Seaborn heatmap...white out the bottom row and right column

Question

I have a heatmap in Seaborn via sns.heatmap. I now want to white out the bottom row and right column but keep the values.

import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd

np.random.seed(2021)
df = pd.DataFrame(np.random.normal(0, 1, (6, 4))
df = df.rename(columns = {0:"a", 1:"b", 2:"c", 3:"d"})
df.index = [value for key, value in {0:"a", 1:"b", 2:"c", 3:"d", 4:"e", 5:"f"}.items()]
sns.heatmat(df, annot = True)
plt.show

I thought I had to include a mask argument in my sns.heatmap call, but I am not having success giving a proper mask, and the mask removes the annotation. I also need to preserve the text indices of my data frame d. How can I get those cells whited out while preserving the text indices?

Answer 1

Here is an approach:

• use the original data for annotation (annot=data)
• create a "norm" using the original data, to be used for coloring
• create a copy of the colormap and assign an "over" color as "white"
• create a copy of the data, and fill the right column and lower row with a value higher than the maximum of the original data (np.inf can't be used, because then no annotation will be placed); use this copy for the coloring; seaborn will magically use the appropriate color for the annotation
• to use the dataframe's column and index names in the heatmap, just use sns.heatmap(..., xticklabels=df.columns, yticklabels=df.index)
• if you don't have a recent seaborn version installed, you might consider using one of matplotlib's standard colormaps, or create one via matplotlib.colors.ListedColormap(). Maybe cmap = ListedColormap(sns.color_palette('rocket', 256))?

In example code:

import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
import numpy as np
from copy import copy

np.random.seed(2021)
df = pd.DataFrame(np.random.normal(0, 1, (6, 4)), columns=[*"abcd"], index=[*"abcdef"])
data = df.to_numpy()
data_for_colors = data.copy()
data_for_colors[:, -1] = data.max() + 10
data_for_colors[-1, :] = data.max() + 10
norm = plt.Normalize(data[:-1, :-1].min(), data[:-1, :-1].max())
# cmap = sns.color_palette('rocket', as_cmap=True).copy()
cmap = copy(plt.get_cmap('RdYlGn'))
cmap.set_over('white')
sns.set_style('white')
sns.heatmap(data=data_for_colors, xticklabels=df.columns, yticklabels=df.index,
            annot=data, cmap=cmap, norm=norm)
plt.show()