Improve bar_label annotation #22447

Valdes-Tresanco-MS · 2022-02-10T22:50:55Z

PR Summary

Improve bar_label method
Fix #22414

PR Checklist

Tests and Styling

Has pytest style unit tests (and pytest passes).
Is Flake 8 compliant (install flake8-docstrings and run flake8 --docstring-convention=all).

Documentation

New features are documented, with examples if plot related.
New features have an entry in doc/users/next_whats_new/ (follow instructions in README.rst there).
API changes documented in doc/api/next_api_changes/ (follow instructions in README.rst there).
Documentation is sphinx and numpydoc compliant (the docs should build without error).

github-actions

Thank you for opening your first PR into Matplotlib!

If you have not heard from us in a while, please feel free to ping @matplotlib/developers or anyone who has commented on the PR. Most of our reviewers are volunteers and sometimes things fall through the cracks.

You can also join us on gitter for real-time discussion.

For details on testing, writing docs, and our review process, please see the developer guide

We strive to be a welcoming and open project. Please follow our Code of Conduct.

tacaswell · 2022-02-10T23:05:20Z

lib/matplotlib/axes/_axes.py

@@ -2608,6 +2608,11 @@ def bar_label(self, container, labels=None, *, fmt="%g", label_type="edge",

        # want to know whether to put label on positive or negative direction
        # cannot use np.sign here because it will return 0 if x == 0
+        a, b = self.yaxis.get_view_interval()
+        yinverted = -1 if a > b else 1


Why not use True / False here? This looks like it is boolean state (and is unlikely to ever grow a third state).

We have two usages below: a boolean and a sign. This tries to push both notions into one variable, which makes it a bit odd: The variable naming suggests a boolean, but the value carries the sign.

I suggest

Suggested change

yinverted = -1 if a > b else 1

yinverted = a > b

and create the sign below; see there.

tacaswell · 2022-02-10T23:40:11Z

Thank you for both reporting and fixing this @Valdes-Tresanco-MS !

I left one small comment in the code about the implementation. Could you also add a test so that we are sure that it does not come back? I do not think we need an image test. I suspect there is a way to query the location of the text, convert it to data coordinates and assert that it is greater than the data it is above.

Valdes-Tresanco-MS · 2022-02-10T23:52:49Z

I described the solution in #22414 and in this comment #22414 (comment). In this comment, I include also an image with the test result.

This is enough?
In the issue, @mwaskom also suggests a deeper revision, but unfortunately, I don't have time for it

Test code

f, axes = plt.subplots(3, 4, figsize=(16, 12))
(ax1, ax2, ax3, ax4), (ax5, ax6, ax7, ax8), (ax9, ax10, ax11, ax12) = axes
bars1 = ax1.bar(["a", "b", "c", "d"], [1, 2, 3, -3])
ax1.bar_label(bars1, padding=0, label_type='center')
ax1.set_title('Vertical Normal yaxis padding=0')

bars2 = ax2.bar(["a", "b", "c", "d"], [1, 2, 3, -3])
ax2.invert_yaxis()
ax2.bar_label(bars2, padding=0, label_type='center')
ax2.set_title('Vertical Inverted yaxis padding=0')

bars5 = ax5.barh(["a", "b", "c", "d"], [1, 2, 3, -3])
ax5.bar_label(bars5, padding=0, label_type='center')
ax5.set_title('Horizontal Normal yaxis padding=0')

bars6 = ax6.barh(["a", "b", "c", "d"], [1, 2, 3, -3])
ax6.invert_yaxis()
ax6.bar_label(bars6, padding=0, label_type='center')
ax6.set_title('Horizontal Inverted yaxis padding=0')

bars9 = ax9.barh(["a", "b", "c", "d"], [1, 2, 3, -3])
ax9.invert_xaxis()
ax9.bar_label(bars9, padding=0, label_type='center')
ax9.set_title('Horizontal Inverted xaxis padding=0')

bars10 = ax10.barh(["a", "b", "c", "d"], [1, 2, 3, -3])
ax10.invert_yaxis()
ax10.invert_xaxis()
ax10.bar_label(bars10, padding=0, label_type='center')
ax10.set_title('Horizontal Inverted yaxis and xaxis padding=0')

bars3 = ax3.bar(["a", "b", "c", "d"], [1, 2, 3, -3])
ax3.bar_label(bars3, padding=10, label_type='center')
ax3.set_title('Vertical Normal yaxis padding=10')

