Rails Design: Protected Modules with Injection

Long-term Manageable Rails through Low Dependency/Entanglement and Loose Coupling

Rails Design - Modules with Boundaries and Injection

Recently, a colleague Severin Räz suggested that we could be using module to create namespaces (to organize our code with high cohesion) and private_constant to enforce API boundaries (enforce low coupling). Based on that I ran some experiments and wrote about them in Rails with Protected Modules and Rails with Minimal Engines)

Similarly, another colleague Jörg Jenni has been encouraging dependency injection to ensure loose coupling, making it easy to replace a service with another and simplifying testing. Dependency-Inversion is an alternative to hardcoding classes into code.

This article builds on the previous articles Rails with Protected Modules and Rails with Mingines, but is not required for this article to make sense.

In any case, this article describes what we (after discussions and a few code iterations) are using as a starting point for maintainable design.

App Description

Let’s assume your application creates valuable documents that need to be archived. We will keep this simple and design a module that can send or retrieve a document to and from an Archive Server. I will use the open source Teedy Server - which can be self-hosted with docker. The API is straight forward and easy to use.

Teedy will be our default adapter, however, we want to make it easy for our app to work with any Archive Server.

Plan our approach

  • We will start by building a simple module to act as an API wrapper.
  • we will build a basic Teedy adapter (feel free to build other for other archive server)
  • we will then integrate it into our Rails app

Getting Started

The code for this experiment can be found at: https://github.com/btihen/protected_rails_modules

Just for fun we will use the newest Rails and Ruby versions, but these techniques should work on any version of Rails.

# install ruby
rbenv install 3.2.2
# or next preview
rbenv install 3.3.0-preview1

# best for production
rbenv local 3.2.2
# fun for learning
rbenv local 3.3.0-preview1

# `--main` uses the rails main branch instead of a released version (--skip-test - I like rspec)
rails new archiver --main --javascript=esbuild --css=tailwind --database=postgresql --skip-test

cd archiver
git add .
git commit -m "initial commit"

bin/rails db:create

Later when Rails 7.1 is released we can set the version in the Gemfile with: gem ‘rails’, ‘~> 7.1’ instead of the current: gem “rails”, github: “rails/rails”, branch: “main”

Configure for Modules / Packages

Let’s start by create a modular structure for our app as described in Rails with Protected Modules - although this is not required.

I generally put everything in either app/packages, app/modules, packages or modules - but you can choose your preference. (I tend to use ‘packages; when using packwerk).

Configure Rails to find our modules:

# config/application.rb
require_relative "boot"
# ...
module ModulesDesign
  class Application < Rails::Application
    # Initialize configuration defaults for originally generated Rails version.
    config.load_defaults 7.1

    # add packages PATH BEFORE autoload config!
    # otherwise it will not find the modules with the error:
    # ActionController::RoutingError (uninitialized constant Landing)
    config.paths.add 'app/packages', glob: '*/{*,*/concerns}', eager_load: true

    config.autoload_lib(ignore: %w(assets tasks))
    # ...
    # Don't generate system test files.
    config.generators.system_tests = nil

Configure Rails to find our module views:

# app/controllers/application_controller.rb
class ApplicationController < ActionController::Base
  # find views within our modules

Move the Rails code into its own directory (in this case app/packages/rails_shim but you can choose your preference):

mkdir -p app/packages/rails_shim

# move the rails code into the rails module
mv app/* app/packages/rails_shim/.

# we need to return assets and javascript to the app directory
mv app/packages/rails_shim/assets app/assets
mv app/packages/rails_shim/javascript app/javascript

Test with bin/dev that it all still works and then commit:

git add .
git commit -m "move rails code into packages"

Landing Page

Let’s create a simple landing page to show that our app is working.

bin/rails g controller landing/home index --no-helper

move into a package

mkdir app/packages/landing
mv app/views app/packages/landing/.
mv app/controllers app/packages/landing/.

configre the homepage route

# config/routes.rb
Rails.application.routes.draw do
  namespace :landing do
    get 'home/index'

  # Defines the root path route ("/")
  root 'landing/home#index'

add to repo now that we have the basics working & tested:

git add .
git commit -m "add landing page package"

Create Blog Articles

Now that we have a modularized app working, let’s add a simple article (blog post) feature.

Let’s make an Article model within the Articles namespace (article content will use Action Text so we can capture the person’s writing) which we can then display as an HTML page or save as a PDF file.

first we will install the required tools:

bin/rails action_text:install
bin/rails db:migrate

# move the active storage code into the rails module
mv app/views/active_storage app/packages/rails_shim/views/.
mv app/views/layouts/action_text app/packages/rails_shim/views/layouts/.

Let’s generate the Articles scaffold (with a namespace to keep this separate from other features):

# content will use Action Text so we can format the writing
bin/rails g scaffold Articles::Post author:string title:string --no-helper
bin/rails db:migrate

move the new code into its module:

mkdir -p app/packages/articles
mv app/views app/packages/articles/.
mv app/models app/packages/articles/
mv app/controllers app/packages/articles/.

