Bug summary

It seems a bug was introduced in Matplotlib 3.8, in which numeric labels for bar charts are no longer getting converted to strings. When trying to add a legend, this causes an AttributeError: 'int' object has no attribute 'startswith'.

Code for reproduction

import numpy as np
import matplotlib.pyplot as plt

plt.bar(np.arange(10), np.random.rand(10), label=1)
plt.legend() # fails

Actual outcome

  File /software/anaconda3/lib/python3.11/site-packages/matplotlib/axes/_axes.py:322 in legend
    handles, labels, kwargs = mlegend._parse_legend_args([self], *args, **kwargs)

  File /software/anaconda3/lib/python3.11/site-packages/matplotlib/legend.py:1361 in _parse_legend_args
    handles, labels = _get_legend_handles_labels(axs, handlers)

  File /software/anaconda3/lib/python3.11/site-packages/matplotlib/legend.py:1291 in _get_legend_handles_labels
    if label and not label.startswith('_'):

AttributeError: 'int' object has no attribute 'startswith'

Expected outcome

Legend should work and show "1" as the label (as it did in Matplotlib 3.7).

Additional information

• It seems this bug was introduced in 3.8.0 (it worked in 3.7.1).
• It's only plt.bar() that's affected; plt.plot() for example works fine.
• Converting manually to a string (label=str(1)) prevents the error.
• Possibly unrelated, but I'm wondering if it has anything to do with this change regarding the startswith conditional.

Operating system

Ubuntu

Matplotlib Version

3.8.0

Matplotlib Backend

Qt5Agg

Python version

3.11.3

Jupyter version

No response

Installation

conda