Make divider without changing size of original axis?

Question

Matplotlib's make_axes_locatable tool allows you to tack a new axis onto the side of an existing axis. However, it resizes the parent axis. Is there a way to avoid this?

Below is a complete example showing the problem and how to reproduce it:

import matplotlib.pyplot as pl
from mpl_toolkits.axes_grid import make_axes_locatable
import matplotlib.axes as maxes


fig = pl.figure()
ax1=pl.subplot(1,3,1)
ax1.imshow([[0,1],[2,0]])
ax1.yaxis.set_visible(False)
ax1.xaxis.set_visible(False)
ax2=pl.subplot(1,3,2)
ax2.imshow([[0,1],[2,0]])
ax2.yaxis.set_visible(False)
ax2.xaxis.set_visible(False)
ax3=pl.subplot(1,3,3)
ax3.imshow([[0,1],[2,0]])
ax3.yaxis.set_visible(False)
ax3.xaxis.set_visible(False)
pl.subplots_adjust(wspace=0)


divider = make_axes_locatable(ax1)
cax1 = divider.new_horizontal(size=0.2, pad=0.0, pack_start=True, axes_class=maxes.Axes)
pl.colorbar(ax1.images[0],cax=cax1)
cax1.yaxis.set_label_position('left')
cax1.yaxis.set_ticks_position('left')
fig.add_axes(cax1)

divider = make_axes_locatable(ax2)
cax2 = divider.new_vertical(size=0.2, pad=0.0, pack_start=True, axes_class=maxes.Axes)
fig.add_axes(cax2)
pl.colorbar(ax2.images[0],cax=cax2,orientation='horizontal')
# thin out the tick labels for visibility
for t in cax2.xaxis.get_majorticklabels()[::2]:
    t.set_visible(False)


divider = make_axes_locatable(ax3)
cax3 = divider.new_horizontal(size=0.2, pad=0.0, pack_start=False, axes_class=maxes.Axes)
pl.colorbar(ax3.images[0],cax=cax3)
fig.add_axes(cax3)

The problem is that the subplots are now different sizes. I think the left and right have shrunk, but the middle is unchanged.

Answer 1

I've been able to avoid the resizing of the parent plot by modifying your code to create new axes for each colorbar & then placing each one manually. It's a bit more work, but I think it's close to the result you are looking for. Note that the actual aesthetic appearance of the plots is a bit different from yours--perhaps because I'm using a more recent version of matplotlib (1.2.1).

%pylab inline
import matplotlib.pyplot as pl

fig = pl.figure()
ax1=pl.subplot(1,3,1)
ax1.imshow([[0,1],[2,0]])
ax1.yaxis.set_visible(False)
ax1.xaxis.set_visible(False)
ax2=pl.subplot(1,3,2)
ax2.imshow([[0,1],[2,0]])
ax2.yaxis.set_visible(False)
ax2.xaxis.set_visible(False)
ax3=pl.subplot(1,3,3)
ax3.imshow([[0,1],[2,0]])
ax3.yaxis.set_visible(False)
ax3.xaxis.set_visible(False)
pl.subplots_adjust(wspace=0)

#Give the colorbar its own axis to avoid resizing the parent axis:
width = 0.02
height = 0.38
vertical_position = 0.32
horizontal_position = 0.1
axColor = pl.axes([horizontal_position, vertical_position, width, height]) #the new axis for first colorbar
pl.colorbar(ax1.images[0],cax=axColor,orientation='vertical')
axColor.yaxis.set_label_position('left')
axColor.yaxis.set_ticks_position('left')

#likewise for the other colorbars with appropriately adjusted positions/ orientations:
horizontal_position= 0.38
vertical_position = 0.29
height = 0.03
width = 0.26
axColor2 = pl.axes([horizontal_position, vertical_position, width, height]) #the new axis for second colorbar
pl.colorbar(ax2.images[0],cax=axColor2,orientation='horizontal')
# thin out the tick labels for visibility
for t in axColor2.xaxis.get_majorticklabels()[::2]:
    t.set_visible(False)

width = 0.02
height = 0.38
vertical_position = 0.32
horizontal_position = 0.905
axColor3 = pl.axes([horizontal_position, vertical_position, width, height]) #the new axis for third colorbar    
pl.colorbar(ax3.images[0],cax=axColor3,orientation='vertical')