Re: [boost] serialization as generialized I/O

From: Robert Ramey (ramey_at_[hidden])
Date: 2005-08-26 22:26:37

Neal Becker wrote:
> Looking again at mersenne_twister, one of the objections that has been
> raised to adding serialization support, is that there is already
> iostream
> I/O support.  It is in principle possible to leverage iostream to do
> serialization.
>
> Unfortunately, that approach is hardly efficient and rather ugly IMHO.
>
> But, what about the other way around?  If we have generic archive
> support, can we use it to eliminate the need for iostream?  I think
> so.
>
> In fact, consider this code (assuming serialization suport added to
> mersenne_twister):
> typedef  boost::mt19937 rng_t;
>
> int main() {
>  rng_t rng1;
>
>  std::ostringstream os;
>  boost::archive::text_oarchive oa(os, boost::archive::no_header);
>  oa << const_cast<const rng_t&>(rng1);
>
>  std::string st (os.str());
>
>  std::cout << rng1 << '\n';
>  std::cout << st << '\n';
> }
>
> As it is, the 2 outputs are almost identical.  There is only one
> difference: the text_oarchive (,no_header) output prints 2 leading
> '0' characters.
>
> Is there some way to suppress this preamble?
>
> If so, then we have a more elegant solution.  All we need to do is
> put in
> the serialization support and remove the iostream support, since
> serialization is just a generalization of i/o.
>
> _______________________________________________
> Unsubscribe & other changes:
> http://lists.boost.org/mailman/listinfo.cgi/boost