Question

FINAL EDIT (Partial Solution):

After a lot of tinkering I’ve been able to accomplish what I want.
It isn’t exactly a fast bit of code as it’s plotting over 17,000 separate bar plots.

Solution

X = np.array(xvals)
Y = np.array(yvals)
Z = np.array(zvals)
Y_bot = np.array(bottomvals)

fig = plt.figure()
ax = fig.add_subplot(111)

jet = plt.get_cmap('jet')

log_norm = colors.LogNorm()

for x,y,z,bot in zip(X,Y,Z,Y_bot):
    log_z = np.log10(z)/np.log10(max(Z))
    colour = jet(log_price)
    ax.bar(x,y,bottom=bot, width=1, edgecolor='none',color=c_col)

a = np.vstack((Z,Z))
im = plt.imshow(a, aspect='auto', cmap=jet, origin='lower', norm=log_norm)
im.set_visible(False)
plt.colorbar(format='%.1f')
plt.axis([values])
plt.show()

This is able to successfully construct the plot I wanted, it automatically does the colour bar and formats it appropriately as well.

—- Begin Original Question —-

I am attempting to incorporate a logarithmic colour map to a series of stacked bar charts in matplotlib (python 2.7). Essentially, each “stack” value has a secondary value which I’m applying a logarithm transform to and then using that as part of a colour map.

There are two loops which run to create the plot:

import matplotlib.pyplot as plt
import matplotlib.pyplot as plt
import numpy as np
import matplotlib.cm as cm
import matplotlib.colors as colors
import matplotlib.colorbar as cb
import matplotlib.ticker as tk    

for i in range(1,x):
    # get data for point i
    for n, j in enumerate(data):
        cval = log10(secondary value)
        plt.bar(i,j, bottom=sum(j[:n]), color=cm.jet(cval)

# Once it has all been plotted I want to set axis, titles and set up a color bar
# To set up the color bar I'm currently using

im = plt.imshow(a, aspect='auto',cmap=jet, origin='lower')
im.set_visible(False)
cbar = plt.colorbar()

Now, the issue is that these values are the log changed values (which is correct as that is what I’ve been using to plot the data).

But, I want to perform the reverse transform (e.g. 10 ** x) and have those valued displayed on the colorbar.

So far I’ve tried using:

cbar.ax.get_yticklabels() # This gets text values, not floats
cbar.ax.get_yticks() # This gets the positions

I know I can update the ticklabels using:

cbar.ax.set_yticklabels(new_ticks)

However, I want a generic way of obtaining the tick values in a way to make the transform.

The current format of the text values returned is:

Text(1,0,u'\u22121.6') # Corresponds to -1.6 in unicode (I think).

Any assistance would be greatly appreciated, this has been doing my head in.

In the end it should look something like this (Note, I’ve manually adjusted these tick values, axis labels have been removed as they’re not important)

EDIT:

I’ve also created a hexbin plot of the data, as you can see it is very very lumpy and generally messy to work with.

To generalise the above problem.

(notation, z =element, Z = vector array etc)

I have a data array Z with two unique indices (x, y).

I would like to plot each value of z at a location specified by its two indices (x, y). z should be coloured logarithmically. I would additionally like a colorbar to be plotted with this plot. Ideally, the spaces should be filled in, similar to the first stacked bar chart.

Note, X is an array of integers (although type is float) ranging from 1.0 to 48.0.
However, Y is an unconstrained float and no order links each value to X. Thus I’m unsure if contour plots are applicable here.

Answer 1

After a lot of tinkering I’ve been able to accomplish what I want.
It isn’t exactly a fast bit of code as it’s plotting over 17,000 separate bar plots.

Solution

X = np.array(xvals)
Y = np.array(yvals)
Z = np.array(zvals)
Y_bot = np.array(bottomvals)

fig = plt.figure()
ax = fig.add_subplot(111)

jet = plt.get_cmap('jet')

log_norm = colors.LogNorm()

for x,y,z,bot in zip(X,Y,Z,Y_bot):
    log_z = np.log10(z)/np.log10(max(Z))
    colour = jet(log_price)
    ax.bar(x,y,bottom=bot, width=1, edgecolor='none',color=c_col)

a = np.vstack((Z,Z))
im = plt.imshow(a, aspect='auto', cmap=jet, origin='lower', norm=log_norm)
im.set_visible(False)
plt.colorbar(format='%.1f')
plt.axis([values])
plt.show()

This is able to successfully construct the plot I wanted, it automatically does the colour bar and formats it appropriately as well.