Jupyter Matplotlibcpp

A header-only library for xeus-cling that lets you:

1. Build the data for plots in C++, and then
2. Use matplotlib to render them inline in a Jupyter notebook.

It's for people that love matplotlib and Jupyter, but want to program in C++.

Huh?

You can put this program into a Jupyter notebook:

#include <matplotlibcpp-jupyter.h>
#include <vector>

namespace plt = matplotlibcpp;

plt::jupyter::shown makePlot() {
    std::vector<double> t(1000);
    std::vector<double> x(t.size());
    
    for(size_t i=0; i < t.size(); i++) {
        t[i] = i / 100.0;
        x[i] = sin(2.0 * 3.14159265 * 1.0 * t[i]);
    }
    
    plt::xkcd();
    plt::plot(t,x);
    plt::title("AN ORDINARY SIN WAVE");
    return plt::jupyter::show();
}

and get this:

(see notebooks/xkcd-example.ipynb).

How?

This header-only library uses:

• xeus-cling, a C++ kernel for Jupyter
• matplotlibcpp, C++ bindings for matplotlib

It adds matplotlibcpp-jupyter.h, a header file that defines the struct and supporting methods to get xeus-cling and Jupyter to render the matplotlib results inline.

It doesn't write to any temp files or anything, it exports directly to an in-memory png buffer using the python method matplotlib.savefig() and then wraps that in a JSON object that Jupyter and xeus-cling understand. This is done via the C python API (most of the heavy lifting is via the Python interpreter).

How do I use it?

The easiest way:

docker run --rm -it -v ${PWD}/notebooks:/notebooks -p 8889:8889 andrewjjenkins/matplotlibcpp-jupyter:latest

Build-it-yourself:

git submodule init
docker build . -t matplotlibcpp-jupyter
docker run --rm -it -v ${PWD}/notebooks:/notebooks -p 8889:8889 matplotlibcpp-jupyter:latest

If you want to use it in your own project:

1. Copy matplotlibcpp-jupyter.h and matplotlibcpp.h into your project.

2. Install xeus-cling and dependencies (see environment.yml if you use conda/mamba/similar).

How do I make other plots? What's the API?

This project doesn't host the matplotlib C++ API.

Basically, look at all the examples and API in matplotlib, and everywhere you see plt::show() or plt::save(), replace it with plt::jupyter::show() instead.

Note that you must return a plt::jupyter::shown object for xeus-cling and Jupyter to render it (see the example above for a function signature).

LICENSE

MIT