I'm trying to test a controller to ensure that only an authorized party can view the correct child object using RSpec. I cant figure out what I'm doing wrong as I'm getting this error:
ActiveRecord::RecordInvalid: Validation failed: Company can't be blank
I have a Plan object and a Company object. The Store can have many plans (think of a pest control Company). I want to test that given a known scenario I can retrieve the plan fo the Company (assuming there is only one).
The Plan looks like this:
class Plan < ActiveRecord::Base
before_save :default_values
# Validation
validates :amount, :presence => true
validates :company, :presence => true
# Plans belong to a particular company.
belongs_to :company, :autosave => true
scope :find_all_plans_for_company, lambda {
|company| where(:company_id => company.id)
}
# Other code ...
end
The Company looks like this:
class Company < ActiveRecord::Base
validates :name, :presence => true
validates :phone1, :presence => true
validates_format_of :phone1, :phone2,
:with => /^[\(\)0-9\- \+\.]{10,20}$/,
:message => "Invalid phone number, must be 10 digits. e.g. - 415-555-1212",
:allow_blank => true,
:allow_nil => true
has_many :users
has_many :plans
end
.. controller looks like this
def index
#plans = Plan.find_all_plans_for_company(current_user.company)
respond_to do |format|
format.html # index.html.erb
format.json { render json: #plans }
end
end
.. and my RSpec test looks like this (excuse me if its full of gimmickery, I'm just splunking around with it and cannot get it to work).
describe PlansController do
def valid_attributes
{
:company_id => 1,
:amount => 1000
}
end
describe "GET index" do
it "should return the Plans for which this users company has" do
#company = mock_model(Company, :id => 1, :name => "Test Company", :phone1 => "555-121-1212")
Company.stub(:find).with(#company.id).and_return(#company)
controller.stub_chain(:current_user, :company).and_return(#company)
plan = Plan.create! valid_attributes
get :index, {}
assigns(:plans).should eq([plan])
end
# Other tests ...
end
end
The problem is, when I try this (or any of the crazy other variants I've tried) I get this error:
ActiveRecord::RecordInvalid: Validation failed: Company can't be blank
I'm not sure why this is happening as I thought the Company.stub call would handle this for me. But apparently not.
What am I missing here and what am I doing wrong? How can I get this test to pass?
Let's peel back the layers on this spec, to make sure things make sense (and to make sure I understand what's going on). First, what are you testing?
it "should return the Plans for which this users company has" do
...
assigns(:plans).should eq([plan])
So you want to check that the plans associated with the company of the current user are assigned to #plans. We can stub or mock out everything else.
Looking at the controller code, we have:
def index
#plans = Plan.find_all_plans_for_company(current_user.company)
What do we need to get this to work, without hitting the database and without depending on the models?
First of all, we want to get a mock company out of current_user.company. This is what these two lines in your spec code do:
#company = mock_model(Company, :id => 1, :name => "Test Company", :phone1 => "555-121-1212")
controller.stub_chain(:current_user, :company).and_return(#company)
This will cause current_user.company to return the mock model #company. So far so good.
Now to the class method find_all_plans_for_company. This is where I'm a bit confused. In your spec, you stub the find method on Company to return #company for id = 1.
But really, wouldn't it suffice just to do something like this in your controller code?:
#plans = current_user.company.plans
If you did it this way, then in your test you could just mock a plan, and then return it as the plans association for your mock company:
#plan = mock_model(Plan)
#company = mock_model(Company, :plans => [ #plan ])
controller.stub_chain(:current_user, :company).and_return(#company)
Then the assignment should work, and you don't need to actually create any model or hit the database. You don't even need to give your mock company an id or any other attributes, which anyway are irrelevant to the spec.
Maybe I'm missing something here, if so please let me know.
Why do you need to mock?
My standard testing setup is to use Database Cleaner which clears out the database from any records created during tests. In this way, the tests are run with real database records which are consequently deleted from the test database after each test.
You might also like taking a look at Factory Girl for creating instances of your models during testing (makes it easy to create 10 company records, for example).
See:
http://rubygems.org/gems/database_cleaner
http://rubygems.org/gems/factory_girl
I have three thoughts coming up that could resolve your issue:
Try adding attr_accessible :company_id to Plan class.
Because mock_model does not actually save to the database when you create a Plan with company_id of 1 it fails validation since it is not present in the database.
Ensure before_save :default_values in Plan class does not mess with company_id attribute of the newly created instance.
