Hide "inner" {x,y}labels in label_outer too.

[anntzer]

ax.label_outer used to hide only "inner" ticklabels; make it hide axes
labels too. (I couldn't find a way to just switch label visibility, so
I just set the labels to the empty string instead.)

[afvincent]

@anntzer I didn't find an example of ax.label_outer in the example gallery so wouldn't your PR be the occasion to add one? For example something like:

"""
When plotting a grid of subplots that share axes, it might be interesting
to display only some of the tick labels as well as some of the axes labels.

The `label_outer` method provides a convenient way of dealing with this
by setting visible only the x tick and axis labels of the bottom row and
the y tick and axis labels of the left column.
"""

import matplotlib.pyplot as plt
import numpy as np

# Dummy data to plot
t = np.linspace(0.0, 5.0, 501)
s1 = 1.25 * np.sin(2*np.pi*t)
s2 = np.exp(-t)
s3 = np.sin(4*np.pi*t)
s4 = np.sin(2*np.pi*t) * np.exp(-t)

# Instantiate a 2x2 subplot grid with both shared x and y axes.
fig, axs = plt.subplots(ncols=2, nrows=2, sharex=True, sharey=True)

# Plot a data signal in each subplot and display:
# - y tick and axis labels only for the first column;
# - x tick and axis labels only for the last row.
for ax, s, title in ((axs[0, 0], s1, 'Left labels ON'),
                     (axs[0, 1], s2, 'All labels OFF'),
                     (axs[1, 0], s3, 'Left & bottom labels ON'),
                     (axs[1, 1], s4, 'Bottom labels ON')):
    ax.plot(t, s)
    ax.set_xlabel('This is x-axis')
    ax.set_ylabel('This is y-axis')
    ax.label_outer()  # Display only the "relevant" tick & axis labels.
    ax.set_title(title)

plt.show()

if one wants to add a simple stand-alone example. And if one doesn't like such a small stand-alone example, then the example subplots_demo may be a good place to add the same kind of example.

For the moment (without the current PR), the provided snippet results in

but this should improve with what you proposed.

[anntzer]

I updated subplots_demo instead.

[afvincent]

Nice! In fact, I hadn't directly proposed to modify the existing example in the case someone thought it would be useful to keep an example of the “cosmetricks” like

plt.setp([a.get_xticklabels() for a in axarr[0, :]], visible=False)
plt.setp([a.get_yticklabels() for a in axarr[:, 1]], visible=False)

which might give more fine-tuning possibilities if one doesn't simply want to label only the “outer” subplots. However if one had to choose, as label_outer will now also handle “inner” axis labels (as well as it also takes care of the possible offset texts), which the former “cosmetrick” doesn't, I really think label_outer should be demonstrated over the former technic.

[anntzer]

Actually the previous approach is "broken" in the sense that it will not handle offset texts (see #4445, #4446), so I'd rather just not have it in the examples here.

[afvincent]

If I wasn't clear in my previous comment: I agree with you about label_outer being a better solution than the former approach because it takes care of the offset texts (as well as of the x- and y-labels with this PR).