FIX: tight_layout having negative width axes #10915

jklymak · 2018-03-29T04:05:16Z

PR Summary

fig.tight_layout() could cause axes to flip sign if the title was too big, and cause the whole thing to Error if the title was really big. This PR adds a check that the axes width and height is still greater than zero after the margins have been set.

It would be nice to just collapse the axes when this happens, but thats actually a bit of work to fake because subplots_adjust 's space parameter is specified as a fraction of average axes width (not the way I'd have expressed this parameter, but...)

Discussion in #4413 suggested that something could be done to ignore the title, but otherwise compute the tight_layout. I think that's theoretically possible, but if someone has a really long title they should probably shorten it or adjust the spacing manually rather than trying to make tight_layout do something magical.

import matplotlib.pyplot as plt

fig, axs = plt.subplots(1, 2, figsize=(6,2))
leng = 95
axs[0].set_title('A'+('a' * leng))
axs[0].set_yticks([])
axs[1].set_title('B' +('b' * leng))
axs[1].set_yticks([])
axs[0].set_xlabel('title length %d' % leng)
fig.tight_layout()
plt.show()

For the second and third plots, a warning is emitted...

PR Checklist

Has Pytest style unit tests
Code is PEP 8 compliant

anntzer · 2018-03-29T05:32:01Z

lib/matplotlib/tight_layout.py

        h_axes = (1 - margin_right - margin_left - hspace * (cols - 1)) / cols
-        kwargs["wspace"] = hspace / h_axes
+        if h_axes < 0.:


0 without dot

tacaswell · 2018-03-31T19:41:29Z

lib/matplotlib/tight_layout.py

+        if v_axes < 0:
+            warnings.warn('tight_layout cannot make axes height small enough '
+                          'to accomodate all axes decorations')
+            kwargs["hspace"] = 0.01


This and line 200 above fall back to different numbers, which was intended?

Thanks! Fixed.

tacaswell · 2018-03-31T19:46:23Z

Is there a way to test that the axis is not flipped? May check that everything in transAxes is positive?

tacaswell · 2018-03-31T19:46:38Z

👍 to behavior change.

jklymak · 2018-03-31T20:07:16Z

@tacaswell for the test given? I’ve looked at the result (as above) and it’s definitely not flipped. It’s hard to see how someone could change the code to make it flip in the test and keep the warning.

timhoffm

Test is good enough for me.

jklymak · 2018-04-01T04:50:44Z

It seems that somehow we have introduced a code-cov test drop...

QuLogic · 2018-04-01T06:57:55Z

lib/matplotlib/tests/test_tightlayout.py

+        assert len(w) == 1
+
+
+def test_big_decorators_verticaltal():


verticaltal?

QuLogic · 2018-04-01T06:58:03Z

lib/matplotlib/tests/test_tightlayout.py

+
+
+def test_big_decorators_horizontal():
+    "Test that warning emited when xlabel too big"


emitted (and in other places)

QuLogic · 2018-04-01T06:58:51Z

lib/matplotlib/tests/test_tightlayout.py

+        assert len(w) == 1
+
+
+def test_big_decorators_horizontal():


codecov drop is because this test name is the same.

QuLogic · 2018-04-01T06:59:25Z

lib/matplotlib/tests/test_tightlayout.py

+        assert len(w) == 1
+
+
+def test_big_decorators_vertical():


Probably the same name as before too if you fix the other's typo.

QuLogic · 2018-04-01T07:00:35Z

lib/matplotlib/tight_layout.py

+        margin_left = 0.4999
+        margin_right = 0.4999
+        warnings.warn('The left and right margins cannot be made large '
+                      'enough to accomodate all axes decorations. ')


accommodate (and below)

jklymak · 2018-07-08T17:02:14Z

@meeseeksdev backport to v2.2.x

…n-v2.2.x Backport PR #10915 on branch v2.2.x

Matplotlib emits errors like the following under specific canvas layouts: ```console Traceback (most recent call last): File "/opt/ros/kinetic/lib/python2.7/dist-packages/rqt_plot/data_plot/mat_data_plot.py", line 107, in resizeEvent self.figure.tight_layout() File "/usr/lib/python2.7/dist-packages/matplotlib/figure.py", line 1756, in tight_layout self.subplots_adjust(**kwargs) File "/usr/lib/python2.7/dist-packages/matplotlib/figure.py", line 1612, in subplots_adjust self.subplotpars.update(*args, **kwargs) File "/usr/lib/python2.7/dist-packages/matplotlib/figure.py", line 230, in update raise ValueError('bottom cannot be >= top') ValueError: bottom cannot be >= top ``` Current patch catches and suppresses that exception. References: - matplotlib/matplotlib#10915 - ros-visualization#35