Related
I still would consider myself new to Rails. I'm implementing a SMS feature in Rails app that reminds clients of their upcoming appointments. My question is, I have the SMS method in my appointment model, but my client model is where the phone attribute is located. How do I call my phone attribute from the appointment model.
Here is my appointment model
class Appointment < ApplicationRecord
enum status: { confirmed: 0, rescheduled: 1, cancelled: 2}
belongs_to :user
belongs_to :client
validates :start_time, presence: true
validates :end_time, presence: true
after_create :reminder
def reminder
#twilio_number = ENV['TWILIO_NUMBER']
account_sid = ENV['TWILIO_ACCOUNT_SID']
#client = Twilio::REST:Client.new account_sid, ENV['TWILIO_AUTH_TOKEN']
time_str = ((self.start_time).localtime).strftime("%I:%M%p on %b. %d, %Y")
reminder = "Hi #{client.name}. Just a reminder that you have an appointment coming up at #{time_str}."
message = #client.api.account(account_sid).messages.create(
:from => #twilio_number,
:to => client.phone_number,
:body => reminder,
)
end
My client model
class Client < ApplicationRecord
has_many :appointments
has_many :users, through: :appointments
scope :clients_by, ->(user) { where(user_id: user.id) }
end
Based on my current associations setup. In the reminder variable couldn't I just call
reminder = "Hi #{client.name}.?
And for
:to => client.phone_number
to access the phone_number attribute?
Yes your assumption is correct, you can just call client.<attribute>.
However, be aware of the dreaded SELECT N+1 issue. So let's say you do something like
Appointment.all.each do {|a| a.reminder }
If you have 50 appointments, this results in 51 calls to the database, one call to load all the appointments and then a bunch of calls to load each client one by one.
To avoid this issue, you can make use of includes, eager_loads, or preload which all load the associated data more efficiently than individual queries.
The difference between those three methods is covered very well in this article http://blog.scoutapp.com/articles/2017/01/24/activerecord-includes-vs-joins-vs-preload-vs-eager_load-when-and-where. I've quoted a TL;DR excerpt below.
I'd roughly summarize my approach to these methods like this:
If I'm just filtering, use joins.
If I'm accessing relationships, start with includes.
If includes is slow using two separate queries, I'll use eager_load to force a single query and compare performance.
There are many edge cases when accessing relationships via ActiveRecord. Hopefully this is enough to prevent some of the more basic performance deadends when using joins, includes, preload, and eager_load.
Following the advice of that article, we'd rewrite my example as
Appointment.all.includes(:client).each do {|a| a.reminder }
I'm new to rails, and I'm currently trying to develop an API based app using Rails 5, on one of my controllers I have a function to filter the allow parameters like so
def provider_params
params.require(:provider).permit(:name, :phone, :email, :website, :address, :provider_id, :bio, :specialty_ids => [])
end
Then posting from Paw I noticed that the arguments that are not attributes of the table are no included in provider_params, the parameter I'm supposed to receive is an array, which is defined by a HABTM relation-ship.
This is how my models look like
specialty.rb
class Specialty < ApplicationRecord
has_and_belongs_to_many :providers
end
provider.rb
class Provider < ApplicationRecord
has_and_belongs_to_many :specialties
end
And this is how the join table was created via migration
class CreateProvidersSpecialties < ActiveRecord::Migration[5.0]
def change
create_table :providers_specialties, :id => false do |t|
t.integer :provider_id
t.integer :specialty_id
end
add_index :providers_specialties, :provider_id
add_index :providers_specialties, :specialty_id
end
end
The JSON I'm posting
{
"name": "the name",
"specialty_ids": [
1,
2
]
}
So as I mentioned, the array specialty_ids doesn't seem to be coming through, and even if it did, I suspect there's still something else I need to do in order for rails to insert the content of specialty_ids in the ProvidersSpecialties Table
So the problem was finally solved by removing the requir call from the method provider_params, since I wasn't wrapping the json-payload in a provider key. Apparently once you add the require(:key) call you would only be able to add parameters that belong to the Model, which is weird since an error should be raised when the key is not present, what was the case with my payload, lacking the provider key.
I am creating a ruby project with the class which is the inheritance of ActiveRecord::Base. How can i write rspec testing and simple coverage for the following code sample without using database.
class Person < ActiveRecord::Base
validates_length_of :name, within: 10..40
end
person = Person.create(:name => "aungaung")
person.save
If you don't want to touch db, FactoryGirl.build_stubbed is your friend.
> person = FactoryGirl.build_stubbed :person
> person.save!
