Re: Proposal: force_cast<>

From: bill_kempf (williamkempf_at_[hidden])
Date: 2002-01-16 11:55:22

--- In boost_at_y..., Mac Murrett <mmurrett_at_v...> wrote:
> On 1/16/02 2:20 AM, "jk_at_s..." <jk_at_s...> wrote:
> 
> > 16 Jan 2002 06:52:11 +0300 Mac Murrett ???????:
> >> Its purpose is to convert anything to anything, and it looks 
like this:
> >> 
> >> // force_cast will convert anything to anything.
> >> 
> >> template<class Return_Type, class Argument_Type>
> >> inline Return_Type force_cast(const Argument_Type &rSrc)
> >>    {    return(*reinterpret_cast<const Return_Type *>
(&rSrc));    }
> > 
> > Then copying will occur. Wouldn't it be better to return by 
reference,
> > like below?
> > 
> > template<class Return_Type, class Argument_Type>
> > inline const Return_Type& force_cast(const Argument_Type &rSrc)
> >   {    return(*reinterpret_cast<const Return_Type *>(&rSrc));    }
> 
> If a user wants this, (s)he can use force_cast<Type &>.  Also, I 
don't think
> that returning a const & by default is fair, as users should have 
the right
> to modify their data :).

I didn't notice this in your original code.  A cast does only that, 
it casts the type.  When you use such a cast you should still be 
manipulating the same memory space: you should only be treating it as 
some new type.

Bill Kempf