We don't see that warning in the tests. That probably means that there is no tests with aspect as a number. It would be good to add such a test

Please reword your commit messages so that they indicate what you're actually doing instead of a generic "Update filename".

I can not get this warning locally

In [47]: import matplotlib.pyplot as plt

In [48]: plt.axes().set_aspect(1)

In [49]: 

It looks like this is a python2 only issue and just a warning.

I am 👎 on merging this.

how about a warning filter instead?

On Fri, Apr 8, 2016 at 12:10 AM, Thomas A Caswell notifications@github.com
wrote:

I am [image: 👎] on merging this.

—
You are receiving this because you commented.
Reply to this email directly or view it on GitHub
#6279 (comment)

Sorry. plt.axes().set_aspect(1) doesn't raise the unicodewarnings, but plt.axes().set_aspect(np.float64(1)) and plt.axes().set_aspect(scipy.sqrt(1)) do.

It seems to be a side effect of from __future__ import unicode_literals, observed as follows:

$ python -c "from __future__ import unicode_literals;import numpy;print numpy.float64(1) == 'normal'"
-c:1: UnicodeWarning: Unicode equal comparison failed to convert both arguments to Unicode - interpreting them as being unequal
False
$ python
Python 2.7.11 (default, Dec  5 2015, 22:44:43) 
[GCC 4.2.1 Compatible Apple LLVM 6.0 (clang-600.0.57)] on darwin
Type "help", "copyright", "credits" or "license" for more information.

It will happen when ever you try to compare a numpy type to a unicode string. I.e. import numpy;print numpy.float64(1) == u'normal' also produces the warning.
One possbile way around it could be to verify that the aspect is a string before doing the comparison i.e. replace:
if aspect == 'normal':
with:
if isinstance(aspect, six.string_types) and aspect == 'normal':

it's not the cleanest and goes breaks normal python ducktypeing but gets the python2 behaviour closer to the python3 one.

Closed by #7658.