Correcting matplotlib colorbar ticks

urschrei picture urschrei · Sep 9, 2013 · Viewed 38.5k times · Source

I've placed a color bar alongside a choropleth map. Because the data being plotted are discrete rather than continuous values, I've used a LinearSegmentedColormap (using the recipe from the scipy cookbook), which I've initialised with my max counted value + 1, in order to show a colour for 0. However, I now have two problems:

enter image description here

  1. The tick labels are incorrectly spaced (except for 5, more or less) – they should be located in the middle of the colour they identify; i.e. 0 - 4 should be moved up, and 6 - 10 should be moved down.

  2. If I initialise the colorbar with drawedges=True, so that I can style its dividers properties, I get this:

enter image description here

I'm creating my colormap and colorbar like so:

cbmin, cbmax = min(counts), max(counts)
# this normalises the counts to a 0,1 interval
counts /= np.max(np.abs(counts), axis=0)
# density is a discrete number, so we have to use a discrete color ramp/bar
cm = cmap_discretize(plt.get_cmap('YlGnBu'), int(cbmax) + 1)
mappable = plt.cm.ScalarMappable(cmap=cm)
mappable.set_array(counts)
# set min and max values for the colour bar ticks
mappable.set_clim(cbmin, cbmax)
pc = PatchCollection(patches, match_original=True)
# impose our colour map onto the patch collection
pc.set_facecolor(cm(counts))
ax.add_collection(pc,)
cb = plt.colorbar(mappable, drawedges=True)

So I'm wondering whether my converting the counts to a 0,1 interval is one of the problems.

Update :

Having tried what Hooked suggested, the 0-value is correct, but subsequent values are set progressively higher, to the the point where 9 is where 10 should be:

enter image description here

Here's the code I used:

cb = plt.colorbar(mappable)
labels = np.arange(0, int(cbmax) + 1, 1)
loc = labels + .5
cb.set_ticks(loc)
cb.set_ticklabels(labels)

And just to confirm, labels definitely has the correct values:

In [3]: np.arange(0, int(cbmax) + 1, 1)
Out[3]: array([ 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10])

Answer

unutbu picture unutbu · Sep 9, 2013

You are suffering from an off-by-one error. You have 10 ticklabels spread among 11 colors. You might be able to correct the error by using np.linspace instead of np.arange. Using np.linspace the third argument is the number of values desired. This reduces the amount of mental gymnastics needed to avoid the off-by-one error:

import matplotlib.pyplot as plt
import numpy as np
import matplotlib.cm as cm
import matplotlib.colors as mcolors

def colorbar_index(ncolors, cmap):
    cmap = cmap_discretize(cmap, ncolors)
    mappable = cm.ScalarMappable(cmap=cmap)
    mappable.set_array([])
    mappable.set_clim(-0.5, ncolors+0.5)
    colorbar = plt.colorbar(mappable)
    colorbar.set_ticks(np.linspace(0, ncolors, ncolors))
    colorbar.set_ticklabels(range(ncolors))

def cmap_discretize(cmap, N):
    """Return a discrete colormap from the continuous colormap cmap.

        cmap: colormap instance, eg. cm.jet. 
        N: number of colors.

    Example
        x = resize(arange(100), (5,100))
        djet = cmap_discretize(cm.jet, 5)
        imshow(x, cmap=djet)
    """

    if type(cmap) == str:
        cmap = plt.get_cmap(cmap)
    colors_i = np.concatenate((np.linspace(0, 1., N), (0.,0.,0.,0.)))
    colors_rgba = cmap(colors_i)
    indices = np.linspace(0, 1., N+1)
    cdict = {}
    for ki,key in enumerate(('red','green','blue')):
        cdict[key] = [ (indices[i], colors_rgba[i-1,ki], colors_rgba[i,ki])
                       for i in xrange(N+1) ]
    # Return colormap object.
    return mcolors.LinearSegmentedColormap(cmap.name + "_%d"%N, cdict, 1024)

fig, ax = plt.subplots()
A = np.random.random((10,10))*10
cmap = plt.get_cmap('YlGnBu')
ax.imshow(A, interpolation='nearest', cmap=cmap)
colorbar_index(ncolors=11, cmap=cmap)    
plt.show()

enter image description here