Re: [Boost-users] [Smart Ptr] Why is weak_ptr a _ptr?

Subject: Re: [Boost-users] [Smart Ptr] Why is weak_ptr a _ptr?
From: Marsh Ray (marsh_at_[hidden])
Date: 2010-10-13 14:03:10

On 10/13/2010 04:42 AM, Szymon Gatner wrote:
> Hi,
>
> Is it just me or is it really the worse naming in whole boost library?

I like good naming as much (really more more than) the average 
programmer, but sometimes there isn't a great choice. In a case like 
this, you're lucky when there's common usage to fall back on.

Finding amusement in the paradoxes that inevitably arise is a good way 
to think about and understand them.

> It is nothing like any other smart pointer.

> No initialization / construction from raw pointer.
>
> No operator * or ->  (the very least one would expect from _ptr class).
>
> No get().
>
> And many more which I now forgot. My point is that is is not a pointer
> at all, not even a bit.

OK, it's not a pointer. Yet it ends in "_ptr"!

Now, wouldn't it be just great if this were the worst, most 
counter-intuitive, inconsistency in the standard language?

However, generations of programmers have grown up with the concepts of 
"strong" and "weak" references. In C and C++ these concepts are have 
been intimately related to pointers (for obvious reasons).

But it turns out not to be possible in C++ to implement weak references 
that look like pointers directly in a way that avoids the possibility of 
seriously undefined behavior.

I don't see anything inherent in the concept of a weak reference that 
says it has to look syntactically similar to a strong reference. In 
fact, it has a more complex interface by definition.

> It only happens to have a member variable returning
> one and that is hardly a reason to call type pointer-like.

No, its interface definition doesn't include any member variables. One 
could cook up an implementation which didn't have any either.

I just think of it (loosely) as an object that has a chance (determined 
at runtime) of being convertible to a shared_ptr. A lot like 
boost::optional really.

> It should rather be called shared_tracker or shared_listener or something
> in that taste ;)

Everybody who's ever used a garbage collected system should know what 
"weak" refers to in this context. I don't know what a "tracker" is 
supposed to be, and this thing doesn't sound much like a "listener" either.

Play the name game to its logical conclusion and eventually it won't 
make a bit of sense to anyone. :-)

- Marsh