libmsgpack

This is a library I yanked from an older project and decided to release in case it's helpful to someone.

Right now, it depends on a single MSVC header (intrin.h) for the declarations of _byteswap_ushort, _byteswap_ulong and _byteswap_uint64 since I had a surprisingly hard time getting MSVC to compile the byte swap into a single bswap instruction otherwise. Everything else is handled internally and the library is completely OS agnostic.

Functionality

Supports automatic encoding/decoding of most non-variable extension MessagePack types, except Float32/64 simply because the original project had no need for it.

All fixed sized types are decoded from the byte stream by calling MessagePack.decode_single( ) and inspecting the returned MPDecodeResult value.

Usage

#include <assert.h>
#include <mp.hpp>

int main() {
    constexpr auto buffer_size = 0x1000lu;
    const auto buffer = new mp::mp_u8[ buffer_size ];

[1] mp::MessagePack mpack { };

[2] mpack.initialize_streams(
        0,
        buffer_size,
        buffer,
        reinterpret_cast< stream::MemoryReader >( &memcpy ),
        reinterpret_cast< stream::MemoryWriter >( &memcpy )
    );

[3] mpack.write_u8( 0xa );
[4] mpack.write_uint( 0xa );

[5] mp::MPDecodeResult dr{ mpack.decode_single (  ) };

[5] assert( dr.marker == mp::MPMarker::Uint8 ); // true, because the first write was not opportunistic.

[6] dr = mpack.decode_single( );

[6] assert( dr.marker == mp::MPMarker::PosFixInt ); // true, because the second write was opportunistic; it chose the smallest representation.
}

1. Create a new MessagePack object.
2. Initialize the internal streams (read/write) with a buffer of buffer_size bytes. Also pass the memory reading/writing functions -- note both streams will utilize the same buffer. This is required.
3. Write a 1 byte unsigned integer to the stream. The total number of bytes written here is two: one for the marker and the other for the value itself.
4. Write a 1 byte unsigned integer to the stream again, however, this time the encoder will pick the smallest representation which is a PosFixInt in this case, resulting in a single byte written.
5. Decode the result, ensure we get the expected marker back.
6. Decode another item, check again we got what we expect.

Tests

This library is missing a more thorough test suite besides the simple unit tests at tests.cpp as well as fuzzing. I plan on adding it eventually, but PRs are welcome.