Active Model Datastore

Makes the google-cloud-datastore gem compliant with active_model conventions and compatible with your Rails 5+ applications.

Why would you want to use Google's NoSQL Cloud Datastore with Rails?

When you want a Rails app backed by a managed, massively-scalable datastore solution. Cloud Datastore automatically handles sharding and replication. It is a highly available and durable database that automatically scales to handle your applications' load. Cloud Datastore is a schemaless database suited for unstructured or semi-structured application data.

Setup
Model Example
Controller Example
Retrieving Entities
Datastore Consistency
Datastore Indexes
Datastore Emulator
Example Rails App
CarrierWave File Uploads
Track Changes
Nested Forms
Datastore Gotchas

Setup

Generate your Rails app without ActiveRecord:

rails new my_app -O

You can remove the db/ directory as it won't be needed.

To install, add this line to your Gemfile and run bundle install:

gem 'activemodel-datastore'

Create a Google Cloud account here and create a project.

Google Cloud requires the Project ID and Service Account Credentials to connect to the Datastore API.

Follow the activation instructions to enable the Google Cloud Datastore API. When running on Google Cloud Platform environments the Service Account credentials will be discovered automatically. When running on other environments (such as AWS or Heroku) you need to create a service account with the role of editor and generate json credentials.

Set your project id in an ENV variable named GCLOUD_PROJECT.

To locate your project ID:

Go to the Cloud Platform Console.
From the projects list, select the name of your project.
On the left, click Dashboard. The project name and ID are displayed in the Dashboard.

If you have an external application running on a platform outside of Google Cloud you also need to provide the Service Account credentials. They are specified in two additional ENV variables named SERVICE_ACCOUNT_CLIENT_EMAIL and SERVICE_ACCOUNT_PRIVATE_KEY. The values for these two ENV variables will be in the downloaded service account json credentials file.

SERVICE_ACCOUNT_PRIVATE_KEY = -----BEGIN PRIVATE KEY-----\nMIIFfb3...5dmFtABy\n-----END PRIVATE KEY-----\n
SERVICE_ACCOUNT_CLIENT_EMAIL = web-app@app-name.iam.gserviceaccount.com

On Heroku the ENV variables can be set under 'Settings' -> 'Config Variables'.

Active Model Datastore will then handle the authentication for you, and the datastore instance can be accessed with CloudDatastore.dataset.

There is an example Puma config file here.

Model Example

Let's start by implementing the model:

class User
  include ActiveModel::Datastore

  attr_accessor :email, :enabled, :name, :role, :state

  def entity_properties
    %w[email enabled name role]
  end
end

Data objects in Cloud Datastore are known as entities. Entities are of a kind. An entity has one or more named properties, each of which can have one or more values. Think of them like this:

'Kind' (which is your table and the name of your Rails model)
'Entity' (which is the record from the table)
'Property' (which is the attribute of the record)

The entity_properties method defines an Array of properties that belong to the entity in cloud datastore. Define the attributes of your model using attr_accessor. With this approach, Rails deals solely with ActiveModel objects. The objects are converted to/from entities automatically during save/query operations. You can still use virtual attributes on the model (such as the :state attribute above) by simply excluding it from entity_properties. In this example state is available to the model but won't be persisted with the entity in datastore.

Validations work as you would expect:

class User
  include ActiveModel::Datastore

  attr_accessor :email, :enabled, :name, :role, :state

  validates :email, format: { with: /\A([^@\s]+)@((?:[-a-z0-9]+\.)+[a-z]{2,})\z/i }
  validates :name, presence: true, length: { maximum: 30 }

  def entity_properties
    %w[email enabled name role]
  end
end

Callbacks work as you would expect. We have also added the ability to set default values through default_property_value and type cast the format of values through format_property_value:

class User
  include ActiveModel::Datastore

  attr_accessor :email, :enabled, :name, :role, :state

  before_validation :set_default_values
  after_validation :format_values
  
  before_save { puts '** something can happen before save **'}
  after_save { puts '** something can happen after save **'}

  validates :email, format: { with: /\A([^@\s]+)@((?:[-a-z0-9]+\.)+[a-z]{2,})\z/i }
  validates :name, presence: true, length: { maximum: 30 }
  validates :role, presence: true

