How to ignore max values from pandas background_gradient

Question

I have this DataFrame:

import pandas as pd
df = pd.DataFrame({'2021-04': {'2021-04': 100.0, '2021-05': float("nan"), '2021-06': float("nan")},
 '2021-05': {'2021-04': 9.599326432568967, '2021-05': 100.0, '2021-06': float("nan")},
 '2021-06': {'2021-04': 7.952995602884856,
  '2021-05': 5.549312064243707,
  '2021-06': 100.0}})

I want to make a heatmap plot of it by line, but ignoring the max values, since they are much higher than the other values.

I also want to give the nans a light color, like white, but I got this part got right.

This is what I got so far:

df.style.background_gradient(cmap ='RdYlGn', axis=1)\
                .highlight_null(null_color='white')

Which produces this table:

How can I apply the gradient ignoring the values equal to 100?

Answer 1

You could use:

df.replace(100, np.nan).style.background_gradient(cmap ='RdYlGn', axis=1)\
                       .highlight_null(color='white')

If the "100" values are of no use.

Answer 2

If I understand correctly, here is a suggestion (highly inspired by @quant's answer here) :

import pandas as pd
import numpy as np

df = pd.DataFrame({'2021-04': {'2021-04': 100.0, '2021-05': float("nan"), '2021-06': float("nan")},
 '2021-05': {'2021-04': 9.599326432568967, '2021-05': 100.0, '2021-06': float("nan")},
 '2021-06': {'2021-04': 7.952995602884856,
  '2021-05': 5.549312064243707,
  '2021-06': 100.0}})

def highlight_max(data, color='white'):
    attr = f'background-color: {color}'
    if data.ndim == 1:
        is_max = data == data.max()
        return [attr if v else '' for v in is_max]
    else:
        is_max = data == data.max().max()
        return pd.DataFrame(np.where(is_max, attr, ''),
                            index=data.index, columns=data.columns)
    
out = (df.style.background_gradient(cmap ='RdYlGn', axis=1)
      .highlight_null(null_color='white')
      .applymap(lambda x: 'color: white' if pd.isnull(x) else 'color: black')
      .apply(highlight_max, axis=None)
     )

>>> print(out)