RSpec test custom validator

Question

I have the following validator in my model:

class ContinuumValidator < ActiveModel::Validator
  def validate(record)
    if !record.end_time.nil? and record.end_time < record.start_time
      record.errors[:base] << "An event can not be finished if it did not start yet..."
    end
  end
end

class Hrm::TimeEvent < ActiveRecord::Base
  validates_with ContinuumValidator
end

How can I test it using Rspec?

Here is what I have tried so far: (thanks to zetetic)

describe "validation error" do
  before do
    @time_event = Hrm::TimeEvent.new(start_time: "2012-10-05 10:00:00", end_time: "2012-10-05 09:00:00", event_type: 2)
  end

  it "should not be valid if end time is lower than start time" do
    @time_event.should_not be_valid
  end

  it "raises an error if end time is lower than start time" do
    @time_event.errors.should include("An event can not be finished if it did not start yet...")
  end
end

But I get the following errors:

1) Hrm::TimeEvent validation error raises an error if end time is lower than start time
   Failure/Error: @time_event.errors.should include("An event can not be finished if it did not start yet...")

   expected #<ActiveModel::Errors:0x007fd1d8e02c50 @base=#<Hrm::TimeEvent id: nil, start_time: "2012-10-05 08:00:00", end_time: "2012-10-05 07:00:00", event_type: 2, employee_id: nil, created_at: nil, updated_at: nil, not_punched: false, validated: false, replace_id: nil>, @messages={}> to include "An event can not be finished if it did not start yet..."

   Diff:
   @@ -1,2 +1,5 @@
   -["An event can not be finished if it did not start yet..."]
   +#<ActiveModel::Errors:0x007fd1d8e02c50
   + @base=
   +  #<Hrm::TimeEvent id: nil, start_time: "2012-10-05 08:00:00", end_time: "2012-10-05 07:00:00", event_type: 2, employee_id: nil, created_at: nil, updated_at: nil, not_punched: false, validated: false, replace_id: nil>,
   + @messages={}>

What am I doing wrong? And how can I achieve my goal? Any help or suggestion would be appreciated. Thanks.

Solution 1

The problem is that you're expecting @time_event.errors to behave like an array of strings. It doesn't, it returns ActiveModel::Errors. As others pointed out, you also need to trigger the validations with a call to valid?:

it "raises an error if end time is lower than start time" do
  @time_event.valid?
  @time_event.errors.full_messages.should include("An event can not be finished if it did not start yet...")
end

Solution 2

This solution works for me (using Mongoid):

The model

class OpLog
...
field :from_status, type: String
field :to_status,   type: String
...
validate :states_must_differ

def states_must_differ
  if self.from_status == self.to_status
    errors.add(:from_status, "must differ from 'to_status'")
    errors.add(:to_status, "must differ from 'from_status'")
  end
end
...
end

The test:

it 'is expected to have different states' do
  expect { create(:oplog, from_status: 'created', to_status: 'created').to raise_error(Mongoid::Errors::Validations) }
end

So in your case I'd write a test like this (if using ActiveRecord):

it 'raises an error if end time is lower than start time' do
  expect { create(Hrm::TimeEvent.new(start_time: "2012-10-05 10:00:00", end_time: "2012-10-05 09:00:00", event_type: 2)) }.to raise_error(ActiveRecord::Errors)
end

Solution 3

There are no errors because you haven't called an event that triggers the errors. This happens normally when a record is created or saved. You may not want to hit the database in your test though and then you can use the method valid? like this:

it "raises an error if end time is lower than start time" do
  @time_event.valid?
  @time_event.errors.should include("An event can not be finished if it did not start yet...")
end

Me personally would put these two tests into one since valid? is called in the first case.

Also a minor: if record.end_time is better than if !record.end_time.nil?. (In my opinion at least.... :-) )

Comments

[siekfried]

You're right I need to add the full_messages to get the error. However like the other said, I need to actually test the validation with valid?

[siekfried]

You are right, but like @Benjamin Sullivan suggested, I also need to add the full_messages to the errors so that my test pass. Thanks also for the minor :)