Bug report

Bug summary

In version 3.2.2, plt.hist(df['class']) will produce a correct histogram where df['class'] is a series of strings. In version 3.3.0, _reshape_2D gives different output in this case, leading to slow computation and a graph with too many bars.

Code for reproduction

import seaborn as sns
import matplotlib.pyplot as plt
df = sns.load_dataset('titanic')
plt.hist(df['class'])

Actual outcome

This is under matplotlib 3.3.0:

This is _reshape_2D under 3.3.0:

Expected outcome

This is under matplotlib 3.2.2:

This is _reshape_2D under 3.2.2:

Can produce the same output for 3.3.0 by putting the series in a list but ideally I think the behavior should remain the same since this unexpectedly also breaks other function behavior relying on this (pandas.DataFrame.hist(), plt.hist())

Matplotlib version

• Operating system: macOS Mojave 10.14.6
• Matplotlib version: 3.3.0
• Matplotlib backend (print(matplotlib.get_backend())): module://ipykernel.pylab.backend_inline
• Python version: 3.7.4
• Jupyter version (if applicable):
• Other libraries: numpy: 1.18.1, pandas: 1.1.0, seaborn: 0.9.0

matplot lib installed via pip