What type of data was failing? Does the test cover the failure mode?

This seems reasonable to me, but it'd be good to make sure we know what failed and that we have a test to gaurd for future..

What type of data was failing? Does the test cover the failure mode?

If you set vmin or vmax in an imshow call to be a quantity with units that caused failures later. This wasn't an issue before 2.2.0.

This seems reasonable to me, but it'd be good to make sure we know what failed and that we have a test to gaurd for future..

The changes I made to test_colors.py should test this failure mode.

If I run the following on master and v2.2.0, it comes back True, so I don't think this test is triggering the bug, or this problem is fixed on master.

import numpy as np
import matplotlib
import matplotlib.colors as mcolors

class MyArray(np.ndarray):
    def __isub__(self, other):
        raise RuntimeError

    def __add__(self, other):
        raise RuntimeError

data = np.arange(-10, 10, 1, dtype=float)
mydata = data.view(MyArray)
for norm in [mcolors.Normalize(vmin=mydata.min(), vmax=mydata.max()),
            mcolors.SymLogNorm(3, vmin=mydata.min(), vmax=mydata.max())]:
    print(np.all(norm(mydata) == norm(data)))

@jklymak thanks! It looks like you actually have to use the Normalize instance with an imshow plot to trigger the bug. Let me see if I can rewrite the test in a way that actually triggers it. I appreciate your patience with my test writing :)

Ha, given that I just got a slap on the wrist for my test writing, please no apologies!

We did change some image normalization stuff recently, hence I wanted to make sure that we didn't miss something. i.e. #10613 worse, that one didn't get in for the RC, but was in response to a bug report after the RC, so its possible we (ahem, I) fixed one thing and broke another.

@jklymak updated the test. Here's a script that I verified should fail on 2.2.0:

import matplotlib
matplotlib.use('Agg')
import numpy as np
import matplotlib.colors as mcolors
from matplotlib import pyplot as plt


class MyArray(np.ndarray):
    def __isub__(self, other):
        raise RuntimeError

    def __add__(self, other):
        raise RuntimeError


data = np.arange(-10, 10, 1, dtype=float)
data.shape = (10, 2)
mydata = data.view(MyArray)
for norm in [mcolors.Normalize(vmin=mydata.min(), vmax=mydata.max()),
             mcolors.SymLogNorm(3, vmin=mydata.min(), vmax=mydata.max())]:
    fig, ax = plt.subplots()
    ax.imshow(mydata, norm=norm)
    fig.canvas.draw()

Yep, that was #10613. Sorry about that. Your fix looks reasonable, as does some test along the lines of the above.