[BUG]: Weirdness with `center` when using non-default colormaps

I have code that looks like this:

import matplotlib
import seaborn as sns

A = np.eye(2)

fig, ax = plt.subplots(figsize=(10,10))
cmap = matplotlib.colors.ListedColormap(["white", "black"])
ax = sns.heatmap(A, cmap=cmap, ax=ax, center=0, cbar=False)

This code produces a heatmap that looks like this:

However, when I change center to None, I correctly get the heatmap that I want:

And, interestingly, when center is 0 and cmap = matplotlib.colors.ListedColormap(["white", "white", "black"]), I get the correct heatmap again.

I’m not familiar enough with the heatmap API to know whether this is a bug, but it seems like it might be?

[mwaskom]

Good to know. So I was wrong, there isn’t a problem interpolating a discrete colormap (good!) and the “weirdness” arises from using center with binary (i.e. 0, and 1) data.

I think we can consider this resolved, but @loftusa if you think better documentation of the center parameter would have avoided this confusion I’d be happy to consider them.

[jhncls]

Here is some example code showing how center works. It can work perfectly with a bicolored colormap, although if can be harder to understand.

import seaborn as sns
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.colors import ListedColormap

x = np.linspace(-np.pi / 2, np.pi / 2, 100)
y = np.linspace(-np.pi / 2, np.pi / 2, 100)
x, y = np.meshgrid(x, y)
A = np.sin(x ** 2 + y ** 2) ** 2
# cmap = ListedColormap(['white', 'black'])
cmap = 'coolwarm'

fig, axes = plt.subplots(1, 4, figsize=(16, 4))
for ax, center in zip(axes.flat, [0, 0.2, 0.5, 0.9]):
    sns.heatmap(A, cmap=cmap, ax=ax, center=center, cbar=True, cbar_kws={'ticks': np.arange(0, 1, 0.1)})
    ax.axis('off')
    ax.set_title(f'center={center:g}')
plt.tight_layout()
plt.show()

With only two colors:

So, setting center=0 gets what you asked for.