* Fix #35 - "bottom cannot be >= top" matplotlib error Matplotlib emits errors like the following under specific canvas layouts: ```console Traceback (most recent call last): File "/opt/ros/kinetic/lib/python2.7/dist-packages/rqt_plot/data_plot/mat_data_plot.py", line 107, in resizeEvent self.figure.tight_layout() File "/usr/lib/python2.7/dist-packages/matplotlib/figure.py", line 1756, in tight_layout self.subplots_adjust(**kwargs) File "/usr/lib/python2.7/dist-packages/matplotlib/figure.py", line 1612, in subplots_adjust self.subplotpars.update(*args, **kwargs) File "/usr/lib/python2.7/dist-packages/matplotlib/figure.py", line 230, in update raise ValueError('bottom cannot be >= top') ValueError: bottom cannot be >= top ``` Current patch catches and suppresses that exception. References: - matplotlib/matplotlib#10915 - #35 * renamed to `safe_tight_layout`, inverted logic to avoid duplicating the call Co-authored-by: Dirk Thomas <dirk-thomas@users.noreply.github.com>

FIX: tight_layout having negative width axes

491c388

jklymak added this to the v3.0 milestone Mar 29, 2018

anntzer reviewed Mar 29, 2018

View reviewed changes

jklymak mentioned this pull request Mar 29, 2018

Long axis title alters xaxis length and direction with plt.tight_layout() #4413

Closed

jklymak force-pushed the fix-tight-layout-too-small branch 2 times, most recently from fbe12fa to 6e08b02 Compare March 29, 2018 15:39

ImportanceOfBeingErnest mentioned this pull request Mar 30, 2018

tight_layout + lots of subplots + long ylabels inverts yaxis #8062

Closed

tacaswell reviewed Mar 31, 2018

View reviewed changes

timhoffm approved these changes Apr 1, 2018

View reviewed changes

jklymak force-pushed the fix-tight-layout-too-small branch from 6e08b02 to 4730719 Compare April 1, 2018 04:00

QuLogic reviewed Apr 1, 2018

View reviewed changes

TST: tight_layout having negative width axes

abf260f

jklymak force-pushed the fix-tight-layout-too-small branch from 4730719 to abf260f Compare April 1, 2018 15:00

dstansby approved these changes Apr 7, 2018

View reviewed changes

efiring merged commit e2a0813 into matplotlib:master Apr 7, 2018

jklymak deleted the fix-tight-layout-too-small branch April 9, 2018 17:11

lumberbot-app bot pushed a commit that referenced this pull request Jul 8, 2018

Backport PR #10915: FIX: tight_layout having negative width axes

6fe8965

lumberbot-app bot mentioned this pull request Jul 8, 2018

Backport PR #10915 on branch v2.2.x #11598

Merged

jklymak mentioned this pull request Jul 8, 2018

Crash when interactively adding a number of subplots #11123

Closed

tacaswell added a commit that referenced this pull request Jul 8, 2018

Merge pull request #11598 from matplotlib/auto-backport-of-pr-10915-o…

cf526b1

…n-v2.2.x Backport PR #10915 on branch v2.2.x

bergercookie mentioned this pull request Apr 3, 2020

Fix #35 - "bottom cannot be >= top" matplotlib error ros-visualization/rqt_plot#52

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

FIX: tight_layout having negative width axes #10915

FIX: tight_layout having negative width axes #10915

jklymak commented Mar 29, 2018 •

edited

Loading

anntzer Mar 29, 2018

tacaswell Mar 31, 2018

jklymak Apr 1, 2018

tacaswell commented Mar 31, 2018

tacaswell commented Mar 31, 2018

jklymak commented Mar 31, 2018

timhoffm left a comment

jklymak commented Apr 1, 2018

QuLogic Apr 1, 2018

QuLogic Apr 1, 2018

QuLogic Apr 1, 2018

QuLogic Apr 1, 2018

QuLogic Apr 1, 2018

jklymak commented Jul 8, 2018



		def test_big_decorators_horizontal():
		"Test that warning emited when xlabel too big"

FIX: tight_layout having negative width axes #10915

FIX: tight_layout having negative width axes #10915

Conversation

jklymak commented Mar 29, 2018 • edited Loading

PR Summary

PR Checklist

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tacaswell commented Mar 31, 2018

tacaswell commented Mar 31, 2018

jklymak commented Mar 31, 2018

timhoffm left a comment

Choose a reason for hiding this comment

jklymak commented Apr 1, 2018

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jklymak commented Jul 8, 2018

jklymak commented Mar 29, 2018 •

edited

Loading