How to test class method using rspec - ruby-on-rails

Hi i am working on RoR project with ruby-2.5.0 and rails 5.0. I have a model forgot_password where a class method is defined to create a record as follows:-
forgot_password.rb
# frozen_string_literal: true
class ForgotPassword < ApplicationRecord
before_create :create_token
def self.create_record
self.create!(expiry: Time.zone.now +
ENV['VALIDITY_PERIOD'].to_i.hours)
end
private
def create_token
self.token = SecureRandom.urlsafe_base64(nil, false)
end
end
I want to write unit testing for it using stub or factory_girl gem.
spec/models/forgot_password_spec.rb
# frozen_string_literal: true
require 'rails_helper'
describe ForgotPassword do
let(:forgot_password) do
described_class.new()
end
describe 'create_record' do
context 'with forgot_password class' do
subject { forgot_password.create_record.class }
it { is_expected.to eq ForgotPassword }
end
end
end
But its throwing error undefined method create_record for #<ForgotPassword:0x000000000622bc98> Please help me how can i test my model. Thanks in advance.

What you have written is a factory method (a class method that returns an instance) you should call it and write expectations about the instance returned:
describe ForgotPassword do
describe ".create_record" do
subject { described_class.create_record! }
it { is_expected.to be_an_instance_of(described_class) }
it "sets the expiry time to a time in the future" do
expect(subject.expiry > Time.now).to be_truthy
end
end
end
However if what you really are looking to do is set a computed default value then there is a much less clunky way:
class ForgotPassword < ApplicationRecord
after_initialize :set_expiry!
private
def set_expiry!
self.expiry(expiry: Time.zone.now).advance(hours: ENV['VALIDITY_PERIOD'].to_i)
end
end
Or with Rails 5:
class ForgotPassword < ApplicationRecord
attribute :expiry, :datetime,
->{ Time.zone.now.advance(hours: ENV['VALIDITY_PERIOD'].to_i) }
end
You can test it by:
describe ForgotPassword do
let(:forgot_password){ described_class.new }
it "has a default expiry" do
expect(forgot_password.expiry > Time.now).to be_truthy
end
end

You can test against described_class directly:
require 'rails_helper'
describe ForgotPassword do
context 'with forgot_password class' do
subject { described_class }
it { is_expected.to eq ForgotPassword }
end
end

Related

Pundit::NotDefinedError: unable to find policy when moving from Pundit 0.3 to 1.0

When I am running rspec wit pundit version 1.0 on one of my project spec classes I get multiple errors which I haven't seen before. However, when I'm switching to the previous version of pundit (0.3) everything works correctly.
Up to now what I have noticed is that with newer version of pundit #error in create function is not correctly assigned (instead of error class, I get an error message string from the error class).
class ErrorsController < ApplicationController
before_action :set_execution_environment
def authorize!
authorize(#error || #errors)
end
private :authorize!
def create
#error = Error.new(error_params)
authorize!
end
def error_params
params[:error].permit(:message, :submission_id).merge(execution_environment_id: #execution_environment.id)
end
private :error_params
in spec/factories:
FactoryGirl.define do
factory :error, class: Error do
association :execution_environment, factory: :ruby
message "exercise.rb:4:in `<main>': undefined local variable or method `foo' for main:Object (NameError)"
end
end
in spec/controllers/error_controller.rb:
describe 'POST #create' do
context 'with a valid error' do
let(:request) { proc { post :create, execution_environment_id: FactoryGirl.build(:error).execution_environment.id, error: FactoryGirl.attributes_for(:error), format: :json } }
context 'when a hint can be matched' do
let(:hint) { FactoryGirl.build(:ruby_syntax_error).message }
before(:each) do
expect_any_instance_of(Whistleblower).to receive(:generate_hint).and_return(hint)
request.call
end
expect_assigns(execution_environment: :execution_environment)
it 'does not create the error' do
allow_any_instance_of(Whistleblower).to receive(:generate_hint).and_return(hint)
expect { request.call }.not_to change(Error, :count)
end
it 'returns the hint' do
expect(response.body).to eq({hint: hint}.to_json)
end
expect_json
expect_status(200)
end
context 'when no hint can be matched' do
before(:each) do
expect_any_instance_of(Whistleblower).to receive(:generate_hint).and_return(nil)
request.call
end
expect_assigns(execution_environment: :execution_environment)
it 'creates the error' do
allow_any_instance_of(Whistleblower).to receive(:generate_hint)
expect { request.call }.to change(Error, :count).by(1)
end
expect_json
expect_status(201)
end
end
I get the error message
Pundit::NotDefinedError:
unable to find policy Pundit::ErrorPolicy for #<Pundit::Error: {"message"=>"exercise.rb:4:in': undefined
local variable or method foo' for main:Object (NameError)",
"execution_environment_id"=>1}>
since error class is not correctly created. After that every test in error class fail.
My policies:
class AdminOrAuthorPolicy < ApplicationPolicy
[:create?, :index?, :new?].each do |action|
define_method(action) { #user.internal_user? }
end
[:destroy?, :edit?, :show?, :update?].each do |action|
define_method(action) { admin? || author? }
end
end
class ErrorPolicy < AdminOrAuthorPolicy
def author?
#user == #record.execution_environment.author
end
end
I have no such an problem with any other class.
I've been dealing with the same problem for the last half hour, albeit using minitest, and the solution was to run spring stop and then rerun my tests. Hope this helps.

