json VS Bitsery

Shouldn't your benchmark use a somewhat larger JSON? Something from https://github.com/boostorg/json/tree/develop/bench/data, for instance.

The mascot for the library appears to be Jason Voorhees. Should I be concerned?

You should be able to build and include boost json as a standalone subproject in CMake if you are using C++17. (Or also possible to use as header only lib)

It gets far more complicated with C++11, since you also need a ton of other boost modules there.

For more Details you can read the Readme of it. https://github.com/boostorg/json

https://github.com/boostorg/json/blob/f55bd4b85edd9b9b9b2d27fb49d66a990aa89001/include/boost/json/impl/object.hpp#L39

I'm gonna throw this out there - Boost.JSON comes with .natvis visualizers for all of its data structures, so you can inspect all of its types in the debugger and get nice insights: https://github.com/boostorg/json/blob/932b97e5ce899f3faebb7b7ab5a68b023131b77f/include/boost/json/json.natvis

Thanks for sharing https://github.com/fraillt/bitsery it seems more powerful but require quite a boilerplate code for even very simple composite stuff. With ByteStream you can totally forget about deep you are in the data-structure hierarchy.

Bitsery is the closest I've seen to it, but you have to use extensions (which i can't yet figure out the API for) to make it achieve byte for byte compatbility

If size and performance matter, then take a look at bitsery. * it might be 9x+ faster than cereal and 18x faster than protobuf. * size-wise, you might save 20-30% by default * on top of that you additional might opt-in into: * bit-level serialization control (e.g. if your values are in the range 1000-2000 it will take you 10bits, or use VLE) * backward-forward compatibility support * pointer support, including raw pointers with the ability to provide custom allocator. * and powerful extensions system, which allows you to further customize things in any way you want it :)

About on par with nlohmann is a binary serialization lib called Bitsery. It's easy to use, and will be a lot more performant than json. The downside is that the serialized data won't be human readable.