Set limits on a matplotlib colorbar without changing the actual plot

Question

I would like create a pseudocolor plot (e.g. contour or contourf) and a colorbar. For practical reason, I want the range of the colorbar different from the underlying mappable.

In the example below, the data Z has a range from 0 to 10000, which is mapped to a colormap. The range of the colorbar is the same.

import numpy
from matplotlib import pyplot

X = numpy.arange(100)
Y = numpy.arange(100)
Z = numpy.arange(100**2).reshape((100,100))

f = pyplot.figure()
ax = f.gca()
cf = ax.contourf(X,Y,Z,100)
cbar = f.colorbar(cf, ticks=[3000,4000,5000,6000])

pyplot.show()

Now, I would like to "zoom in" on the colorbar, i.e. generate a colorbar with a range from 3000 to 6000 only. This new colorbar shall still serve as a legend and give proper colors for each tick (3000 = blue, 6000 = yellow). Neither cbar.set_clim() nor cf.set_clim() accomplish this.

Answer 1

In general suppressing sections of your color bar is a Bad Idea, but here is a super hacky way to do it.

import numpy
from matplotlib import pyplot

X = numpy.arange(100)
Y = numpy.arange(100)
Z = numpy.arange(100**2).reshape((100,100))

f = pyplot.figure()
ax = f.gca()
cf = ax.contourf(X,Y,Z,100)
cbar = f.colorbar(cf, ticks=[3000,4000,5000,6000])
cbar.ax.set_ylim([cbar.norm(3000), cbar.norm(6000)])
cbar.outline.set_ydata([cbar.norm(3000)] * 2 + [cbar.norm(6000)] * 4 + [cbar.norm(3000)] * 3)
cbar.ax.set_aspect(60)  # <- tweak this to get the aspect ratio you want
pyplot.show()

I call this hacky because it touches a whole bunch of the internals of the colorbar. cbar.outline is a Line2D object that is the black box around the colorbar, the set_ydata sets the ydata on the corners to match the sub-region you want to look at. Try it with out that line and see what happens.

You might want to look into colormap's clip feature.