[boost] Re: new approach to smart ptrs -- no new()s is good news

From: mborgerding_at_[hidden]
Date: 2000-01-29 09:33:05

"stephen c. gilardi" <squeege-_at_[hidden]> wrote: 
original article:http://www.egroups.com/group/boost/?start=1980
> >  > if (left == this)
> >  >     delete ptr;
> 
> I could be wrong, but I don't think this is a correct test for 
> uniqueness. 

You are right, but the test (left == right) could safely be used in
either case.

I took the night off. You all have very good suggestions.  I am happy
there seems to be considerable interest in the idea.

I don't want to take too much time here, but I did want to quickly
address a couple of points that have come up since yesterday.

Re: checking ptr before delete called -- in the case where ptr is null,
checking "if (ptr)" will avoid a function call to delete.  I thought it
would be worth the cost overall. Side note: what proportion of non-null
to null pointers would need to be released before the "if (ptr)"
becomes less efficient?  I suspect it is > 1000:1.

Re: "this" vs. "0"  -- I don't have strong feelings either way.  It
seems that using circular linking results in an extremely small gain in
efficiency.  So small that *if* the approach requires a longer learning
curve, it is not worthwhile.  I would defer to popular opinion, and
tend toward whichever approach is considered more readable/maintainable
.  I am biased toward my original "0" approach, but I still do admire
the elegance of the "this" approach.

Re: the timing tests -- I will get to this on Monday.