Now we need to add the ActionText field to the Article model:

# app/packages/articles/models/articles/post.rb
module Articles
  class Post < ApplicationRecord
    has_rich_text :content

We need to update our controller to permit the new content field:

# app/packages/articles/controllers/articles/posts_controller.rb
  # Only allow a list of trusted parameters through.
  def articles_post_params
    params.require(:articles_post).permit(:author, :title, :content)

We will also need to add <%= form.rich_text_area :content %> to the Article form :

# app/packages/articles/views/articles/posts/_form.html.erb
    <%= form.label :title, style: "display: block" %>
    <%= form.text_field :title %>

    <%= form.label :content, style: "display: block" %>
    <%= form.rich_text_area :content %>

    <%= form.submit %>
<% end %>

We also need to update our view to display the content with <%= article.content %>:

# app/packages/articles/views/articles/posts/_post.html.erb

    <%= post.title %>

  <!-- use div since rich text can generate html beyond a p tag-->
    <%= post.content %>

Now let’s test by going to:

# NOTE: with the tradionitional `bin/rails s` the Action Text (Trix) field will not display so use!
# https://stackoverflow.com/questions/73299862/action-text-rails-7-error-vanishing-field

# open our browser and add a new article with some formatting:
open http://localhost:3000/articles/posts/new

cool - let’s test and add this to our repo:

git add .
git commit -m "add Article::Post with ActionText"

Simple PDF Generation

Let’s start by installing the necessary tooling:

# `matrix` was part of ruby, but now its a separate gem (and pawn-html does not load it automatcally)
bundle add 'matrix'
bundle add 'prawn-html'
# for some reason I had to run this to fix a bundler conflict:
bundle install --gemfile Gemfile
# the error was:
# `You specified: image_processing (~> 1.2) and image_processing (>= 0). Gem already added.`

we need to extend our Post model to allow for a PDF file to be attached:

module Articles
  class Post < ApplicationRecord
    has_rich_text :content

let’s create a simple PDF generator:

# app/packages/articles/services/articles/pdf_generator.rb
require 'prawn'
require 'prawn-html'

module Articles
  class PdfGenerator
    def self.generate(model)
      pdf = Prawn::Document.new

      # Add title to the PDF
      pdf.text model.title, size: 24, style: :bold

      # Add author to the PDF
      pdf.text "by: #{model.author}", size: 16, style: :italic

      # Add model content to the PDF using PrawnHtml
      PrawnHtml.append_html(pdf, model.content.to_s)

      # Return the generated PDF

Lets add a route to generate the PDF:

# config/routes.rb
Rails.application.routes.draw do
  # ...
  namespace :articles do
    resources :posts do
      get :download_pdf, on: :member
  # ... other routes

Lets add the action to the controller (lets also nest PostController under Articles namespace):

# app/packages/articles/controllers/articles/posts_controller.rb
module Articles
  class PostsController < ApplicationController
    before_action :set_articles_post, only: %i[show edit update destroy generate_pdf download_pdf]

    def download_pdf
      rendered_pdf = PdfGenerator.generate(@articles_post)

      if rendered_pdf
        formated_time = DateTime.now.strftime('%Y%m%d_%H%M%S')
        filename = "post_#{@articles_post.id}_#{formated_time}.pdf"
        type = 'application/pdf'

        send_data(rendered_pdf, filename:, type:)
        redirect_to articles_post_url(@articles_post), alert: 'OOPS - NO PDF was generation.'
    # ... (other actions and methods)

Now add the generate PDF link to the show view with: <%= link_to 'Generate PDF', generate_pdf_articles_post_path(@articles_post), method: :get %>

# app/packages/articles/views/articles/posts/show.html.erb
<p style="color: green"><%= notice %></p>

<%= render @articles_post %>

  <%= link_to 'Download as PDF', download_pdf_articles_post_path(@articles_post), method: :get %>
  <%= link_to "Edit this post", edit_articles_post_path(@articles_post) %> |
  <%= link_to "Back to posts", articles_posts_path %>
  <%= button_to "Destroy this post", @articles_post, method: :delete %>

Create an Archive Package

Let’s start by creating a new its folder structure:

mkdir -p app/packages/archive

Define Wrapper Needs

Probably the module should handle construction (this allows us to use multiple adapters - perhaps each author is responsbile for their own archive service). And the Adapter wrapper should handle the all other aspects of the API (the interactions with the remote service).

What we need the API to do:

  1. constructor (we want to be able to use multiple remote archive services)
  2. initialize (desired Adapter from config & authenticate with remote archive service)
  3. archive_file (on success report needed information, on error report nil)
  4. retrieve_file (on success report file location/path, on error report nil)
  5. errors (report any errors)

Define Configuration

Before we look at the Constructor, let’s consider what configuration would look like.

