From justinf at wantsa.com  Thu Mar  4 17:48:34 2010
From: justinf at wantsa.com (Justin Francis)
Date: Thu, 04 Mar 2010 14:48:34 -0800
Subject: [mocha-developer] sharing any_instance stubs with subclasses
Message-ID: <4B9038C2.7010508@wantsa.com>

Hello,

I ran into a problem where stubbing `any_instance` of a parent class' 
method would effectively prevent anyone from calling that method on a 
subclass instance (using 0.9.8). I know something similar was brought up 
about two years ago 
(http://rubyforge.org/pipermail/mocha-developer/2008-January/000521.html)

===================
class A
   def t()
     'in A'
   end
end

class B < A
end

A.any_instance.stubs(:t).returns('stubbed')
b = B.new
b.t       # throws error
===================

Here, calling b.t would give the following error:

==================
Mocha::ExpectationError: unexpected invocation: #<AnyInstance:B>.t()
satisfied expectations:
- allowed any number of times, not yet invoked: 
#<AnyInstance:A>.t(any_parameters)
==================

Anyway, I needed this to work, so I implemented a naive solution where 
subclasses will share their parents "AnyInstance" instances and attached 
the patch to this message.

Thought this might be useful to others, even though I realise it does 
not fully solve the issue (in this example, if B.t was defined as an 
overridden method, B.t would still be called, not the stubbed version in A).

Thanks,

Justin Francis


From jamesmead44 at gmail.com  Sun Mar  7 12:53:00 2010
From: jamesmead44 at gmail.com (James Mead)
Date: Sun, 7 Mar 2010 17:53:00 +0000
Subject: [mocha-developer] sharing any_instance stubs with subclasses
In-Reply-To: <4B9038C2.7010508@wantsa.com>
References: <4B9038C2.7010508@wantsa.com>
Message-ID: <1db558f01003070953i32177f1bpd164024d8a244f1a@mail.gmail.com>

Hi Justin,

Thanks for trying to send us a patch. I don't think attachments make
it through on Google Groups. Have you considered forking Mocha on
GitHub [1] and sending me a pull request? Alternatively you could use
Gist [2] or Pastie [3] to show us the code.

I have to confess, I haven't used Mocha's any_instance functionality
for a long time. I find there are usually simpler ways of stubbing
methods and setting expectations. Well done for finding that old
mailing list thread. There's another more recent one here [4].

Have you tried running the Mocha tests with your patch applied? Also
it would be really helpful if you could write new acceptance tests
like these [5] to describe the behaviour you are expecting and/or
identify existing tests which describe behaviour you are not
expecting.

If you're interested in working out how you might avoid the use of
any_instance, it might be worth posting some code and explaining what
you're trying to do and we can see if we can come up with an
alternative approach.

Cheers, James.

[1] http://github.com/floehopper/mocha
[2] http://gist.github.com/
[3] http://pastie.org/
[4] http://groups.google.com/group/mocha-developer/browse_thread/thread/99981af7c86dad5e