copy a color map object does not isolate changes to cm

[tacaswell]

If you do cm2 = copy.copy(cm) and the original color map has been used at least once, then the copy will share a reference to self._lut (which is what is used to __call__ to map from values -> rgba). Any calls to set_over, set_under, or set_bad will be shared by all copies (as they update self._lut in place).

[tacaswell]

Exact work:

• implement __copy__ and maybe __deepcopy__ on ColorMap

Tagging as medium because the color map code is a bit convoluted .

[bendichter]

I can reproduce this issue with:

import matplotlib.pyplot as plt
import copy
import numpy as np

%matplotlib inline
data = np.ma.masked_array([[1,2,3],[2,3,4]], mask=[[1,0,0],[0,0,0]])
cm = plt.cm.Reds
plt.imshow(data,cmap=cm)
plt.show()
cm2 = copy.copy(cm)
cm2.set_bad('b')
plt.imshow(data,cmap=cm)

In the second plot the image contains blue when it should remain white. copy.deepcopy works as expected, though. Would the solution be to have copy call deepcopy?

[dopplershift]

Implementing __copy__ should be enough to get the desired behavior.