From: Johan Nilsson <johan.nilsson@...>
Date: 2005-01-19T18:06:02+09:00
Subject: Writing test cases/suites

Hi,

I pretty often use small helper classes for my unit tests; call them simple 
mock objects or what you like. A trivial example, for the sake of 
illustration:

--- tc_myclass.rb ---

require 'test/unit'

class CountingClient
    attr_reader :times_called
    def initialize
        @times_called = 0
    end
    def on_cb(whatever)
        @times_called += 1
    end
end

class TC_MyClass < Test::Unit::TestCase

    def test_no_of_dispatched_calls

        mc = MyClass.new
        cc = CountingClient.new

        mc.add_listener(cc)
        mc.dispatch("A")
        mc.dispatch("B")

        assert_equal(2, cc.times_called)

    end

end

-----------------
My problem comes now when I create a test suite, e.g.

--- ts_myclasses.rb ---
require 'test/unit'
require 'tc_myclass'
require 'tc_myotherclass'
-----------------------

Imagine now that they both define test helper classes named CountingClient, 
and that the two definitions of CountingClient differ (well, otherwise 
they'd be in a separate file). How should I handle this without naming the 
classes "TC_MyClass_CountingClient" etc...

What I did in C++ (for those who are familiar with that) was to always 
define such classes inside an anonymous namespace, which means that they are 
only accessible within the current compilation unit. A rubyized comparison 
would be:

module # Note: no name

class CountingClient
    ...
end # CountingClient

end # anonymous module

The class CountingClient should then only be accessible within the current 
..rb file and take precedence over previous definitions. But no such beast is 
available - right?


Regards // Johan