opCast for classes

[deadalnix]

On Tuesday, 12 March 2013 at 20:17:33 UTC, Ali Çehreli wrote:
> On 03/12/2013 01:08 PM, "Luís Marques" <luismarques at gmail.com>" 
> wrote:
> > Hi there,
> >
> > I've been away for a while from the D world, so I guess I
> missed some of
> > the new things.
> >
> > Regarding opCast, the documentation says:
> >
> > "This only happens, however, for instances of structs. Class
> references
> > are converted to bool by checking to see if the class
> reference is null
> > or not."
> >
> > Yet this works:
> >
> > class B
> > {
> > bool v;
> >
> > this(bool v)
> > {
> > this.v = v;
> > }
> >
> > T opCast(T)()
> > {
> > return v;
> > }
> > }
> >
> > unittest
> > {
> > B bfalse = new B(false);
> > B btrue = new B(true);
> > assert(cast(bool) bfalse == false);
> > assert(cast(bool) btrue == true);
> > }
> >
> > Is that a new feature? Can I rely on it? Is it documented
> somewhere?
> >
> > Thanks! :-)
> >
> > Regards,
> > Luís
>
> I don't know what the intended behavior but there is a 
> distinction between automatic vs. implicit. These pass as well:
>
>     assert(bfalse);
>     assert(btrue);
>
> So, apparently implicit conversion considers the class variable 
> and explicit conversion considers the class object. And this 
> produces a compilation error:
>
>     B bnull;
>     assert(cast(bool)bnull);
>
> Error: null dereference in function 
> _D6deneme19__unittestL123991_1FZv
>
> Ali

bool toto = bfalse; // Error: cannot implicitly convert 
expression (bfalse) of type module.B to bool

So it isn't the implicit cast kickin here, but a 3rd behavior. 
The kind of behavior that makes D so special and create theses 
edges cases we all love !