Repository navigation
Clarify that constrained_layout and tight_layout conflict with each other #17339
Description
Bug report
Bug summary
Not sure if this an expected behaviour, but the colorbar seems to be ignored when using tight_layout. Example taken from the "Constrained Layout Guide" (https://matplotlib.org/3.1.3/tutorials/intermediate/constrainedlayout_guide.html#colorbars).
Code for reproduction
arr = np.arange(100).reshape((10, 10))
fig, axs = plt.subplots(2, 2, figsize=(4, 4), constrained_layout=True)
for ax in axs.flat:
im = ax.pcolormesh(arr)
fig.colorbar(im, ax=axs, shrink=0.6)
plt.tight_layout()Actual outcome
Expected outcome
This is the outcome before calling plt.tight_layout()
Matplotlib version
- Operating system: Fedora 31
- Matplotlib version: 3.1.3 from conda defaults (3.2.1 from conda-forge behave similarly)
- Matplotlib backend (print(matplotlib.get_backend())): Qt5Agg
- Python version: 3.7.6
- Jupyter version (if applicable):
- Other libraries:
Activity
jklymak commented on May 6, 2020
Yes, thats expected, or at least understood. You can use constrained_layout instead...
Edit: tight_layout and constrained_layout are not compatible.
QuLogic commented on May 6, 2020
Maybe we should raise if calling tight_layout when the figure has constrained_layout enabled?
jklymak commented on May 6, 2020
Thats not particularly the problem here; constrained_layout is a bit of a red herring. tight_layout just doesn't know about the big colorbar so ignores it. In the code above constrained_layout=True is ignored because tight_layout gets precedence (simply because its older).
QuLogic commented on May 7, 2020
Does that generate a warning then?
ericpre commented on May 7, 2020
If tight_layout and constrained_layout are not compatible, it would useful to have it documented - I can do this at some point.
Regardless of using constrained_layout or not, can you confirm the following: if I understand correctly, the colorbar is ignored by tight_layout because it is an instances of matplotlib.axes._axes.Axes and not matplotlib.axes._subplots.AxesSubplot?
jklymak commented on May 7, 2020
Its says constrained_layout is to be used instead of tight_layout in quite a few places. They are not incompatible per se, just tight_layout cancels out constrained_layout.
We could raise a warning I suppose.
tacaswell commented on May 7, 2020
At least on master we do warn:
This limitation of tight_layout was one of the motivations for writing constrained layout. I think action items here are:
ericpre commented on May 7, 2020
Thanks for the explanations!
In the documentation, it is mentioned that both are alternatives but it is not clear, which one would be more useful. Even if there are many examples, it is not straightforward to work out which one would be more appropriate/robust. Maybe also add some simple guidelines or rule of thumb to the documentation?
5 remaining items
DOC: Recommend constrained_layout over tight_layout
DOC: Recommend constrained_layout over tight_layout (#29584)