Repository navigation
[Bug]: ValueError ("cannot convert float NaN to integer") when trying to show horizontally-stacked subplots #31881
Description
Bug summary
Showing two horizontally-stacked subplots with a shared y axis and per-subplot titles fails if the y values are too large. This is a regression in matplotlib 3.11.0; it worked in 3.10.9.
Code for reproduction
fig, axes = plt.subplots(1,2, figsize=(12,5), sharey=True, tight_layout=True)
x = list(range(10))
y = [1e53 for _ in x]
for i, ax in enumerate(axes):
ax.set_title(f'Subplot {i}')
ax.plot(x,y)
plt.show()Actual outcome
Traceback (most recent call last):
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/backend_bases.py", line 1165, in _on_timer
ret = func(*args, **kwargs)
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/backends/backend_macosx.py", line 71, in callback_func
callback()
~~~~~~~~^^
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/backends/backend_macosx.py", line 93, in _draw_idle
self.draw()
~~~~~~~~~^^
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/backends/backend_macosx.py", line 56, in draw
super().draw()
~~~~~~~~~~~~^^
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/backends/backend_agg.py", line 438, in draw
self.figure.draw(self.renderer)
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/artist.py", line 94, in draw_wrapper
result = draw(artist, renderer, *args, **kwargs)
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/artist.py", line 71, in draw_wrapper
return draw(artist, renderer)
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/figure.py", line 3282, in draw
mimage._draw_list_compositing_images(
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^
renderer, self, artists, self.suppressComposite)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/image.py", line 133, in _draw_list_compositing_images
a.draw(renderer)
~~~~~~^^^^^^^^^^
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/artist.py", line 71, in draw_wrapper
return draw(artist, renderer)
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/axes/_base.py", line 3314, in draw
self._update_title_position(renderer)
~~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/axes/_base.py", line 3258, in _update_title_position
ax.yaxis.get_tightbbox(renderer) # update offsetText
~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/axis.py", line 1425, in get_tightbbox
ticks_to_draw = self._update_ticks()
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/axis.py", line 1344, in _update_ticks
major_locs = self.get_majorticklocs()
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/axis.py", line 1673, in get_majorticklocs
return self.major.locator()
~~~~~~~~~~~~~~~~~~^^
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/ticker.py", line 2299, in __call__
return self.tick_values(vmin, vmax)
~~~~~~~~~~~~~~~~^^^^^^^^^^^^
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/ticker.py", line 2307, in tick_values
locs = self._raw_ticks(vmin, vmax)
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/ticker.py", line 2237, in _raw_ticks
nbins = np.clip(self.axis.get_tick_space(),
~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/opt/miniconda3/envs/snewpy/lib/python3.13/site-packages/matplotlib/axis.py", line 2984, in get_tick_space
return int(np.floor(length / size))
ValueError: cannot convert float NaN to integer
Expected outcome
A plot with two (very boring) subplots; as in the following example generated with matplotlib 3.10.9:
Additional information
The bug only seems to happen if three factors come together:
- y becomes large enough to require scientific notation, with the 1e53 part shown at the top of the y axis. (When changing the code to y = [1e2 for _ in x], the error disappears.)
- A title is set for axes[1]. (When changing the code to if i ==0: ax.set_title(f"Subplot {i}"), the error disappears.)
- The subplots are initialised with sharey=True. (When changing it to False, the error disappears.)
The same code has worked correctly in matplotlib 3.10.9 and earlier. (See “Expected Outcome” section.)
I was also able to reproduce this bug on Ubuntu 24.04 in GitHub CI.
Operating system
macOS 15.7.7
Matplotlib Version
3.11.0
Matplotlib Backend
macosx
Python version
3.13.2
Jupyter version
No response
Installation
pip
Activity
ksunden commented on Jun 12, 2026
bisects to d772043 (from #31285)
@scottshambaugh do you have any insight into this?
Looks to me like somewhere we need a better fallback for bbox and/or need to include something more in the bbox, but not sure exactly where
Merge pull request #31885 from gaoflow/fix-31881-hidden-offset-title-nan
JostMigenda commented on Jun 15, 2026
That was very quick; many thanks to everyone involved! 💖
Backport PR #31885: Skip hidden y-axis offset text when positioning t…
Merge pull request #31893 from meeseeksmachine/auto-backport-of-pr-31…
fix(plot): exclude matplotlib 3.11.0, which crashes faceted plots