在python中，如何正确使用`colorbar`和`pcolormesh`？

Question

Here is my code,

from mpl_toolkits.axes_grid1 import make_axes_locatable # colorbar
from matplotlib import pyplot as plt
from matplotlib import cm # 3D surface color
import numpy as np

data1 = np.random.rand(10, 12)
data2 = np.random.rand(10, 12)
data3 = data1 - data2

vmin = min([data1.min(), data2.min(), data3.min()])
vmax = max([data1.max(), data2.max(), data2.max()])
fig, (ax_1, ax_2, ax_error) = plt.subplots(nrows=3, ncols=1, figsize=(6, 6))

ax_1.set_ylabel('x')
mesh_1 = ax_1.pcolormesh(data1.T, cmap = cm.coolwarm)

ax_2.set_ylabel('x')
mesh_2 = ax_2.pcolormesh(data2.T, cmap = cm.coolwarm)

mesh_error = ax_error.pcolormesh(data3.T, cmap = cm.coolwarm)
ax_error.set_ylabel('x')
ax_error.set_xlabel('t')

divider = make_axes_locatable(ax_2)
cax_val = divider.append_axes("right", size="2%", pad=.1)

fig.colorbar(mesh_2, ax=[ax_1, ax_2, ax_error], cax=cax_val)
fig.tight_layout()

plt.show()

and it produces an image

However, what I expect is that it produces the picture below

Can anyone help me with this problem? Thanks in advance for any helpful suggestion!

Answer 1

tight_layout doesn't help with this problem, unfortunately. No tight_layout and no axes_grid works fine:

from matplotlib import pyplot as plt
from matplotlib import cm # 3D surface color
import numpy as np

data1 = np.random.rand(10, 12)
data2 = np.random.rand(10, 12)
data3 = data1 - data2

fig, (ax_1, ax_2, ax_error) = plt.subplots(nrows=3, ncols=1, figsize=(6, 6))

mesh_1 = ax_1.pcolormesh(data1.T, cmap = cm.coolwarm)
mesh_2 = ax_2.pcolormesh(data2.T, cmap = cm.coolwarm)
mesh_error = ax_error.pcolormesh(data3.T, cmap = cm.coolwarm)

fig.colorbar(mesh_2, ax=[ax_1, ax_2, ax_error])
plt.show()

If you want better spacing you can try constrained_layout :

fig, (ax_1, ax_2, ax_error) = plt.subplots(nrows=3, ncols=1, figsize=(6, 6), 
                                           constrained_layout=True)

Constrained layout will also work for just one axes:

fig.colorbar(mesh_2, ax=ax_2)

Answer 2

With the help from @JodyKlymak, I finally solved the problem. The keypoint lies in using shrink , ie fig.colorbar(mesh_2, ax=[ax_1, ax_2, ax_error], shrink=0.3) . Here is the solution

from matplotlib import pyplot as plt
from matplotlib import cm # 3D surface color
import numpy as np

data1 = np.random.rand(10, 12)
data2 = np.random.rand(10, 12)
data3 = data1 - data2

fig, (ax_1, ax_2, ax_error) = plt.subplots(nrows=3, ncols=1, figsize=(6, 6))

mesh_1 = ax_1.pcolormesh(data1.T, cmap = cm.coolwarm)
mesh_2 = ax_2.pcolormesh(data2.T, cmap = cm.coolwarm)
mesh_error = ax_error.pcolormesh(data3.T, cmap = cm.coolwarm)

fig.colorbar(mesh_2, ax=[ax_1, ax_2, ax_error], shrink=0.3)
plt.show()

and it produces