any_instance is not instance-agnostic

I have a problem with testing one of my workers in rails app. It looks like this:
class UserStatisticsWorker
include Sidekiq::Worker
include Sidetiq::Schedulable
def perform(administration_id = nil)
administrations(administration_id).find_each do |administration|
User::StatisticsCalculator.new.recalculate_if_needed(administration.id)
end
end
private
def administrations(administration_id = nil)
administration_id.present? ? Administration.where(id: administration_id) : Administration.all
end
end
And it is tested with rspec:
require 'spec_helper'
describe UserStatisticsWorker do
describe 'perform' do
let!(:administration) { create(:administration) }
let!(:administration_2) { create(:administration) }
context 'when administration_id is present' do
it 'runs User::StatisticsCalculator for one administration' do
expect_any_instance_of(User::StatisticsCalculator).to receive(:recalculate_if_needed).once
subject.perform(administration.id)
end
end
context 'when administration_id is not present' do
it 'runs User::StatisticsCalculator for all administrations' do
expect_any_instance_of(User::StatisticsCalculator).to receive(:recalculate_if_needed).twice
subject.perform
end
end
end
end
The second spec is not pass with following error:
The message 'recalculate_if_needed' was received by #<User::StatisticsCalculator:85721520 > but has already been received by #<User::StatisticsCalculator:0x0000000a383498>
Why is that?
A very good practice is to avoid any_instance_of and instead extract private methods in your worker which can be more easily tested. A refactor would look something like this:
class UserStatisticsWorker
include Sidekiq::Worker
include Sidetiq::Schedulable
def perform(administration_id = nil)
administrations(administration_id).find_each do |administration|
recalculate_if_needed(administration)
end
end
private
def recalculate_if_needed(administration)
User::StatisticsCalculator.new.recalculate_if_needed(administration.id)
end
def administrations(administration_id = nil)
administration_id.present? ? Administration.where(id: administration_id) : Administration.all
end
end
Then, test it like this:
require 'spec_helper'
describe UserStatisticsWorker do
describe 'perform' do
let!(:administration) { create(:administration) }
let!(:other_administration) { create(:administration) }
context 'when administration_id is present' do
it 'tries to recalculate for the specific administration' do
expect(subject).to receive(:recalculate_if_needed).once
subject.perform(administration.id)
end
end
context 'when administration_id is not present' do
it 'tries to recalculate for all administrations' do
expect(subject).to receive(:recalculate_if_needed).twice
subject.perform
end
end
end
end
The problem is that you have set the expectation to happen twice on an instance... but what is actually happening is that it is being called Once on two different instances.
ie this is not the expectation that you're looking for...
see the other answer for what you could try instead.

