Boost serialization end of file

I serialize multiple objects into a binary archive with Boost. When reading back those objects from a binary_iarchive, is there a way to know how many objects are in the archi开发者_如何学编程ve or simply a way to detect the end of the archive ?

The only way I found is to use a try-catch to detect the stream exception. Thanks in advance.

---

I can think of a number of approaches:

1. Serialize STL containers to/from your archive (see documentation). The archive will automatically keep track of how many objects there are in the containers.

2. Serialize a count variable before serializing your objects. When reading back your objects, you'll know beforehand how many objects you expect to read back.

3. You could have the last object have a special value that acts as a kind of sentinel that indicates the end of the list of objects. Perhaps you could add an isLast member function to the object.

4. This is not very pretty, but you could have a separate "index file" alongside your archive that stores the number of objects in the archive.

5. Use the tellp position of the underlying stream object to detect if you're at the end of file:

Example (just a sketch, not tested):

std::streampos archiveOffset = stream.tellg(); 
std::streampos streamEnd = stream.seekg(0, std::ios_base::end).tellg();
stream.seekg(archiveOffset);

while (stream.tellp() < streamEnd)
{
    // Deserialize objects
}

This might not work with XML archives.

---

Do you have all your objects when you begin serializing? If not, you are "abusing" boost serialization - it is not meant to be used that way. However, I am using it that way, using try catch to find the end of the file, and it works for me. Just hide it away somewhere in the implementation. Beware though, if using it this way, you need to either not serialize pointers, or disable pointer tracking.

If you do have all the objects already, see Emile's answer. They are all valid approaches.

---

std::istream* stream_;
boost::iostreams::filtering_streambuf<boost::iostreams::input>* filtering_streambuf_;
...
stream_ = new std::istream(memoryBuffer_);
if (stream_) {
  filtering_streambuf_ = new boost::iostreams::filtering_streambuf<boost::iostreams::input>();
  if (filtering_streambuf_) {
    filtering_streambuf_->push(boost::iostreams::gzip_decompressor());
    filtering_streambuf_->push(*stream_);

    archive_ = new eos::portable_iarchive(*filtering_streambuf_);
  }
}

using zip when reading data from the archives, and filtering_streambuf have such method as

std::streamsize std::streambuf::in_avail()
Get number of characters available to read

so i check the end of archive as

bool    IArchiveContainer::eof() const {
    if (filtering_streambuf_) {
        return filtering_streambuf_->in_avail() == 0;
    }
    return false;
}

It is not helping to know how many objects are last in the archive, but helping to detect the end of them (i'm using eof test only in the unit test for serialization/unserialization my classes/structures - to make sure that i'm reading all what i'm writing)

---

Sample code which I used to debug the similar issue (based on Emile's answer) :

#include <fstream>
#include <iostream>
#include <boost/archive/binary_oarchive.hpp>
#include <boost/archive/binary_iarchive.hpp>

struct A{
    int a,b;
    template <typename T>
    void serialize(T &ar, int ){
        ar & a;
        ar & b;
    }
};


int main(){
    {
        std::ofstream ofs( "ff.ar" );
        boost::archive::binary_oarchive ar( ofs );
        
        for(int i=0;i<3;++i){
            A a {2,3};
            ar << a;
        }
        ofs.close();
    }

    {
        std::ifstream ifs( "ff.ar" );
        ifs.seekg (0, ifs.end);
        int length = ifs.tellg();
        ifs.seekg (0, ifs.beg);
    
        boost::archive::binary_iarchive ar( ifs );

        while(ifs.tellg() < length){
            A a;
            ar >> a;
            std::cout << "a.a-> "<< a.a << " and a.b->"<< a.b << "\n";
        }
    }
    return 0;
}

---

you just read a byte from the file.

If you do not reach the end,

backword a byte then.