咨询无需Anaconda安装Xeus-cling的方法及依赖Anaconda的原因

Great questions—let's break this down clearly, since I know switching package managers when you're comfortable with pip can feel like a huge hassle. I’ve gone through this exact scenario before, so here’s what you need to know:

The short answer is that xeus-cling depends on a stack of compiled C++ libraries (like xeus, cppzmq, and xtl) that don’t play nicely with pip. Pip is built primarily for Python packages, but conda was designed to handle binary dependencies for all languages—including C++. The xeus team maintains pre-compiled conda packages that bundle all these tricky dependencies, so installation is one-and-done for most users. That’s why it’s the go-to in tutorials.

Since you want to stick with your existing pip setup, your best bet is to compile from source (the most reliable non-conda method) or use your system’s package manager if it offers pre-built xeus-cling packages. Below are step-by-step guides for major platforms:

1. Linux (Ubuntu/Debian-based) #

• First, install all required system dependencies and build tools:

  sudo apt update && sudo apt install build-essential cmake git libzmq3-dev libcppzmq-dev nlohmann-json3-dev libxtl-dev libxeus-dev jupyter-notebook
  

• Clone the official xeus-cling repository with recursive submodules enabled (this pulls in all dependent code):

  git clone --recursive <xeus-cling-official-repo-url>
  cd xeus-cling
  

• Compile and install (replace /usr/local with ~/.local to install in your user directory and avoid sudo):

  mkdir build && cd build
  cmake .. -DCMAKE_INSTALL_PREFIX=/usr/local
  make -j$(nproc)  # Uses all CPU cores for faster compilation
  sudo make install
  

• Register the Jupyter kernel:

  sudo jupyter kernelspec install /usr/local/share/jupyter/kernels/xcpp*
  

• Verify success: Run jupyter kernelspec list—you should see entries for xcpp11, xcpp14, xcpp17.

2. macOS (Using Homebrew) #

• Install dependencies via Homebrew (install Homebrew first if you don’t have it):

  brew install cmake git zmq cppzmq nlohmann-json xtl xeus jupyterlab
  

• Clone the repo with recursive submodules:

  git clone --recursive <xeus-cling-official-repo-url>
  cd xeus-cling
  

• Compile and install using Homebrew’s prefix to keep things organized:

  mkdir build && cd build
  cmake .. -DCMAKE_INSTALL_PREFIX=$(brew --prefix)
  make -j$(sysctl -n hw.ncpu)
  make install
  

• Register the kernel for your user:

  jupyter kernelspec install --user $(brew --prefix)/share/jupyter/kernels/xcpp*
  

3. Windows (Using MSYS2) #

Windows is trickier, but MSYS2 provides a Unix-like environment that simplifies compilation:

• Install MSYS2 (follow the setup guide to update the package database first)
• Open the MinGW-w64 x86_64 terminal and install dependencies:

  pacman -S base-devel git cmake mingw-w64-x86_64-gcc mingw-w64-x86_64-zmq mingw-w64-x86_64-cppzmq mingw-w64-x86_64-nlohmann-json mingw-w64-x86_64-xtl mingw-w64-x86_64-xeus mingw-w64-x86_64-jupyter-notebook
  

• Clone and compile:

  git clone --recursive <xeus-cling-official-repo-url>
  cd xeus-cling
  mkdir build && cd build
  cmake .. -G "MinGW Makefiles" -DCMAKE_INSTALL_PREFIX=/mingw64
  mingw32-make -j$(nproc)
  mingw32-make install
  

• Register the kernel:

  jupyter kernelspec install /mingw64/share/jupyter/kernels/xcpp*
  

Key Notes #

• If CMake throws errors about missing dependencies, double-check that all required libraries are installed (the commands above should cover most cases, but distro-specific variations might exist).
• After installation, launch Jupyter Notebook/Lab—you’ll see the C++ kernels (xcpp11/14/17) available to create new notebooks with.
• You don’t need to touch Anaconda at all for this workflow; your existing pip-managed Python environment will work seamlessly with the new kernel.

内容的提问来源于stack exchange，提问作者pg2455