[boost] Re: Another safe_bool alternative?

From: David Abrahams (dave_at_[hidden])
Date: 2003-12-03 14:18:35

Rob Stewart <stewart_at_[hidden]> writes:

> From: Douglas Paul Gregor <gregod_at_[hidden]>
>> 
>>   struct some_safe_bool_nested_class {
>>     int my_data_member;
>>   };
>> 
>>   typedef int some_safe_bool_nested_class::*safe_bool_type;
>> 
>> You don't ever need an instance of some_safe_bool_nested_class.
>
> Here's the context for what you're proposing.  It's code that
> Dave posted in an earlier message.
>
>    template <class T>
>    struct bool_testable
>    {
>    private:
>       struct impl { int x; };
>       typedef int impl::*safe_bool;
>    public:
>       operator safe_bool() const
>       {
> 	 return !*static_cast<T const*>(this) ? 0 : &impl::x;
>       }
>    };
>
> I'm clearly missing something.  How can there be an address for
> impl::x if there's no instance of impl?
>
> Were impl::x declared static, there could be an instance --
> though you'd need to define it somewhere -- so you could take its
> address.  Lacking that, "impl::x" must refer to something and so
> it must either come from a dummy dm in bool_testable or from the
> derived class (through a template parameter).

A pointer to non-static member exists independently of an object.  You
can bind it to an object and dereference it with 

    object.*pointer_to_member

HTH,
-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com