shoulda-matchers: undefined method `attribute_setter' for nil:NilClass

Hello!

I’m receiving an error when a validation fails with the validates_inclusion_of matcher. If the validation is present, everything is fine, but without it when building the failure message you get this error:

NoMethodError:
       undefined method `attribute_setter' for nil:NilClass
     # ./.gem/ruby/2.3.0/gems/shoulda-matchers-3.1.1/lib/shoulda/matchers/active_model/allow_value_matcher.rb:405:in `failure_message'
     # ./.gem/ruby/2.3.0/gems/activesupport-4.2.5.1/lib/active_support/core_ext/object/try.rb:77:in `public_send'
     # ./.gem/ruby/2.3.0/gems/activesupport-4.2.5.1/lib/active_support/core_ext/object/try.rb:77:in `try!'
     # ./.gem/ruby/2.3.0/gems/activesupport-4.2.5.1/lib/active_support/core_ext/object/try.rb:63:in `try'
     # ./.gem/ruby/2.3.0/gems/shoulda-matchers-3.1.1/lib/shoulda/matchers/active_model/validation_matcher.rb:129:in `failure_reason'
     # ./.gem/ruby/2.3.0/gems/shoulda-matchers-3.1.1/lib/shoulda/matchers/active_model/validation_matcher.rb:56:in `block in failure_message'
     # ./.gem/ruby/2.3.0/gems/shoulda-matchers-3.1.1/lib/shoulda/matchers/active_model/validation_matcher.rb:55:in `tap'
     # ./.gem/ruby/2.3.0/gems/shoulda-matchers-3.1.1/lib/shoulda/matchers/active_model/validation_matcher.rb:55:in `failure_message'
     # ./spec/lib/models/address_spec.rb:22:in `block (2 levels) in <module:Models>'

My class and spec look close to this:

class Thing
  include Virtus.model
  include ActiveModel::Model

  ARRAY = ["value"] 

  attribute :thing, String
end

RSpec.describe Thing do
  it { is_expected.to validate_inclusion_of(:thing).in_array(Thing::ARRAY) }
end

I am using active model outside of rails, configuring Shoulda to integrate with RSpec and ActiveModel. Ruby version is 2.3.0 and ActiveModel version is 4.2.5.1

Maybe this is an incompatibility with Virtus that I’m not aware of?

Thanks

About this issue

Original URL
State: closed
Created 8 years ago
Reactions: 17
Comments: 27

Commits related to this issue

Proof of concept for issue #904 Running specs result in the error mentioned on issue #904: ``` Failures: 1) Person should validate that :sex is either ‹"male"› or ‹"female"› Failure/Er... — committed to cabello/shoulda-matchers by cabello 8 years ago

Most upvoted comments

This should be fixed in master now. We’re cooking up an RC so that y’all can try out the fix. We’ll post again when we have that.

mcmire on Sep 16, 2018

What’s the status on this one? Has anyone come up with a solution?

Mange on Sep 14, 2016

it can be workarounded with allow_value matcher tho:

  describe 'my_attr values' do
    it { is_expected.to allow_value('a').for(:my_attr) }
    it { is_expected.to allow_value('b').for(:my_attr) }
    it { is_expected.not_to allow_value('anything other').for(:my_attr) }
  end

ktopolski on Jan 30, 2018

This error will also occur when testing that a presence validation does not exist when it actually does:

# Model
class Event
  validates :date, presence: true
end

# Test
describe Event do
  it { should_not validate_presence_of(:date) }
end

mcmire on Oct 31, 2017

@wilianto

That’s not a bad idea, but we would probably have to make the same change across the board, so I would hesitate to do that.

I’m (slowly) working on a branch that will change how “submatchers” (those are matchers that are used internally by other matches) are processed and should address this issue here and anywhere else it appears.

mcmire on Apr 16, 2018

I just got this error as well when saying should_not validate_confirmation_of(:some_attribute). It doesn’t matter if the test actually passes or not, it will still crash.

mcmire on Aug 29, 2016