Bug summary

The first call to matshow() re-enables interactive mode after it has been disabled by matplotlib.interactive(False). In a Jupyter notebook, this has the result of a figure not intended for display being shown in the output.

Code for reproduction

from matplotlib.artist import Artist
from matplotlib.pyplot import matshow
import matplotlib

# Deactivate interactive mode, so the figure below isn't rendered.
matplotlib.interactive(False)

print(matplotlib.is_interactive())
# Interactive mode get's re-enabled and figure is shown in output.
fig = matplotlib.pyplot.matshow([[1, 2], [3, 4]]).figure
print(matplotlib.is_interactive())

Actual outcome

The first print() will (correctly) show False. By the second print(), interactive mode has been re-enabled and True is printed. The figure saved to variable fig is show in the output.

Expected outcome

I would expect that matplotlib.interactive(False) would have a permanent effect. Interactive mode should stay turned off until re-enabled explicitly. The documentation of matplotlib.interactive() does not mention the behaviour I'm witnessing.

Additional information

Screen shot of the Jupyter notebook I'm using to demonstrate the issue:

Operating system

macOS 14.7.7

Matplotlib Version

3.10.5

Matplotlib Backend

module://matplotlib_inline.backend_inline

Python version

3.13.3

Jupyter version

7.4.5

Installation

pip