RE: [Boost-users] Re: why doesn't scoped_ptr have a custom

From: Reid Sweatman (drunkardswalk_at_[hidden])
Date: 2005-01-08 15:38:39

> -----Original Message-----
> From: boost-users-bounces_at_[hidden] 
> [mailto:boost-users-bounces_at_[hidden]] On Behalf Of Ben 
> Hutchings
> Sent: Thursday, January 06, 2005 6:25 AM
> To: boost-users_at_[hidden]
> Subject: Re: [Boost-users] Re: why doesn't scoped_ptr have a custom
> 
> 
> Pablo Aguilar wrote:
> > Thanks for your reply...
> > 
> > You're right, "reset" would delete the pointer... but when you're 
> > getting
> > rid of a pointer obtained from a factory, like I was in my 
> case, pointers 
> > don't always get disposed of using delete, but rather 
> another function (like 
> > p->release())
> > 
> > And rather than writing special purpose smart pointers for 
> each type 
> > of
> > deleting function (p->dispose(), p->release(), 
> api_function_free(p), etc..) 
> > you can use shared_ptr with a custom deleter (here: 
> > mem_fn(&pointer_type::dispose) or bind(&api_function_free, _1))
> 
> In some cases you can avoid the space and time overhead of 
> shared_ptr by 
> using a smart pointer template for which the custom deleter is a 
> template argument.  In fact you can get even more general than that 
> quite easily; I have a class template called auto_handle which with 
> minimal work can be used to encapsulate all kinds of things 
> with custom 
> deleters, including Windows handles and POSIX file descriptors.

Good point.  Probably most developers in this group have done something like
that; mine is an editor macro customized for my favorite editor, I use it so
often.  It lets me supply the parameterizations using an OS-specific, but
minimally-intrusive interface as I edit.

Reid