  def entity_properties
    %w[email enabled name role]
  end

  def set_default_values
    default_property_value :enabled, true
    default_property_value :role, 1
  end

  def format_values
    format_property_value :role, :integer
  end
end

Controller Example

Now on to the controller! A scaffold generated controller works out of the box:

class UsersController < ApplicationController
  before_action :set_user, only: [:show, :edit, :update, :destroy]

  def index
    @users = User.all
  end

  def show
  end

  def new
    @user = User.new
  end

  def edit
  end

  def create
    @user = User.new(user_params)
    respond_to do |format|
      if @user.save
        format.html { redirect_to @user, notice: 'User was successfully created.' }
      else
        format.html { render :new }
      end
    end
  end

  def update
    respond_to do |format|
      if @user.update(user_params)
        format.html { redirect_to @user, notice: 'User was successfully updated.' }
      else
        format.html { render :edit }
      end
    end
  end

  def destroy
    @user.destroy
    respond_to do |format|
      format.html { redirect_to users_url, notice: 'User was successfully destroyed.' }
    end
  end

  private

  def set_user
    @user = User.find(params[:id])
  end

  def user_params
    params.require(:user).permit(:email, :name)
  end
end

Retrieving Entities

Each entity in Cloud Datastore has a key that uniquely identifies it. The key consists of the following components:

the kind of the entity, which is User in these examples
an identifier for the individual entity, which can be either a a key name string or an integer numeric ID
an optional ancestor path locating the entity within the Cloud Datastore hierarchy

all(options = {})

Queries entities using the provided options. When a limit option is provided queries up to the limit and returns results with a cursor.

users = User.all(options = {})

parent_key = CloudDatastore.dataset.key('Parent', 12345)
users = User.all(ancestor: parent_key)

users = User.all(ancestor: parent_key, where: ['name', '=', 'Bryce'])

users = User.all(where: [['name', '=', 'Ian'], ['enabled', '=', true]])

users, cursor = User.all(limit: 7)

# @param [Hash] options The options to construct the query with.
#
# @option options [Google::Cloud::Datastore::Key] :ancestor Filter for inherited results.
# @option options [String] :cursor Sets the cursor to start the results at.
# @option options [Integer] :limit Sets a limit to the number of results to be returned.
# @option options [String] :order Sort the results by property name.
# @option options [String] :desc_order Sort the results by descending property name.
# @option options [Array] :select Retrieve only select properties from the matched entities.
# @option options [Array] :distinct_on Group results by a list of properties.
# @option options [Array] :where Adds a property filter of arrays in the format[name, operator, value].

find(*ids, parent: nil)

Find entity by id - this can either be a specific id (1), a list of ids (1, 5, 6), or an array of ids ([5, 6, 10]). The parent key is optional. This method is a lookup by key and results will be strongly consistent.

user = User.find(1)

parent_key = CloudDatastore.dataset.key('Parent', 12345)
user = User.find(1, parent: parent_key)

users = User.find(1, 2, 3)

find_by(args)

Queries for the first entity matching the specified condition.

user = User.find_by(name: 'Joe')

user = User.find_by(name: 'Bryce', ancestor: parent_key)

Cloud Datastore has documentation on how Datastore Queries work, and pay special attention to the the restrictions.

Datastore Consistency

Cloud Datastore is a non-relational databases, or NoSQL database. It distributes data over many machines and uses synchronous replication over a wide geographic area. Because of this architecture it offers a balance of strong and eventual consistency.

What is eventual consistency?

It means that an updated entity value may not be immediately visible when executing a query. Eventual consistency is a theoretical guarantee that, provided no new updates to an entity are made, all reads of the entity will eventually return the last updated value.

In the context of a Rails app, there are times that eventual consistency is not ideal. For example, let's say you create a user entity with a key that looks like this:

@key=#<Google::Cloud::Datastore::Key @kind="User", @id=1>

and then immediately redirect to the index view of users. There is a good chance that your new user is not yet visible in the list. If you perform a refresh on the index view a second or two later the user will appear.

"Wait a minute!" you say. "This is crap!" you say. Fear not! We can make the query of users strongly consistent. We just need to use entity groups and ancestor queries. An entity group is a hierarchy formed by a root entity and its children. To create an entity group, you specify an ancestor path for the entity which is a parent key as part of the child key.