So if we want to start with a simple mock adapter (good for testing and perhaps a service we can show to unpaid users to see what it might be like), we could do something like the following since we don’t need to authenticate, nor any other particular infomation, just the configuration name and the adapter class name:

mkdir -p app/packages/archive/config
touch app/packages/archive/config/archive.yml
vim app/packages/archive/config/archive.yml
- adapter_config: Tester
  adapter_klass: TestAdapter

Reviewing the TeedyAPI Docs and with a little experience, we can see that we need to provide the following information:

- adapter_config: DemoAuthor
  adapter_klass: TeedyAdapter
  url: http://localhost:8080
  archive_path: /archive
  username: demo
  password: password

- adapter_config: AdminAuthor
  adapter_klass: TeedyAdapter
  url: http://localhost:8080
  archive_path: /archive
  username: admin
  password: admin

Module Constructor

Let’s see if we can define the API for the module constructor.

  1. we need to find a configuration that matches the name provided
  2. we need to find the adapter class that matches the name provided
  3. we need to instantiate the adapter class with the configuration provided
  4. if needed we need to authenticate with the remote service (& fetch any necessay configs / tokens, etc)
  5. we need to return the adapter instance (or nil if there was an error)?
touch app/packages/archive/archive.rb
# app/packages/archive/archive.rb
require 'yaml'

module Archive
  def self.adapter_for(config_name, config_file = 'app/packages/archive/config/archive.yml')
    config_file_path = Rails.root.join(config_file)
    adapter_configs = YAML.load_file(config_file_path)['archive']
    vendor_config = adapter_configs.find { |config| config['adapter_config'] == config_name }
    vendor_klass = adapter_config[:adapter_klass].constantize
    vendor_adapter = vendor_klass.new(vendor_config)

Define Adapter

Let’s start by creating a simple mock adapter wrapper:

mkdir app/packages/archive/adapters
touch app/packages/archive/adapters/adapter.rb
touch app/packages/archive/adapters/test_adapter.rb
# app/packages/archive/adapters/adapter.rb
module Archive
  module Adapters
    class AdapterError < StandardError; end

    class Adapter
      attr_reader :vendor_adapter
      private     :vendor_adapter

      # we could delegate all the methods to the vendor_adapter - with:
      # `delegate :archive_file, :retrieve_file, :errors, :errors? to: :remote_adapter`
      # but I like to explicity force the API to be the same for all adapters and have
      # one location to trap and handle errors consistently for all adapters

      def initialize(vendor_adapter)
        @vendor_adapter = vendor_adapter

        raise AdapterError, vendor_adapter.errors if vendor_adapter.errors?

      # most document services have need the file_path, mime_type, and some type of metadata to describe the file conext and contents
      def archive_file(model, file_path, mime_type, options: {})
        metadata = {
          model: model.class.name,
          model_id: model.id,
          created_at: model.created_at,
          updated_at: model.updated_at
        response = vendor_adapter.archive_file(file_path, mime_type, metadata)
        ArchiveSend.create!(model:, options:, metadata:,
                            remote_key: response[:remote_key],
                            container_key: response[:container_key])
        return response if response && !response.errors?

        raise AdapterError, remote_adapter.errors

      # most document services have some type of 'container' concept (Teedy calls this a document)
      # to track multiple versions of a given file
      def retrieve_file(file_key, container_key = nil)
        response = vendor_adapter.retrieve_file(remote_keys, container_key)
        return response.to_s if response

        raise AdapterError, vendor_adapter.errors

    private_constant :Adapter

Define Mock Adapter

# app/packages/archive/adapters/test_adapter.rb
require 'prawn-html'

module Archive
  module Adapters
    class TestAdapter
      attr_reader :errors
      private     :errors

      def initialize(_config)
        @errors = []

      def archive_file(file_path, mime_type, metadata)

      def retrieve_file(remote_keys, container_key)
        file_path = Rails.root.join('tmp', 'test.pdf')
        pdf = Prawn::Document.new(page_size: 'A4')
        PrawnHtml.append_html(pdf, '<h1 style="text-align: center">Just a test</h1>')
        pdf.render_file(rails.root.join('tmp', 'test.pdf')

      def errors? = errors.any?

REAL Implimentations

Now you can make any adapter that implement archive_file(model, file_path, mime_type, options: {}), retrieve_file(remote_keys, container_key), .errors? and errors

You now have a blox module with a consistent and clearly defined API & error handling and the rest of your app accesses it via the module methods!

More upfront work, but prrevents coupling and makes it easy to add new adapters and enforces consistency.

Experience - Standard vs Modular

My colleague Jörg Jenni - created a little app messy vs modular - that need needs new features. It has 2 base Apps messy and modulariyed. Its a great way to experience the benefits of modular design.

My solution can be seen at: https://github.com/btihen/coupled_vs_modular


Other supporting resources\

Bill Tihen
Bill Tihen
Developer, Data Enthusiast, Educator and Nature’s Friend

very curious – known to explore knownledge and nature