matplotlib categorical bar chart creates unwanted whitespace

Question

I have a dataframe that looks like this:

import numpy as np
import pandas as pd

location = list(range(1, 34))
location += [102, 172]
stress = np.random.randint(1,1000, len(location))
group = np.random.choice(['A', 'B'], len(location))

df = pd.DataFrame({'location':location, 'stress':stress, 'group':group})
df[['location', 'group']] = df[['location', 'group']].astype(str)

Note: location and group are both strings

I'm trying to create a a bar plot so that location (categorical) is on the x axis, and stress is the height of each bar. Furthermore, I want to color each bar with a different colour for each group

I've tried the following:

f, axarr = plt.subplots(1, 1)
axarr.bar(df['location'], df['stress'])
plt.xticks(np.arange(df.shape[0]) + 1, df['location'])
plt.show()

However, this produces:

I'm not sure why there are blank spaces between the end bars. I'm guessing its because of the 102 and 172 values in location, however, that column is a string so I'm expecting it to be treated as a categorical variable, with all bars placed next to each other regardless of location "value". I tried to correct for this by manually specifying the xtick location and labels but it didn't seem to work

Finally, is there a quick way to colour each bar by group without having to manually iterate over each unique group value?

Y. Luo Y. Luo · Accepted Answer · 2017-06-22T06:25:17

If your location is categorical data, don't make your bar plot with that. Use np.arange(df.shape[0]) to make the bar plot and set ticklabels later:

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

location = list(range(1, 34))
location += [102, 172]
stress = np.random.randint(1,1000, len(location))
group = np.random.choice(['A', 'B'], len(location))

df = pd.DataFrame({'location':location, 'stress':stress, 'group':group})
df[['location', 'group']] = df[['location', 'group']].astype(str)
f, axarr = plt.subplots(1, 1)
bars = axarr.bar(np.arange(df.shape[0]), df['stress'])
for b, g in zip(bars.patches, df['group']):
    if g == 'A':
        b.set_color('b')
    elif g == 'B':
        b.set_color('r')
plt.xticks(np.arange(df.shape[0]) + bars.patches[0].get_width() / 2, df['location'])
plt.setp(axarr.xaxis.get_ticklabels(), rotation=90)
plt.show()

Don't know if there is a concise way to set bar color in bulk. An iteration is not too bad...

matplotlib categorical bar chart creates unwanted whitespace

1 Answers