Instructions on how to compile Jupyter C++ Examples with Boost

[dmitryponv]

Since Building MLPack on Windows and running the basic examples doesn't require Boost,

I am finding a lot of compilation errors when trying to build the advanced Jupyter RL examples which require Boost.

The problem is getting the correct version of boost.

Once I solved most of the errors, including commenting out a duplicate declaration of "tcp::socket s;" on line 203 in gym/client.hpp

I got the error
cannot open file 'libboost_date_time-vc142-mt-s-x64-1_70.lib

The vc142 in the library name indicates it requires boost_1_76_0-msvc-14.2-64.exe

The library can be found here https://sourceforge.net/projects/boost/files/boost-binaries/

Installing this library and linking the boost and lib64-msvc-14.2 folders in my cmake file made it possible to build the examples.

Im giving a step by step answer here because the boost library will likely change in the future. If the devs dont update the instructions, this is how you get it installed.

Note that I tried getting boost v 1.88 (the most current one to date) and I got dozens of linker errors. So it really has to be the correct version.

Here is my cmakelists for reference. Note I put mlpack headers and binaries into my project and you can do the same with boost libraries if you wish. Not I can only build with Release right now, debug is throwing some errors.

---

    cmake_minimum_required(VERSION 3.16)
    
    project(MyMlpackApp)
    
    set(CMAKE_CXX_STANDARD 17)
    set(CMAKE_MSVC_RUNTIME_LIBRARY "MultiThreaded")
    
    # Boost configuration for static linking
    set(Boost_USE_STATIC_LIBS ON)
    set(Boost_USE_MULTITHREADED ON)
    set(Boost_USE_STATIC_RUNTIME ON)
    
    # Boost paths
    set(BOOST_ROOT "D:/Code/mlpackroot/boost")
    set(BOOST_INCLUDEDIR "${BOOST_ROOT}")
    set(BOOST_LIBRARYDIR "${BOOST_ROOT}/lib64-msvc-14.2")
    set(CMAKE_PREFIX_PATH "${BOOST_ROOT}")
    
    find_package(Boost REQUIRED COMPONENTS iostreams date_time)
    include_directories(${CMAKE_SOURCE_DIR}/include ${Boost_INCLUDE_DIRS})
    link_directories(${CMAKE_SOURCE_DIR}/libs "${BOOST_LIBRARYDIR}")
    
    add_executable(MyMlpackApp main.cpp)
    
    target_link_libraries(MyMlpackApp PRIVATE 
        lapack 
        armadillo 
        blas 
        cblas 
        ${Boost_LIBRARIES}
    )
    
    file(MAKE_DIRECTORY "${CMAKE_SOURCE_DIR}/build/Debug")
    file(MAKE_DIRECTORY "${CMAKE_SOURCE_DIR}/build/Release")
    file(GLOB DLL_FILES "${CMAKE_SOURCE_DIR}/dll/*")
    foreach(dll_file IN LISTS DLL_FILES)
      file(COPY "${dll_file}" DESTINATION "${CMAKE_SOURCE_DIR}/build/Debug")
      file(COPY "${dll_file}" DESTINATION "${CMAKE_SOURCE_DIR}/build/Release")
    endforeach()

[rcurtin]

Sorry for the very slow response here. I am not an expert on the gym-related code, but I can say that mlpack doesn't need to link against Boost anymore since the major version 4 release. Unfortunately I am not sure if the RL examples have been updated to the slightly-changed neural network API, but it should not be all that hard. I know this answer is not of the type "here is a fix for your issue!", and is instead "here are some pointers if you want to try to fix the issue!", but that's all I can offer bandwidth-wise at the moment. If you're willing to take a shot at fixing the issues and making it work, I can at least help out with any mlpack-related errors.