Python discrete colormap use

[Nkehoe-QUB]

Hello,

I am trying to use the discrete version of the colourmaps but it just throws errors.

fig,ax = plt.subplots()
cax = ax.pcolormesh(x, y, data.T, norm=cm.LogNorm(), cmap=cmaps.batlow20)
cbar = fig.colorbar(cax)
plt.show()

My understanding from the documentation is that this should split Batlow into 20 discrete colour bins but I just get the following error:

AttributeError: module 'cmcrameri.cm' has no attribute 'batlow20'

Any assistance would be appreciated.

[callumrollo]

Hi @Nkehoe-QUB

You should be able to use matplotlib.color.BoundaryNorm to normalize the colormaps to whatever number of discrete intervals you want. For 20 bins, the example would be like this:

import cmcrameri.cm as cmc
import matplotlib.pyplot as plt
import numpy as np
import matplotlib 

x = np.linspace(0, 1, 100)[np.newaxis, :]
levels = np.linspace(0,1,20) # change 20 to whatever number of levels you want
cmap =cmc.batlow
norm = matplotlib.colors.BoundaryNorm(levels, cmap.N)
plt.imshow(x, aspect='auto', cmap=cmap, norm = norm)
plt.axis('off')
plt.show()

Does this work for your use case?

[burggraaff]

cmap=cmaps.batlow.resampled(20) also works in your application

[callumrollo]

I didn't even know you could do that resampling! Way more elegant. Thanks @burggraaff I'm putting that in the README

[Zerosimi]

I really like the resampling approach. Makes it really simple to have the discrete colormap. I noticed however, that the resampling did not behave as expected for the categorical colormap. I was hoping for the first 5 colors of the categorical colormap, as they are picked to have high contrast, but it selects colors from the whole color range, which in this case does look worse than the non-categorical colormap. See below

import numpy as np
import matplotlib.pyplot as plt
import cmcrameri as cmc

# Data generation
np.random.seed(42)
x = np.random.uniform(-1, 1, 250)
y = np.random.uniform(-1, 1, 250)
categorical_data = np.random.randint(0, 5, size=250)

fig = plt.figure()

im = plt.scatter(x, y, c=categorical_data, cmap=cmc.cm.batlowS.resampled(5))
plt.colorbar(im)
plt.title("batlowS.resampled(5)")

plt.show()
plt.close()

[burggraaff]

@Zerosimi The .resampled method comes from matplotlib and its qualitative/categorical colour maps have the same behaviour, as you can see below where I compare viridis (sequential, continuous) and Dark2 (qualitative, discrete):

Correct me if I'm wrong, but I think it's just a consequence of the way these colour maps are implemented in matplotlib: not as a list of N colours, but as a continuous range from 0..1 with N discrete values assigned to equally spaced sections. If you want to change the behaviour of .resampled, you'll have to do it in matplotlib itself.

As a workaround for selecting the first N colours, you can define a new cmap based on that range. Here's an example where I wanted to resample devon without the near-white bits at the far end:

from cmcrameri import cm as cmc
from matplotlib.colors import LinearSegmentedColormap

N_colours = 12
devon_discrete = LinearSegmentedColormap.from_list("devon_discrete", cmc.devon.colors[:-20]).resampled(N_colours)

Here's a comparison of the result with different limits (the last colour in the first bar might be hard to make out compared to the white background):