[boost] Re: ClonePtr and CopyPtr take 2

From: Kevin Atkinson (kevinatk_at_[hidden])
Date: 2000-03-08 08:21:21

On Wed, 8 Mar 2000, Dave Abrahams wrote:

> on 3/8/00 5:11 AM, Kevin Atkinson at kevinatk_at_[hidden] wrote:
> 
> >> template <typename T>
> >> void ClonePtr<T>::Parms::assign(T * & rhs, const T * lhs) const {
> >> if (typeid(*rhs) == typeid(*lhs)) {
> >> rhs->assign(lhs);
> >> } else {
> >> delete rhs;
> > add the line
> > rhs = 0; // for exception safety incase clone throws
> > 
> > I thought I saved it before I attached it......
> > 
> >> rhs = lhs->clone();
> >> }
> >> }
> 
> You could do better. Wouldn't you rather assignment gave the strong
> guarantee in this case?
> 
> temp = lhs->clone();
> delete rhs;
> rhs = temp;

I guess.  It depends on how you look at it.

> 
> [gee, the raw pointer member rhs is starting to look like it ought to be a
> std::auto_ptr to me ;)]
> 

I don't know how serious you are but by pointers are different than
auto_ptr when ever one of my pointers is assigned to another a copy is
made.  Auto_ptr on the other hand transfers ownership.

---
Kevin Atkinson
kevinatk_at_[hidden]
http://metalab.unc.edu/kevina/