[rspec-users] Testing url_for

Rodrigo Rosenfeld Rosas lbocseg at yahoo.com.br
Tue Oct 27 19:48:06 EDT 2009

Em 27-10-2009 15:17, David Chelimsky escreveu:
> On Oct 27, 2009, at 8:21 AM, Rodrigo Rosenfeld Rosas wrote:
>> Hi David, I'm giving a try to RSpec after we meet each other on Rails 
>> Summit Latin America and I must admit I'm enjoying using 
>> rspec/machinist/faker.
>> Since I have not written any controllers yet, I hadn't taken a chance 
>> to try webrat.
>> But there is a situation that I would like some feedback on how to 
>> deal with it.
>> When registering new users, they will input their e-mail and a 
>> message will be sent for them to confirm their addresses and continue 
>> registering.
>> I use something like:
>> MailNotifier.deliver_email_confirmation_message :confirmation_url => 
>> url_for(:controllers => 'users', :action => 'continue_register', 
>> :user => @user.id, :token => @user.confirmation_token)
>> And the routes are set to ':controller/:action', so that the url 
>> would translate to 
>> '/users/continue_register?user=2&confirmation_token=asdf987asf'.
>> The problem is that Ruby 1.8 will not maintain any specific order for 
>> the parameters. (The application is hosted in a shared server at 
>> hostingrails.com, which hosts Ruby 1.8)
>> I know that I could add a route to generate 
>> '/users/continue_register/2/asdf987asf' instead, but I would still 
>> like to know what would be the alternatives.
> Oi Rodrigo,
> There's no great alternative that I know of. I've always just grabbed 
> the URL using a regexp and then broken it up. Something like this:
> text.should =~ /http:\/\/test\.host\/users\/continue_register\?([^\s]*)/
> query_string = $1
> query_string.should =~ /user=2/
> query_string.should =~ /confirmation_token=asdf987asf/
> You could also use Rack::Utils.parse_query to convert the query_string 
> to hash, or capture the entire url and use the route_to matcher:
> text.should =~ /(http:\/\/test\.host\/users\/continue_register\?[^\s]*)/
> {:get => $1}.should route_to(
>   :controllers => 'users',
>   :action => 'continue_register',
>   :user => @user.id,
>   :token => @user.confirmation_token
> )
> I don't love either of those - I'd sooner change the implementation to 
> something deterministic, but that's me :)

David, thank you very much for your feedback, although not ideal they 
are good approaches.

I understand your point of view in changing to something deterministic 
and I was thinking in doing that, at first. But I would like to know if 
there was a gotcha that I couldn't figure it out. I need to choose 
between clean code in the implementation or in the test...

Using a deterministic approach would make the test clearer, while using 
url_for with Ruby 1.8 would make the implementation clearer... That is 
why I was thinking in creating a route to achieve both, but it would 
exist only to make both clearer and could cause some confusion while 
reading routes.rb...

The first approach you presented is clearer, but wouldn't detect a 
problem that I noticed when testing my mailer. I was using RedCloth's 
textilize, thinking that it would generate a link (which actually 
auto_link does) and I was testing if the link was contained in the body. 
The first approach wouldn't detect that the url was wrong, like 

The second approach goes against Ruby way, in my opinion... :) Too many 
code to test a single statement...

Another approach would be something like:

    match(email.body).to_a[1..-1].to_set.should ==
       ["user=#{@user.id}", "token=#{@user.activation_token}"].to_set

where the first line could be replaced, in this case, with:

although I would prefer another matcher to exist:
have_same_elements(["user=#{@user.id}", "token=#{@user.activation_token}"])

But the problem with this approach is that I would probably need to 
write another test for testing this test ;)

>> How could I verify that the delivered message contains a correct url? 
>> I know that I should follow the url in an acceptance test, but I'm 
>> just trying to test that the message is been correct generated, in a 
>> unit test.
>> Please, let me know if I missed something conceptually while testing 
>> this situation.
>> Thanks for RSpec and the tips about machinist, faker and webrat.
>> Just one more doubt. When using machinist, is it possible to ignore 
>> the blueprint while calling 'make' on an ActiveRecord class? I had to 
>> create a named blueprint reseting all fields set by the blueprint.
> I don't know of a way of to do this, but why do you need to?

Usually, for most tests, I need to create active users.

Only for testing the registering process, I need another structure.

Active users must have login, name, email and password.

Inactive users would only have email.

> You may want to look at a couple of other libraries like Fixjour, 
> Fixture Replacement and Factory Girl - they serve the same function as 
> Machinist, but don't (afaik) add methods to ActiveRecord::Base.

Actually, I've taken a look at Factory Girl, but prefered Machinist, 
which is awesome. I'll take a look at Fixjour and Fixture Replacement. 
Do they have a blueprint-like feature?

I have no problem in creating methods in ActiveRecord::Base. I just 
wanted to call something like:

User.make_without_blueprint(:email => 'test at test.com', :active => 'false')

Thank you for your feedback,


Faça ligações para outros computadores com o novo Yahoo! Messenger 

More information about the rspec-users mailing list