Seaborn gives wrong values on x-axis ticks?

Question

In the code below Matplotlib gives the correct range of 5.0 to 10.0, why is Seaborn different?

import matplotlib.pyplot as plt
import numpy as np
import seaborn as sns
from matplotlib import ticker

sns.set()

fig, (ax1, ax2) = plt.subplots(2)

x = np.linspace(5, 10)
y = x ** 2

sns.barplot(x, y, ax=ax1)
ax1.xaxis.set_major_locator(ticker.MultipleLocator(5))
ax1.xaxis.set_major_formatter(ticker.FormatStrFormatter('%.2f'))

ax2.bar(x, y, width = 0.1)
ax2.xaxis.set_major_locator(ticker.MultipleLocator(5))
ax2.xaxis.set_major_formatter(ticker.FormatStrFormatter('%.2f'))

plt.show()

Answer 1

Seaborn's barplot is a categorical plot. This means it places the bars at successive integer positions (0,1,...N-1). Hence, if you have N bars, the axis will range from -0.5 to N-0.5.

There is no way to tell seaborn to place the bars at different positions; but you can of course fake the labels to let it appear as such. E.g. to label every 5th bar with the value from x:

import matplotlib.pyplot as plt
import numpy as np
import seaborn as sns
from matplotlib import ticker

sns.set()

fig, ax = plt.subplots()

x = np.linspace(5, 10)
y = x ** 2

sns.barplot(x, y, ax=ax)
ax.xaxis.set_major_locator(ticker.FixedLocator(np.arange(0, len(x), 5)))
ax.xaxis.set_major_formatter(ticker.FixedFormatter(x[::5]))
ax.tick_params(axis="x", rotation=90)

plt.tight_layout()
plt.show()

Inversely, it is possible to plot categorical plots with matplotlib. To this end, one needs to plot strings.

ax.bar(x.astype(str), y)
ax.xaxis.set_major_locator(ticker.FixedLocator(np.arange(0, len(x), 5)))
ax.xaxis.set_major_formatter(ticker.FixedFormatter(x[::5]))
ax.tick_params(axis="x", rotation=90)

If you want a numerical bar plot, i.e. a plot where each bar is at the axis position of x, you would need to use matplotlib. This is the default case also shown in the question, where the bars range between 5 and 10. One should make sure to have the width of the bars smaller than the difference between successive x positions in this case.

ax.bar(x, y, width=np.diff(x).mean()*0.8)
ax.xaxis.set_major_locator(ticker.MultipleLocator(1))
ax.xaxis.set_major_formatter(ticker.FormatStrFormatter('%.2f'))
ax.tick_params(axis="x", rotation=90)