bars4 = ax4.bar(["a", "b", "c", "d"], [1, 2, 3, -3])
ax4.invert_yaxis()
ax4.bar_label(bars4, padding=10, label_type='center')
ax4.set_title('Vertical Inverted yaxis padding=10')

bars7 = ax7.barh(["a", "b", "c", "d"], [1, 2, 3, -3])
ax7.bar_label(bars7, padding=10, label_type='center')
ax7.set_title('Horizontal Normal yaxis padding=10')

bars8 = ax8.barh(["a", "b", "c", "d"], [1, 2, 3, -3])
ax8.invert_yaxis()
ax8.bar_label(bars8, padding=10, label_type='center')
ax8.set_title('Horizontal Inverted yaxis padding=10')

bars11 = ax11.barh(["a", "b", "c", "d"], [1, 2, 3, -3])
ax11.invert_xaxis()
ax11.bar_label(bars11, padding=10, label_type='center')
ax11.set_title('Horizontal Inverted xaxis padding=10')

bars12 = ax12.barh(["a", "b", "c", "d"], [1, 2, 3, -3])
ax12.invert_yaxis()
ax12.invert_xaxis()
ax12.bar_label(bars12, padding=10, label_type='center')
ax12.set_title('Horizontal Inverted yaxis and xaxis padding=10')

f.tight_layout()

mwaskom · 2022-02-11T00:18:32Z

Worth pointing out that this is now inconsistent with how annotations work:

ax = plt.gca()
ax.invert_yaxis()
ax.bar([0, 1], [1, .5])
ax.annotate("hello", (1, .5), xytext=(0, 10), textcoords="offset points", ha="center", va="center")

I'm not sure if there is other precedence for point offset adjustments in the library to cross-reference.

QuLogic · 2022-02-11T00:35:26Z

Please give your PRs and commits more descriptive titles; no-one will know what 22414 is in two weeks.

Valdes-Tresanco-MS · 2022-02-11T00:56:51Z

@mwaskom Sorry if I missed something, but why wouldn't it be inconsistent? bar_label makes an annotation passing modified arguments

matplotlib/lib/matplotlib/axes/_axes.py

Lines 2685 to 2688 in 541a786

    
           annotation = self.annotate(fmt % value if lbl is None else lbl, 
        
                                      xy, xytext, textcoords="offset points", 
        
                                      ha=ha, va=va, **kwargs) 
        
           annotations.append(annotation)

If I add this line to your example, everything seems to work fine.

ax = plt.gca()
ax.invert_yaxis()
b = ax.bar([0, 1], [1, .5])
ax.annotate("hello", (1, .5), xytext=(0, 20), textcoords="offset points", ha="center", va="center")
ax.bar_label(b)

Suddenly something related to the handling of the attributes of the annotations escapes me, so I would not like to be left with the doubt.

Additionally, could this be a solution to what you suggest?

matplotlib/lib/matplotlib/axes/_axes.py

Lines 2680 to 2688 in 541a786

    
                   va = 'center' 
        
           if np.isnan(dat): 
        
               lbl = '' 
        
           annotation = self.annotate(fmt % value if lbl is None else lbl, 
        
                                      xy, xytext, textcoords="offset points", 
        
                                      ha=ha, va=va, **kwargs) 
        
           annotations.append(annotation)

if 'ha' not in kwargs:
    kwargs.update([('ha', ha)])
if 'va' not in kwargs:
    kwargs.update([('va', va)])

annotation = self.annotate(fmt % value if lbl is None else lbl,
                           xy, xytext, textcoords="offset points",
                           **kwargs)

Although it is not elegant in principle, it does what you suggest. If the alignment parameters are not explicitly defined, then those determined in the method are passed, otherwise, the user's parameters are used. I only tested a few cases, so I don't know if it could have a wrong behavior in some cases

mwaskom · 2022-02-11T01:26:53Z

Sorry if I missed something, but why wouldn't it be inconsistent? bar_label makes an annotation passing modified arguments

Right, the existing behavior is happening because of the way that annotate works. I'm just pointing out that "fixing" it will introduce an inconsistency. I am not sure that I have an opinion about which behavior is preferable, just noting.

could this be a solution to what you suggest?

Yes but it can be even simpler:

kwargs.setdefault("ha", ha)
kwargs.setdefault("va", va)

tacaswell · 2022-02-11T01:26:58Z

@Valdes-Tresanco-MS Posts on GH are certainly useful! However, the test code needs to end up in the test suite that is run on PR so the computer can tell us if things are broken.

