Super Spread Sheet S³

Or little computing tricks and hacks

Category Archives: Testing

Rails: Different storage for carrierwave assets depending on the environment

I am using carrierwave for storing images, one (or more) per field of a table in a database of the my Rails app. Given that heroku does not allow assets to be uploaded by the user, or dynamically for that matter, I had to use storage in the cloud. Mind you, not a bad thing. I chose AWS, s3 services to store my bucket.

Carrierwave has an extensive explanation on setting up the right configuration variables to allow the storage on the cloud. (Carrierwave in github)

The first problem I had was running the development and staging (or production) sites at the same time: Updates in one were seen in the other in the following case:

  1. Create a field with id=1 in production. The image associated with it, is stored in the cloud with id=1.
  2. Create a field with id=1 in development. the image associated with it, overrides in the cloud, the previously created image.

My solution to solve this problem was to create a second bucket, and add in the file


the setting as follows:

# config/initialize/carrierwave.rb
  if Rails.env.production?
    config.fog_directory = 'bucket-0'
  elsif Rails.env.development?
    config.fog_directory = 'bucket-1'

The next problem was that the test suite run was affecting the development data in the same way as described before. In fact, I wanted the images for the test suites simply to be stored locally and then deleted. The first approach was to include in the same file as above the following:

# config/initialize/carrierwave.rb
  if Rails.env.test? || Rails.env.cucumber? = :file
    config.enable_processing = false
    config.root = "#{Rails.root}/tmp"
  else = :fog

  config.cache_dir = "#{Rails.root}/tmp/uploads"

And to delete after finishing the test, add to the file:


the following:

# spec/spec_helper.rb
RSpec.configure do |config|
  config.after(:all) do
    if Rails.env.test?

That only seemed to work. I think that the display was using cache data. Only recently did I notice the problem when reworking some aspects of the app. The images in development were being overwritten by the test suite.

I had to add the following to the uploader file

# app/uploaders/image_uploader.rb
  if Rails.env.test?
    storage :file
    storage :fog

Et voilá!

Testing with RSpec in Rails, Part 2, Models

(Edited 26 February 2015, added information about accessing database details.)

Following from the post Testing with RSpec in Rails, Part 1, Introduction, it is now time to expand on testing models based on my experience.

I considered three types of test relevant for my application:

  • Factory tests
  • Data fields validations
  • Associations between models

Factory test

The first test is to make sure that a valid record can be created safely and respecting all of the constraints. In other words, that is has a valid factory.

A factory is an object for creating other objects – formally a factory is simply an object that returns an object from some method call, which is assumed to be “new”.

In my opinion, this is not a TDD type test, rather, it follows the database design. Therefore, the table might exist already. This is what I am assuming for the rest of this post.

If the table exists, we need tools to find the table structure in the development database. There are two easy ways that I know of using the rails command. The first is using the console

rails c

short for

rails console

The command

$ rails c
Loading development environment (Rails 4.0.4)
2.0.0-p451 :001 > ActiveRecord::Base.connection.tables
 => ["schema_migrations", "users", ...]

will output an array including all of the current tables. And the command User.column_names

2.0.0-p451 :008 > User.column_names
 => ["id", "name", "email", ...]

will output an array including the column names of the table User.

The second way of accessing this information is by using the command

rails db

short for

rails dbconsole

It starts a console for the database and database adapter specified in config/database.yml depending on the current Rails environment. If testing, one is most likely in the development environment.

$ rails db
psql (9.4.1, server 9.3.5)
Type "help" for help.


To retrieve the fields of the articles table, I used the following command:

=# select column_name from information_schema.columns where table_name='articles';

The output is:

(6 rows)

The command

App_Name_developemnt=# \d

will display a list of all the relations.

With this information and with the information in the model.rb file, the factory can be written for the model.

I used the faker gem to create random names and descriptions ( Given that authors write articles, I also have a factory for articles. The factories look like this:

# spec/factories.rb
require 'faker'

FactoryGirl.define do

  factory :author do
    bio { Faker::Lorem.sentences.to_s }
    image {, 'spec',
      'support', 'images', 'image_2.jpg')) }

  factory :article do
    title "The life of Pepito Perez"
    summary { Faker::Lorem.sentence }

The image field has been created with the carrierwave gem and that is the code needed in the testing environment. The file image_2.jpg must exist in the spec/support/images/ directory.

I did not know what was best practice between having one file called spec/factories.rb or having a file model_name.rb in the spec/factories directory. My searches did not deliver any conclusive results. So I used trial and error. I first opted for one file, then a mix, and then multiple files in the directory, as the number of factories grew, one per model. This is in-line with Rails practice of always having small files.

The actual test to validate the factory is written in file spec/models/article_spec.rb:

require 'rails_helper'

describe Article do
  # Validation tests
  it "has a valid factory" do
    expect(FactoryGirl.create(:article)).to be_valid

The RSpec matcher be_valid verifies that our factory does indeed return a valid object.

Given that I had already created my models, once the test was passing, I changed the code to see the test fail, making sure that specific parts of code were indeed being tested.

Data validations

These tests are straight-forward, validating any of the constraints needed in each field. The model of the article has the constraint that the title is mandatory, as shown below:

# app/models/article.rb
class Article < ActiveRecord::Base
  validates :title, presence: true

The following lines, create a object in the test environment, and gives it an empty title, which is not allowed:

require 'rails_helper'

describe Article do
  it "is invalid without a title" do
    expect(, title: nil)).not_to be_valid

Notice in the article_spec.rb file there are two special methods: create and build.
create builds and saves the object, while build only does that. This allows the modification of attributes before saving the object. The not_to verifies that an empty title should not be allowed in a valid object.

