functional testing with logging in

Hello,

One of the actions in one of my controllers requires a login and I'm
trying to write a test for that action. I don't want to actually go
through the login process. I just want to initialize a session
variable and be done with it :slight_smile:

So, the way the authentication works is pretty simple. In my
application controller I defined this method:

  def authenticate
    if session[:user].nil?
      session[:initial_uri] = request.request_uri
      redirect_to :controller => 'user', :action => 'login'
    end
  end

In the controller, I have :before_filter = "authenticate"

Now, my test so far is:

  def test_show_valid_request
   # Assign the user session variable the user with id = 1
    @request.session[:user] = User.find(1)

   # Santity test -- this does say that the value is an ActiveRecord
object.
    puts @request.session[:user]

    # Try to call the action, which triggers a call to authenticate
    get :show, { :id => '1' }, { :customer_id => '16' }

    # This should fail because no redirection should take place
    assert_redirect_to(:controller => 'user', :action => 'login')
end

What happens is that the "if session[:user]" in "authenticate" returns
nil.

tail -f on logs/test.log has "Session ID: ", which would mean that no
session exists?

Anybody know what I'm missing?

Because your using a before filter, and the session variable is set there is not actually anything being evaluated for the method. Hence the last evaluated expression for the method is nil.

If at the end you returned the session[:user] it should work.

def authenticate
if session[:user].nil?
session[:initial_uri] = request.request_uri
redirect_to :controller => ‘user’, :action => ‘login’

end
session[:user]

end

There are dangers associated with storing your user object in the session. Stale objects and more. There is plenty of discussion about it on the list and elsewhere. It is common practice to just store the ID of the user object in the session and then do a lazy lookup of the user when you need it. It only results in one lookup per request for the user object.

eg something like

def current_user
@current_user ||= User.find_by_id( session[:user_id] ) if session[:user_id]
end

HTH
Daniel

Because your using a before filter, and the session variable is set there is
not actually anything being evaluated for the method. Hence the last
evaluated expression for the method is nil.

Hmm, I don't think that's it because I'm not relying on the return of
value of "authenticate". It just redirects or doesn't. And it does
redirect, which means that session[:user] is nil by the time
authenticate is called. But if I do this:

get :show, { :id => '1' }, { :user => User.find(1) }

it does work. So, for some reason if the session variable is set as
part of the get call, it works.

There are dangers associated with storing your user object in the session. Stale objects and more.

Thanks for the tip. I'll rework my process to take that into account.
Never thought about it!

Thanks for the post, Daniel.

Because your using a before filter, and the session variable is set there is
not actually anything being evaluated for the method. Hence the last
evaluated expression for the method is nil.

Hmm, I don’t think that’s it because I’m not relying on the return of
value of “authenticate”. It just redirects or doesn’t. And it does
redirect, which means that session[:user] is nil by the time

authenticate is called. But if I do this:

get :show, { :id => ‘1’ }, { :user => User.find(1) }

it does work. So, for some reason if the session variable is set as
part of the get call, it works.

There are dangers associated with storing your user object in the session. Stale objects and more.

Thanks for the tip. I’ll rework my process to take that into account.
Never thought about it!

Thanks for the post, Daniel.

In IRB

?> def test

if @test.nil?
“Blah”
end
end
=> nil

test
=> “Blah” # => When @test is not set the method returns a value

@test = “I’m Set”
=> “I’m Set”

test
=> nil #=> When @test is set the method returns nil

Can you see that in your situation, when your session variable is set the method will return nil?

From the docs you can see.
http://api.rubyonrails.org/classes/ActionController/Filters/ClassMethods.html

[before_filter](http://api.rubyonrails.org/classes/ActionController/Filters/ClassMethods.html#M000180) and [ around_filter](http://api.rubyonrails.org/classes/ActionController/Filters/ClassMethods.html#M000186) may halt the
request before controller action is run. This is useful, for example, to
deny access to unauthenticated users or to redirect from http to https.
Simply return false from the filter or call render or redirect.

So if the result of a method evaluates to false ( either false or nil ) then the action of the controller will not be run. Since your redirect_to is inside an if statement whose body is never executed it does not re-direct either. So even though you aren’t explicitly relying on the return value of authenticate, rails is as part of the before_filter behavior.

I’ve never actually tried this. What is returned in this situation?

Daniel

Can you see that in your situation, when your session variable is set the method will return nil?

If I do @request.session[:user_id] = 1, the "if session[:user].nil?"
returns true and the redirect does happen.
If I do get :show, { :id => '1' }, { :customer_id => '16' }, the
method does return nil, but the action of the controller is not
stopped.

That's interesting about halting the action of a controller. It's
strange though. Say I do return false from a before filter, what would
the user see?

I'm not a big expert on all this -- just reporting the results.

Sergei

There is a better way (IMO). Move as much authentication as possible into your model, then create a logged_in? helper function in your application.rb. For your tests, use Mocha (http://mocha.rubyforge.org/) and stub the logged_in? function to return true for tests where the user is meant to be logged in and false for ones where you might have a bogus user.

E.g.,

your controller

def test_user_is_logged_in

controller.stubs(:logged_in?).returns(true)

end

Just a thought…

Steve, thanks for the idea. Right now though I want to stick to the
framework so I can keep the amount of new information to a minimum :slight_smile:

Here is some code from one of my older projects:

in controller

def test_admin

get_logged_in

get :admin

assert_response :success

end

in test_helper.rb

def get_logged_in

request.session[:user] = users(:users_004).id

theoretically, there is more involved in getting logged in… i.e., login time

end

Hope this helps.

Yes, that's what I was doing too -- except are you sure it's
"request.session[:user] = users(:users_004).id" and not
"@request.session[:user] = users(:users_004).id"? Because otherwise I
get "an unknown local variable". Anyways, this general approach is not
working for me for some reason. But I'm OK with passing the session
variables along with the call to "get"

Thanks!
Sergei