从Boost object_pool构造的指针的C ++ Boost二进制序列化

Question

我的应用程序有一个类“MyClass”。它的对象是从 Boost Object_pool 构建的。

我需要通过 Boost Binary Serialization 对此类对象进行序列化/反序列化。

用于序列化 -

我从池中取出一个指针，进行一些操作并通过 Boost 二进制序列化对其进行序列化。到目前为止一切都很好。

用于反序列化 -

我获取该序列化缓冲区并使用 Boost 二进制序列化对其进行反序列化。反序列化成功发生，但在此过程中，新内存正在通过非对象池构造的 Boost 序列化机制分配。
如果我在反序列化之前从池中分配指针，则反序列化会再次在堆上分配内存，并且池中的内存轨迹会丢失。因此，我将无法重用由 boost 序列化机制分配的内存，因为它不是从池中构造的，因此无法返回给对象池。

#include <boost/archive/binary_iarchive.hpp>
#include <boost/archive/binary_oarchive.hpp>
#include <boost/serialization/binary_object.hpp>
#include <boost/pool/object_pool.hpp>
#include <iostream>
#include <sstream>
#include <string>

class MyClass
{
  public :
    friend class boost::serialization::access;
    MyClass():data(9)
    {
      std::cout << std::endl << "MyClass()" << std::endl ;
    }

template<class Archive>
  void serialize(Archive & ar, const unsigned int version)
  {
    ar & data;
  }

~MyClass()
{
}

private :
  int data ;

};


int main()
{
  try
  {

    boost::object_pool<MyClass> pool ;

    // Get object from the Pool.
    MyClass *b = pool.construct();

    // Serialization

    std::stringbuf strbuf;
    boost::archive::binary_oarchive oa( strbuf ) ;
    oa << b ;

    // Deserilaiztion

    //MyClass *a = pool.construct() ;
    MyClass *a ;

    std::stringbuf strbufr(strbuf.str()) ; 
    boost::archive::binary_iarchive ia( strbufr ) ;

     // This calls the constructor and allocates memory for the pointer.
     // If we allocates this from the pool, new memory is still allocated     
     // and we loose track of the pointer fetched from the pool.   
     ia >> a;
  }
 catch ( boost::archive::archive_exception &e )
 {
   std::cout << std::endl << e.what() << std::endl ;
 }

}

我需要什么 -

我想将缓冲区反序列化为从池中分配的指针。

翻阅Boost序列化文档，发现Binary object wrappers的概念，却导致编译错误。

主要功能：-

int main()
{
  try
  {

    boost::object_pool<MyClass> pool ;

    MyClass *b = pool.construct();

    std::stringbuf strbuf;
    boost::archive::binary_oarchive oa( strbuf ) ;
    oa << boost::serialization::binary_object(b, sizeof(MyClass) ) ;

    MyClass *a = pool.construct() ;

    std::stringbuf strbufr(strbuf.str()) ;
    boost::archive::binary_iarchive ia( strbufr ) ;

    ia >> boost::serialization::binary_object(b, sizeof(MyClass) );
  }
  catch ( boost::archive::archive_exception &e )
  {
    std::cout << std::endl << e.what() << std::endl ;
  }

}

它给出了以下编译错误：-

serializationCpy.cpp: In function ‘int main()’:  
serializationCpy.cpp:56:8: error: no match for ‘operator>>’ (operand types are ‘boost::archive::binary_iarchive’ and ‘boost::serialization::binary_object’)
 ia >> boost::serialization::binary_object(b, sizeof(MyClass) );  
    ^
serializationCpy.cpp:56:8: note: candidate is:
In file included from /home/boost_root/include/boost/archive/detail/common_iarchive.hpp:23:0,
             from /home/boost_root/include/boost/archive/basic_binary_iarchive.hpp:30,
             from /home/boost_root/include/boost/archive/binary_iarchive_impl.hpp:21,
             from /home/boost_root/include/boost/archive/binary_iarchive.hpp:20,
             from serializationCpy.cpp:1:
/home/boost_root/include/boost/archive/detail/interface_iarchive.hpp:67:15: note: Archive& boost::archive::detail::interface_iarchive<Archive>::operator>>(T&) [with T = boost::serialization::binary_object; Archive = boost::archive::binary_iarchive]
 Archive & operator>>(T & t){
           ^
/home/boost_root/include/boost/archive/detail/interface_iarchive.hpp:67:15: note:   no known conversion for argument 1 from ‘boost::serialization::binary_object’ to ‘boost::serialization::binary_object&’  

将序列化与对象池一起使用是否正确？
非常感谢您的帮助。

Answer 1

来源于this的回答，我觉得应该是

ia >> boost::serialization::make_binary_object(a, sizeof(MyClass));

或

ia & boost::serialization::make_binary_object(a, sizeof(MyClass));

而不是ia >> boost::serialization::binary_object(b, sizeof(MyClass) );

修正其他错别字，它会变成：

#include <boost/archive/binary_iarchive.hpp>
#include <boost/archive/binary_oarchive.hpp>
#include <boost/serialization/binary_object.hpp>
#include <boost/pool/object_pool.hpp>
#include <iostream>
#include <sstream>

class MyClass
{
public:
    friend class boost::serialization::access;
    MyClass(int data) :data(data)
    {
        std::cout << "\nMyClass(" << data << ")\n";
    }

    MyClass()
    {
        std::cout << "\nMyClass()\n";
    }

    template<class Archive>
    void serialize(Archive & ar, const unsigned int version)
    {
        ar & data;
    }

    void printData()
    {
        std::cout << "Data = " << data << "\n";
    }
private:
    int data;
};

int main()
{
    try
    {
        boost::object_pool<MyClass> pool;

        MyClass *b = pool.construct(6);

        std::stringbuf strbuf;
        boost::archive::binary_oarchive oa(strbuf);
        oa << boost::serialization::binary_object(b, sizeof(MyClass));

        MyClass *a = pool.construct();

        std::stringbuf strbufr(strbuf.str());
        boost::archive::binary_iarchive ia(strbufr);

        ia >> boost::serialization::make_binary_object(a, sizeof(MyClass));

        a->printData();
    }
    catch (boost::archive::archive_exception &e)
    {
        std::cout << std::endl << e.what() << std::endl;
    }
    return 0;
}

输出：

MyClass(6)

MyClass()
Data = 6

DEMO

Answer 2

我建议创建一个value-wrapper 或使用带有自定义序列化逻辑的自定义智能指针。

binary_object 方式可以工作，但仅适用于 POD 数据类型。