Rails & RSpec - Testing Concerns class methods

I have the following (simplified) Rails Concern:
module HasTerms
extend ActiveSupport::Concern
module ClassMethods
def optional_agreement
# Attributes
#----------------------------------------------------------------------------
attr_accessible :agrees_to_terms
end
def required_agreement
# Attributes
#----------------------------------------------------------------------------
attr_accessible :agrees_to_terms
# Validations
#----------------------------------------------------------------------------
validates :agrees_to_terms, :acceptance => true, :allow_nil => :false, :on => :create
end
end
end
I can't figure out a good way to test this module in RSpec however - if I just create a dummy class, I get active record errors when I try to check that the validations are working. Has anyone else faced this problem?
Check out RSpec shared examples.
This way you can write the following:
# spec/support/has_terms_tests.rb
shared_examples "has terms" do
# Your tests here
end
# spec/wherever/has_terms_spec.rb
module TestTemps
class HasTermsDouble
include ActiveModel::Validations
include HasTerms
end
end
describe HasTerms do
context "when included in a class" do
subject(:with_terms) { TestTemps::HasTermsDouble.new }
it_behaves_like "has terms"
end
end
# spec/model/contract_spec.rb
describe Contract do
it_behaves_like "has terms"
end
You could just test the module implicitly by leaving your tests in the classes that include this module. Alternatively, you can include other requisite modules in your dummy class. For instance, the validates methods in AR models are provided by ActiveModel::Validations. So, for your tests:
class DummyClass
include ActiveModel::Validations
include HasTerms
end
There may be other modules you need to bring in based on dependencies you implicitly rely on in your HasTerms module.
I was struggling with this myself and conjured up the following solution, which is much like rossta's idea but uses an anonymous class instead:
it 'validates terms' do
dummy_class = Class.new do
include ActiveModel::Validations
include HasTerms
attr_accessor :agrees_to_terms
def self.model_name
ActiveModel::Name.new(self, nil, "dummy")
end
end
dummy = dummy_class.new
dummy.should_not be_valid
end
Here is another example (using Factorygirl's "create" method" and shared_examples_for)
concern spec
#spec/support/concerns/commentable_spec
require 'spec_helper'
shared_examples_for 'commentable' do
let (:model) { create ( described_class.to_s.underscore ) }
let (:user) { create (:user) }
it 'has comments' do
expect { model.comments }.to_not raise_error
end
it 'comment method returns Comment object as association' do
model.comment(user, "description")
expect(model.comments.length).to eq(1)
end
it 'user can make multiple comments' do
model.comment(user, "description")
model.comment(user, "description")
expect(model.comments.length).to eq(2)
end
end
commentable concern
module Commentable
extend ActiveSupport::Concern
included do
has_many :comments, as: :commentable
end
def comment(user, description)
Comment.create(commentable_id: self.id,
commentable_type: self.class.name,
user_id: user.id,
description: description
)
end
end
and restraunt_spec may look something like this (I'm not Rspec guru so don't think that my way of writing specs is good - the most important thing is at the beginning):
require 'rails_helper'
RSpec.describe Restraunt, type: :model do
it_behaves_like 'commentable'
describe 'with valid data' do
let (:restraunt) { create(:restraunt) }
it 'has valid factory' do
expect(restraunt).to be_valid
end
it 'has many comments' do
expect { restraunt.comments }.to_not raise_error
end
end
describe 'with invalid data' do
it 'is invalid without a name' do
restraunt = build(:restraunt, name: nil)
restraunt.save
expect(restraunt.errors[:name].length).to eq(1)
end
it 'is invalid without description' do
restraunt = build(:restraunt, description: nil)
restraunt.save
expect(restraunt.errors[:description].length).to eq(1)
end
it 'is invalid without location' do
restraunt = build(:restraunt, location: nil)
restraunt.save
expect(restraunt.errors[:location].length).to eq(1)
end
it 'does not allow duplicated name' do
restraunt = create(:restraunt, name: 'test_name')
restraunt2 = build(:restraunt, name: 'test_name')
restraunt2.save
expect(restraunt2.errors[:name].length).to eq(1)
end
end
end
Building on Aaron K's excellent answer here, there are some nice tricks you can use with described_class that RSpec provides to make your methods ubiquitous and make factories work for you. Here's a snippet of a shared example I recently made for an application:
shared_examples 'token authenticatable' do
describe '.find_by_authentication_token' do
context 'valid token' do
it 'finds correct user' do
class_symbol = described_class.name.underscore
item = create(class_symbol, :authentication_token)
create(class_symbol, :authentication_token)
item_found = described_class.find_by_authentication_token(
item.authentication_token
)
expect(item_found).to eq item
end
end
context 'nil token' do
it 'returns nil' do
class_symbol = described_class.name.underscore
create(class_symbol)
item_found = described_class.find_by_authentication_token(nil)
expect(item_found).to be_nil
end
end
end
end

