pyxi

#include <cstdint>
#include <string>
#include <vector>

#include "pyxi.hpp"

struct Person
{
    std::string name;
    uint32_t id;
};

class Group
{
public:
    void add(const std::string& name)
    {
        people_.push_back({ name, nextId_++ });
    }

    void serialize(pyxi::Serializer& ser)
    {
        ser << people_;
    }

    void deserialize(pyxi::deserialize& des)
    {
        des >> people_;
        nextId_ = people_.back().id + 1;
    }
private:
    std::vector<Person> people_;
    uint32_t nextId_ = 0;
}

int main()
{
    Group group;
    group.add("Guy");
    group.add("Man");

    std::vector<uint8_t> bytes = pyxi::align(pyxi::serialize(group), 8);

    // bytes
    // =======================
    // 00 00 00 00 00 00 00 02 // people_.size()         == 2
    // 00 00 00 00 00 00 00 03 // people_[0].name.size() == 3
    // 47 75 79                // people_[0].name        == "Guy"
    //          00 00 00 00 00
    // 00 00 00                // people_[0].id          == 0
    //          00 00 00 00 00
    // 00 00 03                // people_[1].name.size() == 3
    //          4D 61 6E       // people_[1].name        == "Man"
    //                   00 00
    // 00 00 00 00 00 01       // people_[1].id          == 1
    //                   00 00 // align(..., 8)
}

FAQ

Where are the serialize and deserialize functions for Person?

There aren't any. For standard layout objects like Person, the members can be inferred and handled like any other type. This works with nesting too, meaning you could have a vector of some other custom type as well.

Hang on, 8 bytes for the size of collections?

Currently, yes. This is because size_t is used to ensure the size fits. Don't worry, I'm planning on adding a utility object like Bits and Spare to handle custom size types.