mckinsey / qbstyles Goto Github PK

A new style registration mechanism was introduced in matplotlib 3.7 that simplifies distributing styles a lot:

Third-party packages can now distribute style files that are globally available as follows. Assume that a package is importable as import mypackage, with a mypackage/__init__.py module. Then a mypackage/presentation.mplstyle style sheet can be used as plt.style.use("mypackage.presentation").

See https://matplotlib.org/stable/users/prev_whats_new/whats_new_3.7.0.html#style-files-can-be-imported-from-third-party-packages, matplotlib/matplotlib#4781, matplotlib/matplotlib#24257

Tangentially related: #12

If I install matplotlib through anaconda, my project gives a type error for argument 2 in pyplot line 671. After uninstalling the conda version and reinstalling the pip version the themes work as expected.

I am using the Spyder IDE. The same error occurs when run from the anaconda console.

I followed global installation instruction. The styles do not apply on the first plot I run. I need to re-run plt.style.use(['qb-common', 'qb-dark']) for it to work.

Consider the following minimal working example which demonstrates unexpected behaviour when using qbstyles.mpl_style with plt.savefig():

import matplotlib.pyplot as plt
from qbstyles import mpl_style
mpl_style(dark=True)
a = [1,2,3,4,5]
plt.plot(a)
plt.savefig('filename.png')

The output image is missing most elements and does not have the qbstyle applied:

Same image with qbstyles activated and deactivated: x-axis and y-axis are not indexed properly in the first image

# LINE PLOT
def line_plot(ax):
    rng = np.random.RandomState(4)
    x = np.linspace(0, 10, 500)
    y = np.cumsum(rng.randn(500, 4), 0)
    ax.set_title('Line Graph')
    ax.set_xlabel('— Time')
    ax.set_ylabel('— Random values')
    ax.legend(['Bitcoin', 'Ethereum', 'Dollar', 'Oil'])
    ax.set_xlim([0, 10])
    ax.set_ylim([-20, 60])
    ax.plot(x, y)

fig, ax = plt.subplots(figsize=(16, 9))
line_plot(ax)

Problem:

• Some of the colours are hard to differentiate for people with colour blindness/deficiency (e.g. 1 & 2 in the cycle for red-green)

Proposed solution:

• Change cycle to have contrasting colours that are robust to different types of colour blindness. For example: http://bconnelly.net/posts/creating_colorblind-friendly_figures/img/cb-palette.png (more on this below)

Background material:

• If it works again, this used to be a great site: http://jfly.iam.u-tokyo.ac.jp/color/
  Archived version: https://web.archive.org/web/20190301225833/jfly.iam.u-tokyo.ac.jp/color/