Compiling a program

Requirements

Matplotlib for C++ requires a working Python installation as well as Matplotlib. Python2.7 and Python3 (>= 3.6) have been tested, but other versions should work as well. In the linking process the exact version of Python to use can be specified by linking the according library.

On Unix it is recommended to install Python via the package manager to assert that all dependencies are installed properly.

<package-manager> install python3 python3-dev  # or -devel depending on the platform

If Python is installed from source problems in the linking may occur. How to resolve these is explained in the next section, or in this code-block.

Install matplotlib via pip

pip3 install matplotlib  # or pip for Python 2

Includes and Linking

The header matplotlibcpp.h depends on the Python header, Python.h, the corresponding Python library libpython, and on numpy/arrayobject.h. If not in the standard include paths, the paths to the header files, the path to the library, and the library itself have to be specified for the compiler using the options -I, -L and -l respectively. Note, that all Python constituents should be of the same Python version. Matplotlib for C++ supports both, Python 2.7 and Python 3 versions.

In detail:

  • The Python header Python.h

    The Python header comes with the Python installation. If it cannot be found on your system try installing the Python development packages. The location of this header has to be specified using the option -I.

    Typical locations:

    • Linux: /usr/local/include/python3.7
    • Mac: if installed with Homebrew /usr/local/Cellar/python/3.7.3/Frameworks/Python.framework/Versions/3.7/include/python3.7m

  • The Python library libpython*.so

    The program has to be linked against the compiled Python library. Depending on the Python version the name of the library differs, for Python 3.7 it is libpython3.7.so (or libpython3.7m.so). Then link the library by specifying -lpython3.7 (or -lpython3.7m).

    Additionally to the linking the location of the library must be specified if not installed in the usual directory. For Linux systems this is usually not necessary, for Mac however it mostly is. The location of the library has to be specified using the option -L.

    If Python has not been installed using the package manager (but e.g. from source) twofold problems with linking the library can occur. The first are missing dependencies of the Python library, these can be added via -lpthread -lutil -ldl. The second is that dynamic libraries have to be exported which is resolved by adding -Xlinker -export-dynamic.

    Typical locations:

    • Linux: Path usually already included
    • Mac: /usr/local/Cellar/python/3.7.3/Frameworks/Python.framework/Versions/3.7/lib

  • Numpy array numpy/arrayobject.h

    By default Matplotlib for C++ uses Numpy arrays. This requires the above header file. However it is possible to avoid this header by defining -DWITHOUT_NUMPY.

    • Linux: /usr/local/lib/python3.7/site-packages/numpy/core/include
    • Mac: If installed via Homebrew, same as for Linux.

Examples

On Linux with the GNU compiler g++ and C++11.

# using Python 2.7
g++ main.cpp -std=c++11 -I/usr/local/include/python2.7 \
  -I/usr/local/lib/python2.7/site-packages/numpy/core/include -lpython2.7

# using Python3.7 and no Numpy
g++ main.cpp -std=c++11 -DWITHOUT_NUMPY -I/usr/local/include/python2.7 -lpython2.7

On Mac with the GNU compiler g++ and C++14.

g++ main.cpp -std=c++14 \
 -I /usr/local/Cellar/python/3.7.3/Frameworks/Python.framework/Versions/3.7/include/python3.7m \
 -I /usr/local/lib/python3.7/site-packages/numpy/core/include \
 -L /usr/local/Cellar/python/3.7.3/Frameworks/Python.framework/Versions/3.7/lib \
 -lpython3.7

With exporting dynamic libraries and linking to all dependencies of the Python library on a Linux system:

g++ main.cpp -std=c++11 -I/usr/local/include/python3.7m \
  -I/usr/local/lib/python3.7/site-packages/numpy/core/include \
  -lpython3.7m \
  -lpthread -lutil -ldl \ # library dependencies
  -Xlinker -export-dynamic \ # export dynamic libraries