Before using the save method, assign the parent_key_id attribute an ID. Let's say that 12345 represents the ID of the company that the users belong to. The key of the user entity will now look like this:

@key=#<Google::Cloud::Datastore::Key @kind="User", @id=1, @parent=#<Google::Cloud::Datastore::Key @kind="ParentUser", @id=12345>>

All of the User entities will now belong to an entity group named ParentUser and can be queried by the Company ID. When we query for the users we will provide User.parent_key(12345) as the ancestor option.

Ancestor queries are always strongly consistent.

However, there is a small downside. Entities with the same ancestor are limited to 1 write per second. Also, the entity group relationship cannot be changed after creating the entity (as you can't modify an entity's key after it has been saved).

The Users controller would now look like this:

class UsersController < ApplicationController
  before_action :set_user, only: [:show, :edit, :update, :destroy]

  def index
    @users = User.all(ancestor: User.parent_key(12345))
  end

  def show
  end

  def new
    @user = User.new
  end

  def edit
  end

  def create
    @user = User.new(user_params)
    @user.parent_key_id = 12345
    respond_to do |format|
      if @user.save
        format.html { redirect_to @user, notice: 'User was successfully created.' }
      else
        format.html { render :new }
      end
    end
  end

  def update
    respond_to do |format|
      if @user.update(user_params)
        format.html { redirect_to @user, notice: 'User was successfully updated.' }
      else
        format.html { render :edit }
      end
    end
  end

  def destroy
    @user.destroy
    respond_to do |format|
      format.html { redirect_to users_url, notice: 'User was successfully destroyed.' }
    end
  end

  private

  def set_user
    @user = User.find(params[:id], parent: User.parent_key(12345))
  end

  def user_params
    params.require(:user).permit(:email, :name)
  end
end

See here for the Cloud Datastore documentation on Data Consistency.

Datastore Indexes

Every cloud datastore query requires an index. Yes, you read that correctly. Every single one. The indexes contain entity keys in a sequence specified by the index's properties and, optionally, the entity's ancestors.

There are two types of indexes, built-in and composite.

Built-in

By default, Cloud Datastore automatically predefines an index for each property of each entity kind. These single property indexes are suitable for simple types of queries. These indexes are free and do not count against your index limit.

Composite

Composite index multiple property values per indexed entity. Composite indexes support complex queries and are defined in an index.yaml file.

Composite indexes are required for queries of the following form:

queries with ancestor and inequality filters
queries with one or more inequality filters on a property and one or more equality filters on other properties
queries with a sort order on keys in descending order
queries with multiple sort orders
queries with one or more filters and one or more sort orders

NOTE: Inequality filters are LESS_THAN, LESS_THAN_OR_EQUAL, GREATER_THAN, GREATER_THAN_OR_EQUAL.

Google has excellent doc regarding datastore indexes here.

The datastore emulator generates composite indexes in an index.yaml file automatically. The file can be found in /tmp/local_datastore/WEB-INF/index.yaml. If your localhost Rails app exercises every possible query the application will issue, using every combination of filter and sort order, the generated entries will represent your complete set of indexes.

One thing to note is that the datastore emulator caches indexes. As you add and modify application code you might find that the local datastore index.yaml contains indexes that are no longer needed. In this scenario try deleting the index.yaml and restarting the emulator. Navigate through your Rails app and the index.yaml will be built from scratch.

Datastore Emulator

Install the Google Cloud SDK.

$ curl https://sdk.cloud.google.com | bash

You can check the version of the SDK and the components installed with:

$ gcloud components list

Install the Cloud Datastore Emulator, which provides local emulation of the production Cloud Datastore environment and the gRPC API. However, you'll need to do a small amount of configuration before running the application against the emulator, see here.

$ gcloud components install cloud-datastore-emulator

Add the following line to your ~/.bash_profile:

export PATH="~/google-cloud-sdk/platform/cloud-datastore-emulator:$PATH"

Restart your shell:

exec -l $SHELL

To create the local development datastore execute the following from the root of the project:

$ cloud_datastore_emulator create tmp/local_datastore

To create the local test datastore execute the following from the root of the project:

$ cloud_datastore_emulator create tmp/test_datastore

To start the local Cloud Datastore emulator:

$ cloud_datastore_emulator start --port=8180 tmp/local_datastore

Example Rails App

There is an example Rails 5 app in the test directory here.

