Error on colorbar

[bidhya]

Bug report

Bug summary
I get an error on calling the colorbar() function that was working perfectly with older version of matplotlib, numpy

I suspect, this TypeError is rasied when calling the plt.colorbar() method on an image of dtype boolean.

Code for reproduction

plt.imshow(bg_img) #data type of bg_img is bool
plt.colorbar();

Actual outcome

Plots the image, but gives following error with plt.colorbar()

# If applicable, paste the console output here
#
#
TypeError                                 Traceback (most recent call last)
<ipython-input-47-384f1bb7cd5d> in <module>
      1 print(bg_img.dtype)
      2 plt.imshow(bg_img)
----> 3 plt.colorbar();

~\Anaconda3\lib\site-packages\matplotlib\pyplot.py in colorbar(mappable, cax, ax, **kw)
   2095         ax = gca()
   2096 
-> 2097     ret = gcf().colorbar(mappable, cax = cax, ax=ax, **kw)
   2098     return ret
   2099 colorbar.__doc__ = matplotlib.colorbar.colorbar_doc

~\Anaconda3\lib\site-packages\matplotlib\figure.py in colorbar(self, mappable, cax, ax, use_gridspec, **kw)
   2130                              'panchor']
   2131         cb_kw = {k: v for k, v in kw.items() if k not in NON_COLORBAR_KEYS}
-> 2132         cb = cbar.colorbar_factory(cax, mappable, **cb_kw)
   2133 
   2134         self.sca(current_ax)

~\Anaconda3\lib\site-packages\matplotlib\colorbar.py in colorbar_factory(cax, mappable, **kwargs)
   1549         cb = ColorbarPatch(cax, mappable, **kwargs)
   1550     else:
-> 1551         cb = Colorbar(cax, mappable, **kwargs)
   1552 
   1553     cid = mappable.callbacksSM.connect('changed', cb.on_mappable_changed)

~\Anaconda3\lib\site-packages\matplotlib\colorbar.py in __init__(self, ax, mappable, **kw)
   1080                 kw['alpha'] = mappable.get_alpha()
   1081 
-> 1082             ColorbarBase.__init__(self, ax, **kw)
   1083 
   1084     def on_mappable_changed(self, mappable):

~\Anaconda3\lib\site-packages\matplotlib\colorbar.py in __init__(self, ax, cmap, norm, alpha, values, boundaries, orientation, ticklocation, extend, spacing, ticks, format, drawedges, filled, extendfrac, extendrect, label)
    408         # The rest is in a method so we can recalculate when clim changes.
    409         self.config_axis()
--> 410         self.draw_all()
    411 
    412     def _extend_lower(self):

~\Anaconda3\lib\site-packages\matplotlib\colorbar.py in draw_all(self)
    432         # sets self._boundaries and self._values in real data units.
    433         # takes into account extend values:
--> 434         self._process_values()
    435         # sets self.vmin and vmax in data units, but just for
    436         # the part of the colorbar that is not part of the extend

~\Anaconda3\lib\site-packages\matplotlib\colorbar.py in _process_values(self, b)
    826                 self.norm.vmin,
    827                 self.norm.vmax,
--> 828                 expander=0.1)
    829 
    830             b = self.norm.inverse(self._uniform_y(self.cmap.N + 1))

~\Anaconda3\lib\site-packages\matplotlib\transforms.py in nonsingular(vmin, vmax, expander, tiny, increasing)
   2900         vmax = expander
   2901 
-> 2902     elif vmax - vmin <= maxabsvalue * tiny:
   2903         if vmax == 0 and vmin == 0:
   2904             vmin = -expander

TypeError: numpy boolean subtract, the `-` operator, is deprecated, use the bitwise_xor, the `^` operator, or the logical_xor function instead.

Expected outcome

Matplotlib version 3.0

• Operating system: Windows 10 64-bit
• Matplotlib version:
• Matplotlib backend (print(matplotlib.get_backend())): module://ipykernel.pylab.backend_inline
• Python version: 3.6.6 Anaconda
• Jupyter version (if applicable):
• Other libraries: numpy (1.15.2)

using Anaconda

[jklymak]

Can you make a complete example: http://sscce.org

[bidhya]

Example

Using Jupyter notebook

%matplotlib inline
import matplotlib.pyplot as plt
import numpy as np

### Create a random array
a = np.random.rand(100, 100)
plt.imshow(a)
plt.colorbar();

##This second part will give error for the colorbar
### Create a boolean array
b = a > .5
print(b.dtype) # bool
plt.imshow(b) #This works
plt.colorbar() #TypeError even though it worked with older version of matplotlib/numpy

[bidhya]

sorry closed it accidentally.

[ImportanceOfBeingErnest]

Just to mention, plt.imshow(b.astype(int)) works fine. So if in doubt, using actual numbers is safer.
Still it looks like a regression.

[jklymak]

Not sure if we support boolean arrays, but colorbar hasn't worked with this since at least 2.2.0. Suggest you cast to float or int.

[anntzer]

fwiw this was already discussed in #8820.

[bidhya]

Yes, true. Upon casting to float or int, the colorbar works properly. So I believe there is no plan to support colorbar for either boolean or categorical data.

[jklymak]

Categorical would be a big change. Boolean would probably be easier, but frankly it’s probably better to defer to numpy who are not allowing Boolean math any longer.

[anntzer]

Let's close the issue here as it was already discussed less than a year ago and we decided to defer to numpy's choice.
If you want to re-discuss this decision, the mailing list (with references to both discussions here) may be a better place to catch the core-devs' attention.