Repository navigation
RuntimeError: adjustable='datalim' is not allowed when both axes are shared. #11416
Description
Bug report
Bug summary
Matplotlib 2.2.2 fails to plot the figure described below. The code works with Matplotlib 2.0.2.
The issue seems to be the combination of sharex=True, sharey=True and axis('equal').
Code for reproduction
import matplotlib.pyplot as plt
import numpy as np
nrows = 3
ncols = 2
fig, axes = plt.subplots(ncols=ncols, nrows=nrows, sharex=True, sharey=True)
n = 20
np.random.seed(1234)
data = np.random.uniform(size=(nrows, ncols, n, n))
for i in range(nrows):
for j in range(ncols):
ax = axes[i, j]
ax.imshow(data[i, j])
ax.axis("equal")
plt.show()Actual outcome
With Matplotlib 2.2.2, the figure is not drawn and the following exception is raised:
Traceback (most recent call last):
File "C:\ProgramData\Anaconda3\lib\site-packages\matplotlib\backends\backend_qt5.py", line 519, in _draw_idle
self.draw()
File "C:\ProgramData\Anaconda3\lib\site-packages\matplotlib\backends\backend_agg.py", line 433, in draw
self.figure.draw(self.renderer)
File "C:\ProgramData\Anaconda3\lib\site-packages\matplotlib\artist.py", line 55, in draw_wrapper
return draw(artist, renderer, *args, **kwargs)
File "C:\ProgramData\Anaconda3\lib\site-packages\matplotlib\figure.py", line 1475, in draw
renderer, self, artists, self.suppressComposite)
File "C:\ProgramData\Anaconda3\lib\site-packages\matplotlib\image.py", line 141, in _draw_list_compositing_images
a.draw(renderer)
File "C:\ProgramData\Anaconda3\lib\site-packages\matplotlib\artist.py", line 55, in draw_wrapper
return draw(artist, renderer, *args, **kwargs)
File "C:\ProgramData\Anaconda3\lib\site-packages\matplotlib\axes\_base.py", line 2546, in draw
self.apply_aspect()
File "C:\ProgramData\Anaconda3\lib\site-packages\matplotlib\axes\_base.py", line 1631, in apply_aspect
raise RuntimeError("adjustable='datalim' is not allowed when both"
RuntimeError: adjustable='datalim' is not allowed when both axes are shared.
Expected outcome
Drawn figure, no error, as with Matplotlib 2.0.2
Matplotlib version
- Operating system: Windows 7 64 bits
- Matplotlib version: 2.2.2
- Matplotlib backend (print(matplotlib.get_backend())): Qt5Agg
- Python version: 3.6
- Jupyter version (if applicable):
- Other libraries: Anaconda 5.2
Activity
jklymak commented on Jun 11, 2018
This was a purposeful change in #10033. However, it should have had an API change entry - our apologies.
I'm also not sure that your use case should error. I think we need to still think about all the sharing/datalim interactions to make sure we always do the right thing.... ping @efiring
efiring commented on Jun 11, 2018
This combination is internally consistent only in the special case where all shared Axes have identical box aspect ratios in screen space. This is a common case, but by no means universal. Rather than blocking datalim at the outset, it might be possible to check for the consistency condition in apply_aspect. If it is met, then apply_aspect might be able to proceed as if there were no sharing. But it can get very complicated, because axes A and B might share both x and y, but C might share only x, etc.
I'm not sure it is worth trying to untangle this web in order to relax the simple blanket restriction against combining sharing of both axes with adjustable datalim. Maybe the only sensible step in that direction would be to handle the single special case in which all sharing involves both x and y in a set of Axes in a single Figure.
Or maybe apply_aspect can be replaced with something based on a general constraint solver, which could raise an exception whenever it runs into a conflict.
bgriffen commented on Jun 25, 2018
I have the same problem as @nbud. It is fairly common to have say a 2x2 grid of images displayed via imshow and you want to preserve the x-y scale after a zoom event say (e.g. R,G,B in three panels then RGB in the fourth). I'll switch back for 2.0.2 in the mean time.
efiring commented on Jun 26, 2018
Have you considered using adjustable='box'?
1fish2 commented on Jul 16, 2018
matplotlib 1.5.3 and 2.2.2 seem in conflict over this, or maybe I just don't understand which way to go.
I'm updating a codebase from matplotlib 1.5.3 to 2.2.2. adjustable='box-forced' is no longer defined so I switched to 'box':
fig, axesList = plt.subplots(n_generation, sharey=True, sharex=True, subplot_kw={'aspect': 0.4, 'adjustable': 'box'})That works in matplotlib 2.2.2 but matplotlib 1.5.3 raises:
It'd be good for the new code to pass our Jenkins build before I merge in the change and update the shared pyenv to the new libraries, so I tried 'datalim'. Then matplotlib 2.2.2 raises the opposite error:
I don't see docs on these choices and their compatibility with shared axes.
Matplotlib version
jklymak commented on Feb 11, 2019
To me the error here is that ax.axis('equal') calls ax.set_aspect('equal', adjustable='datalim'). Why does it do that instead of the default 'box'?
efiring commented on Feb 11, 2019
I think this is just a matter of leaving long-standing behavior in place. This definition of 'equal' predates the Axes refactoring 6 years ago. (I think it originated in Matlab compatibility.) Most or all of the behavior of ax.axis() has been kept unchanged from early days. There are two arguments that offer variations on 'equal': 'scaled', and 'image'. Both of those use 'box' adjustable.
jklymak commented on Mar 3, 2019
Anyhow, the workaround is to call ax.set_aspect('equal'). We should probably deprecate ax.axis('equal'), since its hard to see why we need two ways to do the same thing. But I think this can wait for 3.2...
ImportanceOfBeingErnest commented on Aug 20, 2019
@timhoffm recently updated the docstring of axis("equal") and cohorts in #15032. Would that qualify to close this issue, or is there another action item?
timhoffm commented on Sep 7, 2019
#15032 only adds documentation that ax.axis('equal') is ax.set_aspect('equal', adjustable='datalim').
That does not help too much for a user with the above problem because his code does not mention datalim. And the error message does not mention aspect/equal. Any idea how to hint at a solution in the error message without assuming too much?
stormshawn commented on Oct 8, 2021
I agree with @timhoffm. I am getting the same error and my code doesn't mention datalim
jklymak commented on Oct 8, 2021
@stormshawn are you using axis('equal') instead of ax.set_aspect('equal')? Is there a compelling reason to do so?
stormshawn commented on Oct 8, 2021
@jklymak I was using set_aspect('equal', adjustable="datalim") and that didn't work so I used set_aspect('equal') and it compiled. Though I'm not sure if that is the solution I should use here.
jamesclarke22 commented on Dec 26, 2024
I am trying to share both X and Y axis but an error occurred when I used axs.set_aspect('equal'):
set_aspect(..., adjustable='datalim') or axis('equal') are not allowed when both axes are shared. Try set_aspect(..., adjustable='box')..
However, non of the option fixed the error. Any help would be very appreciable.