Model.extend
  shippingDocs: DS.attr "file"
  weightPass: DS.attr "file"

At this point, you're ready to go and the models you declared will work exactly as you expected

truck = @store.find "truck", 33
truck.get("rails").set "driver", "bob"
truck.get("rails").save()
truck.save()
truck.destroy()

But so far, Data Complex has done nothing that Data couldn't do for you (after all, you had to declare everything you needed to using regular Ember Data).

However, DataComplex allows you to declare strategies regarding how your group of models should be found, saved, updated and destroyed (updated and destroyed not yet implemented, sorry, lol):

# strategies/truck.coffee
TruckStrategy = DSC.Strategy.extend
  # the one argument is a promise to your model from upstream
  onFindById: (masterTruck) ->
    masterTruck.then (truck) ->
      truck.get("fire").then -> truck
    .then (truck) ->
      truck.get("rails").then -> truck
    .then whatever
  onFindAll: (trucks) -> ...
  onFindByQuery: (trucks) -> ...
  # runs when save is called on your new model,
  # but before it is scheduled and marked inflight in the adapter
  # unsavedTruck is NOT a promise and is instead an DS.Model
  beforeCreate: (unsavedTruck) -> ...

Note that, you should always return a promise from onFind. When that promise resolves (or rejects), DataComplex will automatically resolve out to the original truck you tried to @store.find on.

In this way, you can think of your master truck as a purely lazy data structure (evaluted to Weak-Head Normal Form) and the strategy object as a way of evaluating (aka normal-forming) that lazy structure. If you've done parallel haskell, this is the Eval Monad applied here to front-end models.

Tactic Mixins for Strategy

Strategies wind up sharing a lot of common code. So far, DSC ships with 2 tactics (mixins) for your convenience.

  1. Creative Delegation
TruckStrategy = DSC.Strategy.extend DSC.CreativeDelegationTactic,
  # your stuff
  ...

CreativeDelegation provides a beforeCreate method so you can pass in embedded objects and get them all created:

store.createRecord "truck"
  s3:
    shippingDocs: somefile
    weightPass: someotherfile
  rails:
    licensePlate: 'yoshi-420'
    driver: 'mario'
  fire:
    latitude: 66.6
    longitude: 66.6
    licensePlate: 'yoshi-420'
    driver: 'mario'
  .save() # creates the master truck, and slave s3, rails, and fire trucks, and ties them all together
  .then -> # stuff
  .catch (error) ->
    error.message # some long-winded thing
    error.orphans # models that saved successfully
    error.deadChildren # an array of objects with data on the children that failed to persist

If any of the children fail, the entire save process will fail, and you can catch and cleanup as necessary.

A custom error object with what I personally needed is provided you.

  1. Fallback Cache
TruckStrategy = DSC.Strategy.extend DSC.FallbackCacheTactic,
  order: ['fire', 'rails']
  # your stuff
  ...

You must specify an order to your caching and FallbackCache will provide onFindById and onFindAll methods.

Given your specified cache order, suppose truck#77 exists only in your rails upstream, but not fire upstream...

store.find "truck", 77

Will find your truck, miss the cache hit to fire, try rails, succeed, then store.createRecord "fires/truck" to the fire cache.

If all your cache tiers fall through (aka you have no slaves on your master), the promise will reject.

  1. TODOS: probably add an UpdativeDelegation and FamilyObliteration tactics to round out the basic use cases.

Regarding promises and strategies

This library ships with a small async library for using es6 generators to deal with stuff. Here is an example:

DSC.async ->
  rover = yield find "dog", "rover"

  DSC.ifA rover
  .thenA ->
    rover.get "id" # rover
  .elseA ->
    rover instanceof Error # true
    rover.message # unable to find rover
    rover.reason # returns the caught reason of why rover failed
  .end()

important If you intend on using es6 generators and coffeescript concurrently in your ember app, there is currently an issue where ember-cli-coffeescript doesn't pipe its output into ember-cli-babel. I've already done a pull-request to ember-cli-coffeescript, but for now, you need to use my fork of ember-cli-coffeescript at:

https://github.com/foxnewsnetwork/ember-cli-coffeescript

also important we're still waiting on ember-cli-babel to properly polyfill calls to the rengeratorRuntime in areas that want to use es6 generators. You can see the progress of this issue here:

https://github.com/babel/ember-cli-babel/issues/24

For now, the work-around is to manually copy-and-paste the following file:

https://github.com/facebook/regenerator/blob/master/runtime.js

from facebook's regenerator repo into your vendor/ directory, and then edit your brocfile so it says:

// Brocfile.js
// other stuff
app.import("vendor/regenerator.js");
// other stuff

the case for using DSC.ifA instead of javascript's native 'if' is that ifA is a lot more context-aware.

ifA runs the thenA block if the thing you gave it is:

  • a promise resolving to something truthy
  • truthy (that is, not blank, not null, but 0 is ok)
  • a function that returns something truthy

ifA runs the elseA block (if you provided one) if the thing you gave it:

  • is not truthy (null, "", false)
  • a promise resolving to something not truthy
  • a function that returns something not truthy
  • something that is an instanceof Error

if you don't call end() (or run()) at the end of an DSC.ifA block, then no evaluation happens and you just have an arrow (see next section). For all intents and purposes, arrows are objects with a run function and can be composed in various interesting ways that a mere function can't. This allows you to do really clever but incredibly ill-advised meta programming.

Regarding arrows

For those interested in writing elegant (although newbie-unfriendly) code with promises, it should be noted promises are more than monadic (read about them here: http://en.wikipedia.org/wiki/Monad_(functional_programming) ) they are also an arrow (read about them here: https://www.haskell.org/arrows/ ).

Completely incomprehensible academic horseshit aside, the arrow idea is actually extremely helpful to writing non-cancerous code with promises. DataComplex ships with a small Arrow library to faciliate this.

In continuing with the truck example, you can take advantage of arrows in your strategies:

import { lift, ifA } from 'ember-data-complex/utils/arrows'

# strategies/truck.coffee
TruckStrategy = DSC.Strategy.create
  attemptFirstTruck: (masterTruck) ->
    @store.find "fires/truck", masterTruck.get "id"
  attemptRailsTruck: (masterTruck) ->
    @store.find "rails/truck", masterTruck.get "id"
  onFind: (masterTruck) ->
    lift masterTruck
    .await ifA @attemptFireTruck.bind(@)
    .await doNothing.elseA @attemptRailsTruck.bind(@)

I will probably write up a more involved / explanatory example later, but the idea behind lifting (wrapping into a context) something to an arrow is to be able to employ a bastardized semi-es7-await-syntax control flow to async computations.

Installation

  • git clone this repository
  • npm install
  • bower install

Running

Running Tests

  • ember test
  • ember test --server

Building

  • ember build

For more information on using ember-cli, visit http://www.ember-cli.com/.