Remove font warning when legend is added while using Tex

[Struan-Murray]

PR Summary

Addresses issue #17518 regarding a warning being shown when legend is added while using Tex. Code by @anntzer.

PR Checklist

• Has Pytest style unit tests
• Code is Flake 8 compliant
• [N/A] New features are documented, with examples if plot related
• [N/A] Documentation is sphinx and numpydoc compliant
• [N/A] Added an entry to doc/users/next_whats_new/ if major new feature (follow instructions in README.rst there)
• [N/A] Documented in doc/api/api_changes.rst if API changed in a backward-incompatible way

[QuLogic]

Is there a way we can test this?

[Struan-Murray]

I can confirm that after building from source, with diff

git diff master e17518 
diff --git a/lib/matplotlib/offsetbox.py b/lib/matplotlib/offsetbox.py
index 7bcf3f3de..60ddf06ef 100644
--- a/lib/matplotlib/offsetbox.py
+++ b/lib/matplotlib/offsetbox.py
@@ -876,7 +876,8 @@ class TextArea(OffsetBox):
 
     def get_extent(self, renderer):
         _, h_, d_ = renderer.get_text_width_height_descent(
-            "lp", self._text._fontproperties, ismath=False)
+            "lp", self._text._fontproperties,
+            ismath="TeX" if self._text.get_usetex() else False)
 
         bbox, info, d = self._text._get_layout(renderer)
         w, h = bbox.width, bbox.height

and confirming that I have moved on from matplotlib 3.2.1:

Python 3.8.2 (default, Apr 27 2020, 15:53:34) 
[GCC 9.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import matplotlib
>>> matplotlib.__version__
'3.2.1.post2648+g80935b1f9'

That the MWE:

from matplotlib import rc
import matplotlib.pyplot as plt

rc('font', **{'family':'serif','serif':'Computer Modern'})
rc('text', usetex=True)

plt.figure()
plt.plot(0,0,label='Input')
plt.title('Commenting out legend removes warning')
plt.text(0, 0, 'Text test')
plt.xlabel('Label Test')
plt.xticks(ticks=[-1,1],labels=['TestA','$MATH_{\pi}$'])

plt.legend()

plt.show()

now does not give an error while using plt.legend().

[QuLogic]

I mean an automated test.

[Struan-Murray]

I don't know how to do that, where should I start?

[jklymak]

Hi @Struan-Murray There are instructions in the developer guide. But really, just do this if you are interested a) not sure that a test is easy for this and b) its a bit of a process if you are just trying to get a bug squashed...

[Struan-Murray]

Hey @jklymak, I'm reading up on pytest just now in the developer guide, I assume this is what you meant unless it's the online tests further down, let me know. I just started using matplotlib a few days ago (it's working great so far, thanks!) so I might be a bit out of my depth. I'm understanding that I would need to try find a way to assert that the code does not throw a warning when the second argument (ismath) is set to "TeX" as opposed to "False" where it does, I don't know how to do this within a test environment, yet. From my side the bug is squashed, I'll probably continue reading up on testing but might just observe any further changes as it seems a little too complicated for what was just a minor annoyance. Thanks for the help.

[jklymak]

Well the nice thing about warnings is that they lead to a failed test, so you could just add your routine above (boil it down as simple as you can) to the end of lib/matplotlib/tests/test_legend.py; no need to assert anything. The only gotchya is that I think you need to put a bad font in?

[Struan-Murray]

Smaller MWE:

from matplotlib import rc
import matplotlib.pyplot as plt

rc('font', **{'family':'serif','serif':'Computer Modern'})
rc('text', usetex=True)

plt.figure()
plt.plot(0,0,label='Input')
plt.legend()
plt.show()

So as I understand it, this test should give a warning to show that it will still alert the user that an incorrect font has been selected? I see what you mean, I'll have to find a way to select a font that is guaranteed to not be installed to show that it still throws a warning code, maybe I'll just put some random characters in. Either way it's late here so I'll work on it later.

[jklymak]

ping @Struan-Murray thanks for working on this - any progress on an automated test?

[Struan-Murray]

Hi @jklymak. I tried to create a test for this at the time but unfortunately have not had the time to get it to work, especially for something which was such a small issue. Had I worked on anything like this before I would likely be able to do it but I am interested to see how this is handled (if anyone is willing) so I am capable of creating tests in the future.

[QuLogic]

I rebased and added a test for you. Note, this is a little harder than @jklymak guessed, as it's not a warning, but a warning-level log message.

Since Qulogic wrote the test.

[dstansby]

Thanks for the PR!