Valdes-Tresanco-MS · 2022-02-11T03:32:41Z

kwargs.setdefault("ha", ha)
kwargs.setdefault("va", va)

@mwaskom I found a small problem with the implementation of the ha and va parameters defined by the user. Since the annotation, in this case, is done on each bar, assigning the items ha and va envelopes overwrites the configuration, making all labels appear the same, whether they are for positive or negative values. The solution is to create a copy of kwargs on each iteration to define these particular variables in each iteration. However, the fact that the labels are defined in the same way for all the values I think can be confusing for users.

Valdes-Tresanco-MS · 2022-02-11T03:59:19Z

@tacaswell I think it's ready.

Valdes-Tresanco-MS · 2022-02-11T04:28:05Z

@tacaswell I see you added it to the 3.6.0 milestone. Since it is an improvement of a method that was already implemented, is it possible to add it in the current version (3.5.x)? It would be a great help for my projects.
Thanks

timhoffm · 2022-02-12T22:48:14Z

lib/matplotlib/axes/_axes.py

@@ -2608,6 +2608,11 @@ def bar_label(self, container, labels=None, *, fmt="%g", label_type="edge",

        # want to know whether to put label on positive or negative direction
        # cannot use np.sign here because it will return 0 if x == 0
+        a, b = self.yaxis.get_view_interval()
+        yinverted = -1 if a > b else 1


We have two usages below: a boolean and a sign. This tries to push both notions into one variable, which makes it a bit odd: The variable naming suggests a boolean, but the value carries the sign.

I suggest

Suggested change

yinverted = -1 if a > b else 1

yinverted = a > b

and create the sign below; see there.

lib/matplotlib/axes/_axes.py

Valdes-Tresanco-MS · 2022-02-12T23:01:58Z

@timhoffm I see, definitely better

timhoffm · 2022-02-12T23:08:40Z

@Valdes-Tresanco-MS since you wrote that you don't have much time, I took the liberty and pushed my proposed changes as well as error handling if the user passes alignment parameters to your branch. I hope that's ok.

lib/matplotlib/axes/_axes.py

QuLogic · 2022-03-02T01:33:18Z

lib/matplotlib/axes/_axes.py

+                    if y_inverted:
+                        va = 'top' if dat > 0 else 'bottom'  # also handles NaN
+                    else:
+                        va = 'top' if dat < 0 else 'bottom'  # also handles NaN


I'm not entirely certain this is correct, because there are no tests with NaN on an inverted axis.

@Valdes-Tresanco-MS di this get addressed? If so, please move back out of draft. Thanks!

There was a new test added, but the problem is that the result that is being checked is the same as what occurs when you run the code on main. So I'm not certain if the test is asserting that this PR changes things correctly (and the test just happens to match main coincidentally), or the test is asserting that this PR doesn't break how things currently work (and the match is intentional).

So what I guess I'm saying is it would be better if the test were paramterized across all the 4 cases here.

Valdes-Tresanco-MS · 2022-03-30T14:45:15Z

@jklymak Sorry, I don't understand what I should do. Since it was indicated I included the test (a64db70). The failure in the check seems to be related to a CLI failure not to the test itself. This commit should I not include it?

jklymak · 2022-03-30T15:02:21Z

I just wanted to be sure @QuLogic 's comment had been addressed....

Valdes-Tresanco-MS · 2022-03-30T15:07:29Z

I just wanted to be sure @QuLogic 's comment had been addressed....

Yes. Should I mark the conversations as resolved? It's my first PR here

jklymak · 2022-03-30T15:11:38Z

Making as resolved depends how trivial the comment is. If its a typo I go ahead and mark as resolved. If the commenter might want to think about it more I leave it open. Thanks!

Valdes-Tresanco-MS · 2022-03-30T15:17:19Z

Making as resolved depends how trivial the comment is. If its a typo I go ahead and mark as resolved. If the commenter might want to think about it more I leave it open. Thanks!

Thank you for the clarification. I remain attentive to your comments to help with what I can in this PR.

timhoffm · 2022-03-30T21:41:39Z

lib/matplotlib/axes/_axes.py

+        for key in ['horizontalalignment', 'ha', 'verticalalignment', 'va']:
+            if key in kwargs:


Might alternatively written as

Suggested change

for key in ['horizontalalignment', 'ha', 'verticalalignment', 'va']:

if key in kwargs:

unsupported = set(['horizontalalignment', 'ha', 'verticalalignment', 'va']):

for key in unsupported.intersection(kwargs):

But no strong preference. I'll leave this up to you.