$ bundle
$ cloud_datastore_emulator create tmp/local_datastore
$ cloud_datastore_emulator create tmp/test_datastore
$ ./start-local-datastore.sh
$ rails s

Navigate to http://localhost:3000.

CarrierWave File Uploads

Active Model Datastore has built in support for CarrierWave which is a simple and extremely flexible way to upload files from Rails applications. You can use different stores, including filesystem and cloud storage such as Google Cloud Storage or AWS.

Simply require active_model/datastore/carrier_wave_uploader and extend your model with the CarrierWaveUploader (after including ActiveModel::Datastore). Follow the CarrierWave instructions for generating an uploader.

In this example it will be something like:

rails generate uploader ProfileImage

Define an attribute on the model for your file(s). You can then mount the uploaders using mount_uploader (single file) or mount_uploaders (array of files). Don't forget to add the new attribute to entity_properties and whitelist the attribute in the controller if using strong parameters.

require 'active_model/datastore/carrier_wave_uploader'

class User
  include ActiveModel::Datastore
  extend CarrierWaveUploader

  attr_accessor :email, :enabled, :name, :profile_image, :role
  
  mount_uploader :profile_image, ProfileImageUploader

  def entity_properties
    %w[email enabled name profile_image role]
  end
end

You will want to add something like this to your Rails form:

<%= form.file_field :profile_image %>

Track Changes

TODO: document the change tracking implementation.

Nested Forms

Adds support for nested attributes to ActiveModel. Heavily inspired by Rails ActiveRecord::NestedAttributes.

Nested attributes allow you to save attributes on associated records along with the parent. It's used in conjunction with fields_for to build the nested form elements.

See Rails ActionView::Helpers::FormHelper::fields_for for more info.

NOTE: Unlike ActiveRecord, the way that the relationship is modeled between the parent and child is not enforced. With NoSQL the relationship could be defined by any attribute, or with denormalization exist within the same entity. This library provides a way for the objects to be associated yet saved to the datastore in any way that you choose.

You enable nested attributes by defining an :attr_accessor on the parent with the pluralized name of the child model.

Nesting also requires that a <association_name>_attributes= writer method is defined in your parent model. If an object with an association is instantiated with a params hash, and that hash has a key for the association, Rails will call the <association_name>_attributes= method on that object. Within the writer method call assign_nested_attributes, passing in the association name and attributes.

Let's say we have a parent Recipe with Ingredient children.

Start by defining within the Recipe model:

an attr_accessor of :ingredients
a writer method named ingredients_attributes=
the validates_associated method can be used to validate the nested objects

Example:

class Recipe
  attr_accessor :ingredients
  validates :ingredients, presence: true
  validates_associated :ingredients

  def ingredients_attributes=(attributes)
    assign_nested_attributes(:ingredients, attributes)
  end
end

You may also set a :reject_if proc to silently ignore any new record hashes if they fail to pass your criteria. For example:

class Recipe
 def ingredients_attributes=(attributes)
   reject_proc = proc { |attributes| attributes['name'].blank? }
   assign_nested_attributes(:ingredients, attributes, reject_if: reject_proc)
 end
end

Alternatively,:reject_if also accepts a symbol for using methods:

class Recipe
  def ingredients_attributes=(attributes)
    reject_proc = proc { |attributes| attributes['name'].blank? }
    assign_nested_attributes(:ingredients, attributes, reject_if: reject_recipes)
  end

  def reject_recipes(attributes)
    attributes['name'].blank?
  end
end

Within the parent model valid? will validate the parent and associated children and nested_models will return the child objects. If the nested form submitted params contained a truthy _destroy key, the appropriate nested_models will have marked_for_destruction set to True.

Datastore Gotchas

Ordering of query results is undefined when no sort order is specified.

When a query does not specify a sort order, the results are returned in the order they are retrieved. As Cloud Datastore implementation evolves (or if a project's indexes change), this order may change. Therefore, if your application requires its query results in a particular order, be sure to specify that sort order explicitly in the query.

Agrimatics / activemodel-datastore