Rotation of colorbar tick labels in matplotlib

Question

I would like to rotate the colorbar tick labels so that they read vertically rather than horizontally. I have tried as many variations as I can think of with cbar.ax.set_xticklabels and cbar.ax.ticklabel_format and so on with rotation='vertical' but haven't quite landed it yet.

I've provided a MWE below:

import numpy as np
import matplotlib.pyplot as plt

#example function
x,y = np.meshgrid(np.linspace(-10,10,200),np.linspace(-10,10,200))
z = x*y*np.exp(-(x+y)**2)

#array for contourf levels
clevs = np.linspace(z.min(),z.max(),50)

#array for colorbar tick labels
clevs1 =np.arange(-200,100,10)

cs1 = plt.contourf(x,y,z,clevs)

cbar = plt.colorbar(cs1, orientation="horizontal")
cbar.set_ticks(clevs1[::1])

plt.show()

Any pointers would be greatly appreciated - I'm sure this must be pretty simple...

Solution 1

You can use cbar.ax.set_xticklabels to change the rotation (or set_yicklabels if you had a vertical colorbar).

cbar.ax.set_xticklabels(clevs1[::1],rotation=90)

EDIT:

To set the ticks correctly, you can search for where in your clevs1 array the first tick should be using np.argmin, and use that to index clevs1 when you set_xticklabels:

tick_start = np.argmin(abs(clevs1-clevs[0]))
cbar.ax.set_xticklabels(clevs1[tick_start:],rotation=90)

Solution 2

If you're happy with tick locations and labels and only want to rotate them:

cbar.ax.set_xticklabels(cbar.ax.get_xticklabels(), rotation='vertical')

Solution 3

This is the idiomatic way to rotate tick labels as of Matplotlib 3.4 (and very likely earlier versions too)

cbar.ax.tick_params(rotation=45)

Comments

[Kieran Hunt]

I had looked at this, but it makes the tick labels incorrect.

[tmdavison]

what should they be? you have clevs1, which goes from -200 to 90, but your example goes from -100 to 0?

[tmdavison]

and should the tick labels be different to the actual contour levels? It might be better to use vmin and vmax for contourf to control the range shown on the colorbar

[Kieran Hunt]

The correct values are -100 to 0; clevs1 is (assuming we don't know the range of data plotted) a list of acceptable values to use as tick labels.

[tmdavison]

@KieranHunt OK, so its just a matter of finding the first value in clevs1 that corresponds to the lowest value of clevs. See my edit on how to do that.