Associations between models

An interesting test that I discovered in this post, is testing the associations between models using RSpec. As a beginner, I always want to double check that I made the right associations. For this, I used the gem shoulda found in github.

Add the gem to the Gemfile:

group :test do
  gem 'shoulda-matchers'

and run bundle.

Then add to the spec/models/article_spec.rb, the following test:

require 'rails_helper'

describe Article do

  # Associations test
  it { should belong_to(:author) }

Testing with RSpec in Rails, Part 1, Introduction

Testing in Rails is both hard and easy. Hard because it takes as much time as coding does, but easy because there are many tools that make the most complicated things straightforward. There are so many tools though, that sometimes it is hard to choose. Neat paradox.

When setting up testing for my app, I did not find in one place a comprehensive explanation of the Rails testing process as a whole. This is an attempt to fix the situation.

Why testing?

The issue is more what and when to test. Michael Hartl has an introduction to testing in his Ruby on Rails tutorial, which describes why testing and the kinds of tests he is likely to do. In summary,

  • first test the controllers and the models, in other words unit testing;
  • second test the functionality across models, views and controllers, the integration tests;
  • and third, the views, but if they are likely to change, they can be skipped.

He also mentions the importance of writing regression tests on bugs found and having tests in place before any refactoring.

RSpec testing infrastructure

Set up

When a new Rails project is created with default settings, a test directory is created, coupled to work with minitest. From the word GO, I started using RSpec (, since I used the Ruby on Rails tutorial mentioned above which used RSpec as testing framework. I believed it has changed since.

I did not want that default directory, as I wanted the RSpec set up.

By issuing the command

rails g controller StaticPages about --no-test-framework

the test files related to the StaticPages controller will not be created. To create the right files, RSpec must be installed. That is done by including its gem in the Gemfile. RSpec takes advantage of a series of helpers to run tests automatically. The gems are specified following:

group :development, :test do
  gem 'rspec-rails'
  gem 'factory_girl_rails'

group :test do
  gem 'selenium-webdriver', '2.35.1'
  gem 'faker'
  gem 'capybara', '>= 2.2.0'
  gem 'guard-rspec'
  gem 'launchy'

and then run bundle.

In my opinion, the most interesting about Rails testing, is the interaction with databases and RSpec plus helpers makes this easy.

By default, every Rails application has three environments: development, test, and production. The database for each one of them is configured in config/database.yml.

That fact is simply brilliant, as the development database is NOT the same as the test database. One can create automatically hundreds of records to test for specific features in isolation.

To run any testing, the databases need to be created, so run

rake db:create:all

which will create the databases which do not exist and inform you of the ones already created. The information on the databases description is taken from the file config/database.yml, which should include the information about test, development and production databases. Make sure that the three are name differently!

Then run

rails generate rspec:install

which generates the following configuration files:


All tests and helpers will reside in the spec directory. This is directory where RSpec searches for the tests to run.

NOTE: After making changes to any of the models in development, you have to migrate the changes to the test database as well by running

rake db:migrate RAILS_ENV=test

With this, the testing infrastructure is set up.


RSpec uses mainly the words “describe” and “it” so we can express concepts like a conversation:

“Describe an order.”
“It sums the prices of its line items.”

The describe method creates an ExampleGroup. Within the block passed to describe you can declare examples using the it method. Under the hood, an example group is a class in which the block passed to describe is evaluated.

The broad syntax of the test is as follows:

describe Object do
  it "Descriptive message of the test" do
    code with expectations

Each “it” line only expects one example. Best practice is to test one thing at a time to make it simple to find errors. Although the descriptive message is technically optional, omitting defeat the purpose of individual testing. Previous RSpec examples had the “should” beginning the message, however that just clutters the output. A direct verb suffices.

Actual testing

Now to writing the tests. But where to start? With how to run a test.

To run a test, use the command


from the root directory of the app. If used alone, it will run all the tests found in the spec directory. You can also specify a directory or a filename including its path with respect to root. RSpec will run all tests found in that directory in the first case, or just the file specified in the second.

The testing framework automatically creates directories to sort out the tests. My spec directory looks like this:

controllers/  factories/       models/         requests/       support/
helpers/      rails_helper.rb  spec_helper.rb  views/

Next is what to test: unit testing, integrations testing, views, regression testing.

Unit testing: Models

Models are the building blocks of the application. They are also easier to test since their behaviour should be well defined in any application. I considered them to be first priority to test.

Everyday Rails

The blog Everyday Rails considers the following to be essential model tests:

  • the factory should generate a valid object
  • data validation
  • class and instance method

Full text deployed in a post called Testing with RSpec in Rails, Part 2, Models.

Unit testing: Controllers

Post in development.

Integration tests

Post in conception.
This should test the functionality across models, views and controllers.

View’s tests?

Will I create a post?
Code in the views tend to change often, so there are different schools of thought on whether to test or not.

Regression tests

Automatic test runs with Guard

Guard runs the test suite upon the detection of a modification of file in the spec directory or as specified in the Guardfile. It also sets the testing environment just once, speeding up the running of subsequent tests. To set up (gem already included in the Gemfile) run:

bundle exec guard init

which creates the Guardfile describing how and when Guard is to run. In his tutorial, Guard: Michael Hartl’s Rails tutorial, he explains the set up in more detail, although using minitest instead of RSpec.

To start guard just type in a terminal

$ guard

It will create a shell and guard will start listening to any changes in the spec directory or any other file specified in the Guardfile. The Guarfile created in the set up is a very good starting point.

Type enter in the shell to run all the tests in the spec directory. Ctl-D to exit.

(Explanation on Guard needs expansion.)