Re: [Boost-users] optional + serialize

From: Jasper, Alexander (alexander.jasper_at_[hidden])
Date: 2005-08-05 06:17:55

Hi Robert!

thanks for the reply. I've to agree with you that my solution to problem was an ugly hack.
It certainly is very inefficient and not exception safe. Furthermore it requires T::operator=.

I took a look at the file you mentioned. But unfortunately I lack the knowledge of boost internals and, more important, I'm unfamiliar with the highly advanced c++ stuff, like inplace new or however the collection stuff is done. Therefore I, unfortunately, cannot suggest a better solution. I guess I've to bear with that performance hit, for now. I hope s.o. with better boost knowledge than me will address this issue in a future version.

kind regards

Alex

> -----Ursprüngliche Nachricht-----
> Von: boost-users-bounces_at_[hidden]
> [mailto:boost-users-bounces_at_[hidden]]Im Auftrag von 
> Robert Ramey
> Gesendet: Donnerstag, 4. August 2005 01:49
> An: boost-users_at_[hidden]
> Betreff: Re: [Boost-users] optional + serialize
> 
> 
> 
> a) I've read your post
> b) I see your point that the serialization of optional 
> doesn't handle types 
> without default constructors
> c) I agree that it should be fixed.
> d) don't like the way you fixed it.  I'm sure should work but 
> I don't like 
> the usage of the heap.  For a better way to do this take a look at 
> collections_load_imp.hpp which adderss this problem in a 
> different context.
> 
> Its too late to fixe for 1.33.
> 
> Robert Ramey
> 
> Jasper, Alexander wrote:
> > Hi Jeff, All!
> >
> > thanks a lot for the reply. I assume the bug fix is in 1.33?
> >
> > I'm still having issues with boost.optinal and boost.serialize
> > working together (in 1.32).
> > Is for deserialization of boost::optional<T> T required to have
> > default ctor?.
> >
> > This prevents boost::optional<boost::posix_time::time_period> from
> > being serialized. The extraction code fails to compile. I would an
> > ugly way around this (please see my last email). Does anybody have
> > ideas for better solutions?
> >
> > I've attached a modified version of #include
> > <boost/serialization/optional.hpp> to this email, that contains code
> > changes to make to work with types w/o T having a default ctor. But
> > I'm really unhappy with that solution and I don't know what I broke
> > with this change...
> >
> > thanks in advance
> >
> > Alex
> >
> >> -----Ursprüngliche Nachricht-----
> >> Von: boost-users-bounces_at_[hidden]
> >> [mailto:boost-users-bounces_at_[hidden]]Im Auftrag von
> >> Jeff Garland
> >> Gesendet: Freitag, 29. Juli 2005 15:40
> >> An: boost-users_at_[hidden]
> >> Betreff: Re: [Boost-users] questions concerning optional + 
> date_time
> >> +serialize
> >>
> >>
> >> On Fri, 29 Jul 2005 15:21:38 +0200, Jasper, Alexander wrote
> >>> ...
> >>> Another thing is that I'm unable to serialize not_a_date_time
> >>> ptimes. The extractions fails here. I'm using the binary
> >> archive. Is
> >>> that supported? The documentation of date time only states
> >>> compatibility with text and XML archive...
> >>
> >> There was a recent (jun 21) bug-fix in date-time that was
> >> preventing correct
> >> serialization of special values -- even with text archives --
> >> so it's likely
> >> that's what you are seeing.  I haven't tried the binary
> >> archive, but there's
> >> no reason it shouldn't work.  I believe you should be able to
> >> just pickup the
> >> latest version of gregorian/greg_serialize.hpp and
> >> posix_time/time_serialize.hpp to try it out.
> >>
> >> Jeff
> >> _______________________________________________
> >> Boost-users mailing list
> >> Boost-users_at_[hidden]
> >> http://lists.boost.org/mailman/listinfo.cgi/boost-users
> >>
> >
> >
> >
> >
> >> _______________________________________________
> >> Boost-users mailing list
> >> Boost-users_at_[hidden]
> >> http://lists.boost.org/mailman/listinfo.cgi/boost-users 
> 
> 
> 
>