> #=> person obj
> Person.all
> #=> [] # Not saved in db
So, to test validation
it "validates name at length" do
person = FactoryGirl.build_stubbed :person, name: "aungaung"
expect{person.save!}.to raise_error(ActiveRecord::RecordInvalid)
end
Note build_stubbed is good at model's unit testing. For anything UI related, you can't use this method and need to save to db actually.
Here's a short example of testing the validations on an ActiveRecord model. You can certainly go into much more depth, and there are plenty of ways to make the tests more elegant, but this will suffice for a first test.
describe Person do
describe "#name" do
specify { Person.new(:name => "Short").should_not be_valid }
specify { Person.new(:name => "Long" * 12).should_not be_valid }
specify { Person.new(:name => "Just Right").should be_valid }
end
end
I'm unclear on what this method actually does or when to use it.
Lets say I have these models:
Person < ...
# id, name
has_many :phone_numbers
end
PhoneNumber < ...
# id, number
belongs_to :person
validates_length_of :number, :in => 9..12
end
When I create phone numbers for a person like this:
#person = Person.find(1)
#person.phone_numbers.build(:number => "123456")
#person.phone_numbers.build(:number => "12346789012")
#person.save
The save fails because the first number wasn't valid. This is a good thing, to me. But what I don't understand is if its already validating the associated records what is the function validates_associated?
You can do has_many :phone_numbers, validate: false and the validation you're seeing wouldn't happen.
Why use validates_associated then? You might want to do validates_associated :phone_numbers, on: :create and skip validation on update (e.g. if there was already bad data in your db and you didn't want to hassle existing users about it).
There are other scenarios. has_one according to docs is validate: false by default. So you need validates_associated to change that.
When using accepts_nested_attributes_for, I got stuck when having a validation which required the original to be present. The code will help clear up that sentence.
class Foo < ActiveRecord::Base
has_one :bar
accepts_nested_attributes :bar
end
class Bar < ActiveRecord::Base
#property name: string
belongs_to :foo
validates_presence_of :foo #trouble line!
end
#now when you do
foo = Foo.create! :bar_attributes => {:name => 'steve'}
#you get an error because the bar validation failed
I would like to write a validation that goes something like...
class Bar < ActiveRecord::Base
validates_presence_of :foo, :unless => :being_built_by_foo?
end
I am currently using rails3.beta4
Thank you
Alas I don't have an answer to this post, but the I came up with another way so I didn't need the validation.
Since bar should never be without a foo then any request to create a bar without a foo_id is an error. In the real example a foo is a project, and bar is a bid. It is a nested resource, but I wanted to give access to json apps to be able to query the info from the /bids location so the router looked like.
resources :bids
resources :projects do
resources: bids
end
and then I just had to make sure all html access used project_bids_path or form_for [:project,#bid] etc. This next part is largely untested but so far the desired behavior is there. I got the idea from Yehuda's post on generic actions http://yehudakatz.com/2009/12/20/generic-actions-in-rails-3/
#I'm sure there is a better way then map.connect
map.connect "projects/invalid_id", :controller => "projects", :action => "invalid_id"
resources :projects
resources :bids
end
#couple of changes from Yehuda
def redirect(*args, &block)
options = args.last.is_a?(Hash) ? args.pop : {}
path = args.shift || block
path_proc = path.is_a?(Proc) ? path : proc {|params| path % params }
status = options[:status] || 301
lambda do |env|
req = Rack::Request.new(env)
#Get both the query paramaters and url paramaters
params = env["action_dispatch.request.path_parameters"].merge req.params
url = path_proc.call(params.stringify_keys)
#Doesn't add the port back in!
#url = req.scheme + '://' + req.host + params
#content-type might be a bad idea, need to look into what happens for different requests
[status, {'Location' => url, 'Content-Type' => env['HTTP_ACCEPT'].split(',').first}, ['Moved Permanently']]
end
end
def bid_path
redirect do |params|
if params['project_id']
"/projects/#{params['project_id']}/bids/#{params['id']}"
else
'/projects/invalid_id'
end
end
end
match "bids", :to => bid_path
match "bids/:id", :to => bid_path
however, after doing all of this I most definitely don't think it worth it. I think nested_attributes breaks things and can be improved if that validation doesn't work, but after looking through the code for a little while I'm not sure exactly how to fix it or if it's worth it.
first of all, when using nested_attributes, you'll get the presence of the container. in the example: when you save Foo and there's also a nested form for Bar, then Bar is built by Foo.
I think there's no need to make this kind of validation if you're sure to use Bar only in contexts with Foo.
btw, try to write validation as follow (new preferred syntax for Rails3):
validates :foo, :presence => true
hope this helps,
a.