Re: [boost] [serialization+date_time] Serialization Review: serialization of dates

From: JOAQUIN LOPEZ MU?Z (joaquin_at_[hidden])
Date: 2004-04-25 15:14:03

----- Mensaje original -----
De: Jeff Garland <jeff_at_[hidden]>
Fecha: Domingo, Abril 25, 2004 9:25 pm
Asunto: Re: [boost] [serialization+date_time] Serialization Review: serialization of dates

> On Sun, 25 Apr 2004 20:34:34 +0200, JOAQUIN LOPEZ MU?Z wrote
> 
> > Regarding the need to include "greg_serialize.hpp", I think
> > a better approach, from the point of view of the user, is
> > that serialization support is already embedded into
> > "boost/date_time/gregorian/gregorian_types.hpp" and
> > automatically provided if one of the archive headers
> >
> >...snip details... 
> >
> > See what I mean? For this to work with minimum
> > overhead, we'd need some forward headers for serialize,
> > save, load, etc. to be included by 
> > "boost/date_time/gregorian/gregorian.hpp".
> 
> Yes, I agree on the user perspecitive. I'm sorry I forgot to say 
> that in the
> first few emails.  I did say it here: 
> http://lists.boost.org/MailArchives/boost/msg63885.php
> 
> > I'm not sure whether this forward headers are already
> > provided by Robert's library.
> 
> Why would I need 'forward headers' for this? 
> 

Well, since serialization support is entirely
contained in function templates, you don't need
any actual definition till POI (I might be wrong here,
but I guess I'm not.) This can help reduce compilation
time if you're not going to serialize dates
at all. Matter of fact, seeing the includes
in your example

#include "boost/archive/basic_archive.hpp"
#include "boost/serialization/split_free.hpp"
#include "boost/serialization/string.hpp"

I think the two first are unnecessary (could you
try this out? thx)

Joaquín M López Muñoz
Telefónica, Investigación y Desarrollo