Re: [Boost-users] [smart_ptr] Resetting an intrusive_ptr

From: Nat Goodspeed (ngoodspeed_at_[hidden])
Date: 2006-05-02 13:20:20

> -----Original Message-----
> From: boost-users-bounces_at_[hidden] [mailto:boost-users-
> bounces_at_[hidden]] On Behalf Of Scott Meyers
> Sent: Tuesday, May 02, 2006 1:00 PM
> To: boost-users_at_[hidden]
> Subject: Re: [Boost-users] [smart_ptr] Resetting an intrusive_ptr
> 
> Peter Dimov wrote:
> > The reason for p = 0 working in intrusive_ptr's case is that
> intrusive_ptr
> > has an assignment operator taking a raw pointer. The other smart
pointer
> > types do not.
> 
> I understand that, my question was more about why this operation is
> supported only for intrusive_ptr.

[Nat] The library discourages casual conversion from Foo* to
shared_ptr<Foo> because it is a Bad Thing when you instantiate a second
shared_ptr<Foo> for the same Foo* value. Now you have two different
reference counts floating around on the heap. When either goes to zero,
every shared_ptr<Foo> value based on the other reference count becomes a
land mine.

On the other hand, intrusive_ptr is designed to be able to cope with
casual conversion of exactly that type. That's why, unlike shared_ptr,
the intrusive_ptr constructor that accepts a dumb pointer is not marked
'explicit'. Simple dumb-pointer assignment works for the same reason.

Concerning the missing reset() method, I can only surmise that the
author considered it unnecessary, given assignment support. I for one
wouldn't object to it being added.