[rspec-users] Name collision - how would you handle this?
dchelimsky at gmail.com
Sun Aug 8 20:54:55 EDT 2010
On Aug 8, 2010, at 11:13 AM, Matt Wynne wrote:
> On 8 Aug 2010, at 16:53, David Chelimsky wrote:
>> On Aug 8, 2010, at 10:40 AM, Matt Wynne wrote:
>>> On 8 Aug 2010, at 16:38, David Chelimsky wrote:
>>>> On Aug 7, 2010, at 4:10 PM, David Chelimsky wrote:
>>>>> Hey all,
>>>>> It turns out that if you have
>>>>> * Rails (2 or 3)
>>>>> * Ruby-1.9
>>>>> * a model named Message
>>>>> * let(:message) or def message in an example group
>>>>> * a Rails assertion in an example in that group
>>>>> * note that rspec-rails' matchers delegate to Rails' assertions
>>>>> You'll get an error saying "wrong number of arguments (1 for 0)"
>>>>> This is because the rails assertion, which, when running with Ruby-1.9, delegates to Minitest::Assertions#assert_block, which delegates to a message() method that it defines. So the message() method defined by let() overrides the message() method in the Assertions module, and results in unexpected and undesirable outcomes.
>>>>> So - what should we do? I don't think changing Minitest is really an option, as too many assertion libraries already wrap Minitest assertions. I don't think RSpec should be in the business of monitoring methods end-users define to make sure they're not overriding pre-existing methods (what if you override a method intentionally?). The only thing I'm left with is document this particular case and hope for the best, but that feels unsatisfactory as well.
>>>>> Recommendations? Words of wisdom?
>>>> FYI - here's the issue that spawned this thread: http://github.com/rspec/rspec-rails/issues/152
>>> Can you use the Assertions module some other way than mixing it into the example (thereby polluting it with the Assertions module's methods?)
>> I like the idea in the abstract, but most of the rails assertions rely on some state that is local to the example (@response, @controller, @request, etc, etc). RSpec _could_ gather up all those instance variables and pass them into an assertion-wrapper object, but then it would be highly coupled to that implementation and would lead us down a familiar and unfriendly path of forcing rspec-rails releases for every rails release. That's a world I hope to leave behind with Rails 3 :)
> So leave the rails assertions mixed into the example, but forward all the calls to the MiniTest::Assertions methods to some other object that has them mixed in. Won't that work?
Here's a prototype implementation: http://github.com/rspec/rspec-rails/commit/0cd384536cf532435ec8f290a9c357b60872acd7
It's on a branch (http://github.com/rspec/rspec-rails/tree/assertion-delegate) because I'm not convinced this is the right way to go yet, but I'd like some feedback from anyone who wishes to peruse and comment.
>> It would also eliminate the option to use the Rails assertions directly in examples.
>> Oh, well :)
>>> +44(0)7974 430184
>> rspec-users mailing list
>> rspec-users at rubyforge.org
> +44(0)7974 430184
> rspec-users mailing list
> rspec-users at rubyforge.org
More information about the rspec-users