That should be:

unsupported = {'horizontalalignment', 'ha', 'verticalalignment', 'va'}

Merged as is since the existing code style is good enough.

timhoffm · 2022-03-30T21:45:21Z

Python 3.8 on ubuntu-18.04 test failure is unrelated.

timhoffm · 2022-04-05T23:23:38Z

@Valdes-Tresanco-MS thanks, and congratulations on your first contribution to Matplotlib! We hope to see you back.

Valdes-Tresanco-MS · 2022-04-05T23:27:00Z

@Valdes-Tresanco-MS thanks, and congratulations on your first contribution to Matplotlib! We hope to see you back.

Definitely... I will try to contribute what I can in the future.

Fix matplotlib#22414

09fa965

github-actions bot reviewed Feb 10, 2022

View reviewed changes

tacaswell added this to the v3.6.0 milestone Feb 10, 2022

tacaswell reviewed Feb 10, 2022

View reviewed changes

Valdes-Tresanco-MS changed the title ~~Fix #22414~~ Improve bar_label annotation Feb 11, 2022

added tests for inverted axes

2b77283

QuLogic added the status: needs workflow approval For PRs from new contributors, from which GitHub blocks workflows by default. label Feb 11, 2022

timhoffm reviewed Feb 12, 2022

View reviewed changes

Cleanup logic

9bc4c41

Document that bar_label() does not support alignment parameters

4560691

oscargus approved these changes Feb 28, 2022

View reviewed changes

QuLogic reviewed Mar 2, 2022

View reviewed changes

Valdes-Tresanco-MS added 2 commits March 1, 2022 21:19

refactoring

fae56a0

added test for nan values with yaxis inverted

a64db70

jklymak marked this pull request as draft March 30, 2022 11:50

jklymak added the status: needs revision label Mar 30, 2022

jklymak marked this pull request as ready for review March 30, 2022 15:10

jklymak removed the status: needs revision label Mar 30, 2022

jklymak requested a review from QuLogic March 30, 2022 15:10

timhoffm approved these changes Mar 30, 2022

View reviewed changes

timhoffm merged commit 33b3015 into matplotlib:main Apr 5, 2022

timhoffm removed the status: needs workflow approval For PRs from new contributors, from which GitHub blocks workflows by default. label Apr 5, 2022

		for key in ['horizontalalignment', 'ha', 'verticalalignment', 'va']:
		if key in kwargs:

Uh oh!

Improve bar_label annotation #22447

Improve bar_label annotation #22447

Uh oh!

Conversation

Valdes-Tresanco-MS commented Feb 10, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

PR Summary

PR Checklist

Uh oh!

github-actions bot left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

tacaswell commented Feb 10, 2022

Uh oh!

Valdes-Tresanco-MS commented Feb 10, 2022

Uh oh!

mwaskom commented Feb 11, 2022

Uh oh!

QuLogic commented Feb 11, 2022

Uh oh!

Valdes-Tresanco-MS commented Feb 11, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mwaskom commented Feb 11, 2022

Uh oh!

tacaswell commented Feb 11, 2022

Uh oh!

Valdes-Tresanco-MS commented Feb 11, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Valdes-Tresanco-MS commented Feb 11, 2022

Uh oh!

Valdes-Tresanco-MS commented Feb 11, 2022

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Valdes-Tresanco-MS commented Feb 12, 2022

Uh oh!

timhoffm commented Feb 12, 2022

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Valdes-Tresanco-MS commented Mar 30, 2022

Uh oh!

jklymak commented Mar 30, 2022

Uh oh!

Valdes-Tresanco-MS commented Mar 30, 2022

Uh oh!

jklymak commented Mar 30, 2022

Uh oh!

Valdes-Tresanco-MS commented Mar 30, 2022

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

timhoffm commented Mar 30, 2022

Uh oh!

timhoffm commented Apr 5, 2022

Uh oh!

Valdes-Tresanco-MS commented Apr 5, 2022

Uh oh!

Uh oh!

Valdes-Tresanco-MS commented Feb 10, 2022 •

edited

Loading

Valdes-Tresanco-MS commented Feb 11, 2022 •

edited

Loading

Valdes-Tresanco-MS commented Feb 11, 2022 •

edited

Loading