[tuple] [serialization] serialize a tuple
How to serialize a tuple? Where are no boost/serialization/tuple.hpp, nor boost/tuple/tuple_serialization.hpp...
Am Monday 24 August 2009 18:13:02 schrieb Dmitry Vinogradov:
How to serialize a tuple? Where are no boost/serialization/tuple.hpp, nor boost/tuple/tuple_serialization.hpp...
if there is no builtin support for tuple (I don't know), have look at the serialization function of std::pair in serialization/utility.hpp to see how it's done. or in the documentation under "free function".
Stefan Strasser wrote:
Am Monday 24 August 2009 18:13:02 schrieb Dmitry Vinogradov:
How to serialize a tuple? Where are no boost/serialization/tuple.hpp, nor boost/tuple/tuple_serialization.hpp... if there is no builtin support for tuple (I don't know), have look at the serialization function of std::pair in serialization/utility.hpp to see how it's done. or in the documentation under "free function".
Thanks, I'll do it by myself. BTW, do you know the reason why tuple serialization is missing in the Boost library?
On Aug 24, 2009, at 2:23 PM, Dmitry Vinogradov wrote:
Stefan Strasser wrote:
Am Monday 24 August 2009 18:13:02 schrieb Dmitry Vinogradov:
How to serialize a tuple? Where are no boost/serialization/tuple.hpp, nor boost/tuple/tuple_serialization.hpp... if there is no builtin support for tuple (I don't know), have look at the serialization function of std::pair in serialization/ utility.hpp to see how it's done. or in the documentation under "free function".
Thanks, I'll do it by myself. BTW, do you know the reason why tuple serialization is missing in the Boost library?
IIRC the reason is (was?) that Boost.Tuple is implemented using a cons- list and can be most efficiently serialized using this implementation detail. Serializing any TR1-compatible tuple implementation means that you'd need to build around the get<N> templates, which is more work for the compiler. Alternately you can do a bunch of fancy Boost.Preprocessor tricks to make both work, but that seems like a lot of work. That being said, I'm all for adding support to Boost.Serialization for serializing Boost.Tuple even if it does use knowledge of the cons-list implementation, it would allow me to remove my own implementation of serialization for Boost.Tuple from quite a few projects. Of course boost::fusion::tuple provides tuple_size< >, which makes serializing it really easy, so maybe switching to boost::fusion::tuple is the easiest (best?) approach. -Nick
On Mon, Aug 24, 2009 at 2:38 PM, Nick Edmonds
On Aug 24, 2009, at 2:23 PM, Dmitry Vinogradov wrote:
Stefan Strasser wrote:
Am Monday 24 August 2009 18:13:02 schrieb Dmitry Vinogradov:
How to serialize a tuple? Where are no boost/serialization/tuple.hpp, nor boost/tuple/tuple_serialization.hpp...
if there is no builtin support for tuple (I don't know), have look at the serialization function of std::pair in serialization/utility.hpp to see how it's done. or in the documentation under "free function".
Thanks, I'll do it by myself. BTW, do you know the reason why tuple serialization is missing in the Boost library?
IIRC the reason is (was?) that Boost.Tuple is implemented using a cons-list and can be most efficiently serialized using this implementation detail. Serializing any TR1-compatible tuple implementation means that you'd need to build around the get<N> templates, which is more work for the compiler. Alternately you can do a bunch of fancy Boost.Preprocessor tricks to make both work, but that seems like a lot of work. That being said, I'm all for adding support to Boost.Serialization for serializing Boost.Tuple even if it does use knowledge of the cons-list implementation, it would allow me to remove my own implementation of serialization for Boost.Tuple from quite a few projects.
Of course boost::fusion::tuple provides tuple_size< >, which makes serializing it really easy, so maybe switching to boost::fusion::tuple is the easiest (best?) approach.
Boost.Fusion has functions to convert such things to things it can handle, and yes, Boost.Fusion is so much better for this stuff, I actually use it for functionality like this.
participants (4)
-
Dmitry Vinogradov -
Nick Edmonds -
OvermindDL1 -
Stefan Strasser