Re: [boost] [review] Review of Unordered library begins today December7

From: JOAQUIN LOPEZ MU?Z (joaquin_at_[hidden])
Date: 2007-12-11 14:25:04

----- Mensaje original -----
De: Daniel James <daniel_james_at_[hidden]>
Fecha: Martes, Diciembre 11, 2007 8:01 pm
Asunto: Re: [boost] [review] Review of Unordered library begins	today
December7
Para: boost_at_[hidden]

> Joaquín Mª López Muñoz wrote:
> > I think unordered containers are *not* allowed to rehash when 
> erasing:> see N2461, 23.1.3/8
> >
> >   "...Rehashing invalidates iterators, changes ordering between 
> elements,...">
> > and 23.1.3/12
> >
> >   "...The erase members shall invalidate only iterators and 
> references>   to the erased elements."
> >
> > so the conclusion is that erase can't rehash (incidentally, the 
> > implications of rehashing when erasing a range would be
> > difficult to cope with).
> >   
> 
> If a container implemented a rehash which doesn't invalidate 
> iterators 
> (which is certainly possible, if unlikely) then could it rehash? 
> (Of  course, such a container would have to maintain the order of the 
> elements).

I don't think so: even if you manage to implement an
iterator-conscious, order-preserving, sensible rehashing mechanism,
getting new memory can throw std::bad_alloc, which is not allowed
(only throwing from Hash or Pred permitted)--unless you catch
such exceptions internally... but this is rather artificial. My
opinion is that the writers of this piece of the C++0x standard
didn't consider rehashing as a permissible operation when
erasing.

Joaquín M López Muñoz
Telefónica, Investigación y Desarrollo