Auto-detection of locales from WWW browser and testing with Cucumber

I test my application with Cucumber and it worked before I've added auto-detection of locales from WWW browser in application_controller.rb:

  before_filter :set_locale

  private

    def set_locale
      xxx = request.env['HTTP_ACCEPT_LANGUAGE'].scan(/^[a-z]{2}/).first
      if xxx.match /^(en|fr)$/
        I18n.locale = xxx
      else
        I18n.locale = 'en'
      end
    end

I have a scenario:

  Scenario: Successful sign up
    Given I am an anonymous user
    And I am on the home page
    When ...

When I run cucumber, I get an error:

Given I am an anonymous user                   # features/step_definitions/user_steps.rb:7
And I am on the home page                      # features/step_definitions/webrat_steps.rb:6
  private method `scan' called for nil:NilClass (NoMethodError)
  C:/workspace/jeengle/app/controllers/application_controller.rb:33:in `set_locale'
  c:/worktools/ruby/lib/ruby/1.8/benchmark.rb:308:in `realtime'
  (eval):2:in `/^I am on (.+)$/'
  features/manage_users.feature:8:in `And I am on the home page'

I have tried to do it in before statement in step_definitions folder:

Before do
  request.env['HTTP_ACCEPT_LANGUAGE'] = "en"
end

but I've got another error:

  开发者_开发百科undefined method `env' for nil:NilClass (NoMethodError)

Does anybody know how to initialize/emulate request.env['HTTP_ACCEPT_LANGUAGE'] in Cucumber?

---

UPDATED

Cucumber test passed when I rewritted set_locale method:

  xxx = request.env['HTTP_ACCEPT_LANGUAGE']    
  if xxx
    xxx = xxx.scan(/^[a-z]{2}/).first
    if xxx.match /^(en|ru)$/
      I18n.locale = xxx
  end
  else
    I18n.locale = 'en'
  end

It's not a solution but it works.

---

Actually the problem lies in Webrat, not Cucumber. The sequence of events is (roughly)

• Cucumber runs your feature
• When the "I am on the home page" step is reached, it invokes Webrat to make the request to the controller
• Webrat constructs a request and sends it to the controller
• Step fails because the request does not have an "Accept-Language" header

Apparently Webrat does not add this header when it builds the request. However, Webrat gives you a workaround: the "header" method, with which any header can be set for the duration of the request.

So to make this work add a step that puts in the header, e.g.:

Given /^an Accept Language header$/ do
  header "Accept-Language", "en;en-us" # or whatever value you need for testing
end`

Run this step before you visit a page and Webrat is no longer confused.

btw I got this from The Rspec Book, which really does a good job explaining how this BDD stuff fits together.

---

Another way is "same-same, but different". You can add before statement in your step definition file:

Before do
  header 'Accept-Language', 'en-US' 
end

This will be executed before each scenario and they remain clear and clean.