Re: [boost] [Review][bimap]

From: Joaquín Mª López Muñoz (joaquin_at_[hidden])
Date: 2007-02-26 12:43:44

Thorsten Ottosen ha escrito:

> Thorsten Ottosen wrote:
> > JOAQUIN LOPEZ MU?Z wrote:
> >
> >>----- Mensaje original -----
> >
> >
> >>So, the problem is far (IMHO) from simple and involves
> >>other important and as of yet not completely settled
> >>issues like move constructors and extended allocators
> >>(see http://www.open-
> >>std.org/jtc1/sc22/wg21/docs/papers/2006/n2045.html,
> >>problem #5).
> >
> >
> > Ok, so the problem is essentially
> > cause by the lack of contruction functions for the
> > allocator.
> >
> > And placement new + ~T() + allocate() + deallocate()
> > is not a viable approach to avoid using construct()/destroy()?
>
> JOAQUIN,
>
> Do you know why placement new is insufficient?

Hello Thorsten, excuse my skipping your previous mail, I didn't
notice it.

Umm... yes, the problem boils down to lack of extended construction
functions and could be in principle overcome by not using the allocator
and relying on placement new instead.

Let me consider this with care. I'll put this issue of delayed
construction
insertion in my "to study" list.

Best,

Joaquín M López Muñoz
Telefónica, Investigación y Desarrollo