matplotlib-backend-notcurses

This python module allows you to show the plots generated by python's matplotlib in many modern and older terminals by using notcurses.

To install it, you will need to do one of the following

• $ pip install --user matplotlib-backend-notcurses
• clone this repo into your python's site-packages directory
• clone this repo and add the parent directory to sys.path or $PYTHONPATH

Then, configure matplotlib to use the module by either setting the environment variable MPLBACKEND to module://matplotlib-backend-notcurses or by initializing matplotlib as follows.

import matplotlib
matplotlib.use('module://matplotlib-backend-notcurses')
import matplotlib.pyplot as plt

Please make sure that you have the programs ncplayer and notcurses-info, both from notcurses, in your PATH.

If you've installed this module correctly, you can now use the following sample code to draw a plot in your terminal.

$ export MPLBACKEND='module://matplotlib-backend-notcurses'
$ python -i
>>> import numpy as np; import pandas as pd
>>> n = 10000
>>> df = pd.DataFrame({'x': np.random.randn(n),
                       'y': np.random.randn(n)})
>>> df.plot.hexbin(x='x', y='y', gridsize=20)
<plot is shown>

If you set your matplotlib to interactive mode via matplotlib.pyplot.ion() or by running python as python -i, non-empty figures are drawn on construction where possible. This allows you to use pandas' plot() calls directly, without calling plt.show(), and still enables you to manually construct and plt.show().

If your matplotlib is in non-interactive mode, you can construct your figures as usual, and then call plt.show() to render them to your terminal. This works from both a repl and when running scripts.

Figures are resized to the size of your terminal by default. If you'd rather control the sizing of figures manually, set the MPLBACKEND_NOTCURSES_SIZING environment variable to manual.

Internally, this backend is somewhat based on matplotlib's IPython support: it's a hybrid of image and GUI backend types. It works by using matplotlib's Agg backend to render the plot, and then calls notcurses's ncplayer to place the rendered image on your terminal. This means that plotting works as expected, but the image drawn to your terminal isn't interactive and animations aren't supported.

This is a port of my matplotlib-backend-kitty to notcurses tooling, which provides support for more terminal graphics protocols and reasonable auto-detection and fallbacks for these.