Matplotlib set_color_cycle versus set_prop_cycle

Question

One of my favorite things to do in Matplotlib is to set the color-cycle to match some colormap, in order to produce line-plots that have a nice progression of colors across the lines. Like this one:

Previously, this was one line of code using set_color_cycle:

ax.set_color_cycle([plt.cm.spectral(i) for i in np.linspace(0, 1, num_lines)])

But, recently I see a warning:

MatplotlibDeprecationWarning: 
The set_color_cycle attribute was deprecated in version 1.5. 
Use set_prop_cycle instead.

Using set_prop_cycle, I can achieve the same result, but I need to import cycler, and the syntax is less compact:

from cycler import cycler
colors = [plt.cm.spectral(i) for i in np.linspace(0, 1, num_lines)]
ax.set_prop_cycle(cycler('color', colors))

So, my questions are:

Am I using set_prop_cycle correctly? (and in the most efficient way?)

Is there an easier way to set the color-cycle to a colormap? In other words, is there some mythical function like this?

ax.set_colorcycle_to_colormap('jet', nlines=30)

Here is the code for the complete example:

import numpy as np
import matplotlib.pyplot as plt

ax = plt.subplot(111)
num_lines = 30

colors = [plt.cm.spectral(i) for i in np.linspace(0, 1, num_lines)]

# old way: 
ax.set_color_cycle(colors)

# new way:
from cycler import cycler
ax.set_prop_cycle(cycler('color', colors))

for n in range(num_lines):
    x = np.linspace(0,10,500)
    y = np.sin(x)+n
    ax.plot(x, y, lw=3)

plt.show()

Solution

Because the new property cycler can iterate over other properties than just color (e.g. linestyle) you need to specify the label, i.e. the property over which to cycle.

ax.set_prop_cycle('color', colors)

There is no need to import and create a cycler though; so as I see it the only drawback of the new method it that it makes the call 8 characters longer.

There is no magical method that takes a colormap as input and creates the cycler, but you can also shorten your color list creation by directly supplying the numpy array to the colormap.

colors = plt.cm.Spectral(np.linspace(0,1,30))

Or in combination

ax.set_prop_cycle('color',plt.cm.Spectral(np.linspace(0,1,30)))

Comments

[jds]

plt.cm.get_cmap('spectral') for anyone who gets an error trying to access the spectral function.

[Miguel Gonzalez]

'spectral'.capitalize()

[DanHickstein]

Ah, so now I can pretty easily specify a whole combination of different colors, linewidths, linestyles, etc. That does seem like an advantage. And the only price to pay is 8 more characters - I can live with that :). And thanks for the tip about passing the linspace to the colormap. In the end, I think that you saved me a few characters!

[jhaagsma]

I get this error, sadly: AttributeError: module 'matplotlib.cm' has no attribute 'spectral' Has matplotlib changed yet again?

[ImportanceOfBeingErnest]

@jhaagsma The spectral colormap has been removed. You may use any other valid colormap instead, e.g. Spectral or nipy_spectral. I updated the answer.