Re: [boost] [serialization] Proposal to add support for lexical_cast

From: Peter Dimov (pdimov_at_[hidden])
Date: 2005-07-13 12:27:27

Martin Wille wrote:

> Usage example:
>
>  // udt.hpp
>  struct udt { };
>  std::ostream &operator<<(std::ostream &, udt const &);
>  std::istream &operator>>(std::istream &, udt const &);
>
>  // someclass.hpp
>  #include "udt.hpp"
>  #include <boost/serialization.hpp>
>
>  struct someclass
>  {
>     template<typename Archive>
>     void serialize(Archive &ar, const unsigned int version)
>     {
>       ar & boost::serialization::lexical_cast<std::string>(my_udt);
>     }
>
>  private:
>     udt my_udt;
>  };

This puts the serialization support for udt at the wrong level. If you have 
twelve classes that have four udt members each, you'll have to repeat 
"boost::serialization::lexical_cast<std::string>()" 48 times. And 
vector<udt> will still not work.

It's better to define load+save for udt.

If you do that, you'll see that the proposed wrapper adds almost no value. 
The 'almost' part comes from track_never, which enables you to sidestep the 
usual problem of serializing temporaries. :-)

>  - Do you see a use case for using an "external" type other than
>    std::string with this wrapper, e.g
>      ar & boost::serialization::lexical_cast<int>(my_udt); // ?

I'm not sure I do.

>  - Assuming there are some positive responses about the usefulness
>    of the lexical_cast wrapper, would this addition be suitable for
>    a fast-track review (it is very tiny)?

Fast-track components need to already be in use as an implementation detail 
of an existing Boost library, I believe.