Rspec and testing instance methods

Here is my rspec file:
require 'spec_helper'
describe Classroom, focus: true do
describe "associations" do
it { should belong_to(:user) }
end
describe "validations" do
it { should validate_presence_of(:user) }
end
describe "instance methods" do
describe "archive!" do
before(:each) do
#classroom = build_stubbed(:classroom)
end
context "when a classroom is active" do
it "should mark classroom as inactive" do
#classroom.archive!
#classroom.active.should_be == false
end
end
end
end
end
Here is my Classroom Factory:
FactoryGirl.define do
factory :classroom do
name "Hello World"
active true
trait :archive do
active false
end
end
end
When the instance method test runs above, I receive the following error: stubbed models are not allowed to access the database
I understand why this is happening (but my lack of test knowledge/being a newb to testing) but can't figure out how to stub out the model so that it doesn't hit the database
Working Rspec Tests:
require 'spec_helper'
describe Classroom, focus: true do
let(:classroom) { build(:classroom) }
describe "associations" do
it { should belong_to(:user) }
end
describe "validations" do
it { should validate_presence_of(:user) }
end
describe "instance methods" do
describe "archive!" do
context "when a classroom is active" do
it "should mark classroom as inactive" do
classroom.archive!
classroom.active == false
end
end
end
end
end
Your archive! method is trying to save the model to the database. And since you created it as a stubbed model, it doesn't know how to do this. You have 2 possible solutions for this:
Change your method to archive, don't save it to the database, and call that method in your spec instead.
Don't use a stubbed model in your test.
Thoughtbot provides a good example of stubbing dependencies here. The subject under test (OrderProcessor) is a bona fide object, while the items passed through it are stubbed for efficiency.

Using Shoulda to test for an object expecting a method

In the following example the #user.expects(:send_invitations!).once assertion is failing, even though the invitations are being sent by the app and the #send_invitations variable is being assigned. Would you expect #user.send_invitations! to be invoked at this point or is #user.expects(:send_invitations!) being used incorrectly?
The Controller
class RegistrationsController < ApplicationController
before_filter :require_active_user
def welcome
if params[:send_invitations]
current_user.send_invitations!
#send_invitations = true
end
end
end
The Controller Test
require 'test_helper'
class RegistrationsControllerTest < ActionController::TestCase
context "With a logged-in user" do
setup { login_as #user = Factory(:user) }
context ":welcome" do
context "with params[:send_invitations] present" do
setup { get :welcome, { :send_invitations => true } }
should "send invitations" do
#user.expects(:send_invitations!).once # tests say this isn't being invoked
assert assigns(:send_invitations) # but tests say this IS being assigned
end
end
context "without the presence of params[:send_invitations]" do
setup { get :welcome }
should "not send invitations" do
#user.expects(:send_invitations!).never # passes fine
assert !assigns(:send_invitations!) # also passes fine
end
end
end
end
end

Resources