Bug report

Bug summary

#17266 requires tick labels set by FixedFormatter to be equal to the ticks set by FixedLocator. FixedLocator is used in many cases where categorical data is plotted, e.g. bar and box plot. A way of removing all tick labels (while maintaining ticks and gridlines) is to do ax.set_xticklabels([]) or ax.set_yticklabels([]), by passing just an empty list. This works on 3.2.2, but raises a ValueError with 3.3.0rc1.

The new behavior is not mentioned in the "What's new?" section of 3.3.0rc1. I suspect this may break existing code, as it broke seaborn and this stackoverflow thread seems very popular.

Code for reproduction

# Example modified from bar example page

import matplotlib.pyplot as plt
import numpy as np

labels = ['G1', 'G2', 'G3', 'G4', 'G5']
men_means = [20, 34, 30, 35, 27]
women_means = [25, 32, 34, 20, 25]

x = np.arange(len(labels))  # the label locations
width = 0.35  # the width of the bars

fig, ax = plt.subplots()
rects1 = ax.bar(x - width/2, men_means, width, label='Men')
rects2 = ax.bar(x + width/2, women_means, width, label='Women')

ax.set_xticks(x)
ax.set_xticklabels([])

Actual outcome

# 3.3.0rc1
ValueError: The number of FixedLocator locations (5), usually from a call to set_ticks, does not match the number of ticklabels (0).

Expected outcome

Unset tick labels but not error out, or raise a deprecation warning that this won't be possible in a future release.

Matplotlib version

• Operating system: macOS
• Matplotlib version: 3.3.0rc1
• Matplotlib backend (print(matplotlib.get_backend())): MacOSX
• Python version: 3.8.3

Installed using pip