Strange ActiveRecord::AssociationTypeMismatch

I am getting a very strange error when running a spec:

Failure/Error: entity = Factory.create(:entity, :name => "Test Entity", :creator => user)
 ActiveRecord::AssociationTypeMismatch:
   ::User(#97318850) expected, got User(#92770800)

This is the code that results in the above error. Factory is a factory_girl factory.

  user = Factory(:开发者_Go百科user, :username => "kai", :email => "xxx@yyy.com", :password => "testing")
  entity = Factory.create(:entity, :name => "Test Entity", :creator => user)

When I use :creator => User.first then everything works as expected. I printed out User.first and user, but see no difference.

Any suggestions what the heck is wrong here?

Update

I also got this error when running this simple request spec

describe "Entities" do
  it "should succeed" do
    entity = Factory.create(:entity, :name => "Test Entity 1")
    visit root_path
  end

  it "should also succeed" do
    entity = Factory.create(:entity, :name => "Test Entity 2")
    property = Factory.create(:property, :entity => entity)
  end
end

This time I get

Failure/Error: property = Factory.create(:property, :entity => entity)
 ActiveRecord::AssociationTypeMismatch:
   Entity(#103620190) expected, got Entity(#96047070)

when I delete visit root_path everything works fine (also when running each spec on its own). It just seems to be a problem for request specs. The other specs (model, controller) seems to run fine. I use Capybara 1.0.0.beta1 and RSpec 2.5.

What does this number behind the class name mean?

This is an error that occurs when two different versions of the model have been loaded. I used to hit it in an older version of Rails 3, since the development environment's model reloader was slightly glitched. The numbers after the class name refer to different versions of the class.

It stands to reason that this sort of error might come up in development mode, but it shouldn't in test mode, because, by default, classes are cached. See the config/environments/test.rb file to ensure that cache_classes is set to true.

Also check that you're on the latest version of Rails, 3.0.7. This may be a bug that has since been fixed. While we're at it, check that you're on factory_girl 1.3.3. When using the API totally correctly, which I think you're doing, the only possibilities left are that something is misconfigured or that it's a bug in someone else's code.

Rather than disable class caching, which can be aggravating while in development, the problem might disappear if you get your object fresh before using it. In my case I was loading an object from an association:

desired_object = foo.bar

Finding the item instead removed the problem and didn't require caching classes.

desired_object = Bar.find(foo.bar_id)

I know it isn't ideal, but perhaps this will help someone find out just why it is happening at all.

For the combination of Rails + Spring + factory_girl this is fixed since version v4.4.1 of factory_girl_rails (Feb 2014) see https://github.com/thoughtbot/factory_girl_rails/pull/121

Similar error might occur if you are using Spring or any other Rails application preloader, make sure you restart it.

spring stop
spring start
# or usually bin/rails s or bin/rails c for console

The issue for me was in my factory definition where I was using additional factories to populate id fields. I accidentally referred to an attribute that didn't exist on the table (account, instead of account_id). See example below.

factory Omni::CustomerAccount do
  sequence(:display_name) {|n| "test #{n}"}
  customer_id :customer  # this is correct
  account :account  # wrong - this should say account_id :account
end

Hope this helps someone.