On Apr 5, 2010, at 8:09 AM, Pat Maddox wrote:
>
> Sounds like a lot of work
>
>
> On Apr 3, 2010, at 9:08 PM, Julian Leviston wrote:
>
>> Sorry I meant send AND __send__
>>
>> Julian.
>>
>> On 04/04/2010, at 11:45 AM, Julian Leviston wrote:
>>
>>>
>>> On 04/04/2010, at 7:32 AM, David Chelimsky wrote:
>>>
>>>> On Sat, Apr 3, 2010 at 8:56 AM, Vojto Rinik <[email protected]> wrote:
>>>>> Hello RSpec users!
>>>>> I have one abstract class and a few classes that inherit from that
>>>>> abstract
>>>>> one.
>>>>
>>>> Ruby doesn't have abstract classes. You can have a base class that you
>>>> don't _intend_ to instantiate directly, but there's nothing stopping
>>>> you from doing so, so it's not like an abstract class in java, which
>>>> you actually can't instantiate directly.
>>>>
>>>> I've seen some cases where the initialize method is made private so
>>>> you can't just call Foo.new, so it sort of feels like an abstract
>>>> class, but even in that case you can still use send() to instantiate
>>>> one in a test:
>>>>
>>>> AbstractIshClass.send(:new)
>>>>
>>>
>>> How about if you overrode new and __new__ ?
>>>
>>> Julian.
How about:
class AbstractClassException < RuntimeError; end
class AbstractishClass
def initialize
raise AbstractClassException.new("don't do that.")
end
end
That way, even sending a message to new causes a failure. And that's what
you're looking for, right?
_______________________________________________
rspec-users mailing list
[email protected]
http://rubyforge.org/mailman/listinfo/rspec-users
