Re: [boost] Re: compressed_pair requirements?

From: JOAQUIN LOPEZ MU?Z (joaquin_at_[hidden])
Date: 2005-05-27 15:48:02

----- Mensaje original -----
De: Peter Dimov <pdimov_at_[hidden]>
Fecha: Viernes, Mayo 27, 2005 7:47 pm
Asunto: Re: [boost] Re: compressed_pair requirements?

> Howard Hinnant wrote:
> 
> > 2.  Uninitialized is better because it is much faster, 
> especially if
> > you have containers of such elements.  You can always explicitly
> > initialize if you want.
> 
> Technically... you can't put an uninitialized (non-character) 
> scalar in a 
> container; copying an uninitialized scalar is undefined behavior. 
> 

A similar issue also renders the following snippet illegal
(unless someone has filed a DR):

typedef set<int> int_set; 
pair<int_set::iterator,int_set­::iterator> p;

According to 20.2.2.2, the default ctor of std::pair works
as if implemented like 

pair():first(int_set::iterator()),second(int_set::iterator()){}

and int_set::iterator(), being a singular value, cannot
be copied. A technical nitpick save for the fact that
I actually was once static asserted by an overzealous checked
STL implementation (don't remember which, I think it was
Comeau). IMHO, std::pair default ctor should be defined as

pair():first(),second(){}

Joaquín M López Muñoz
Telefónica, Investigación y Desarrollo