Two issues with the axes offset indicator

[anntzer]

Consider:

plt.plot(10001 + np.arange(10), 10001 + np.arange(10))
plt.setp(plt.gca().get_yticklabels(), visible=False)
plt.show()

The x axis is labeled 0..9 + 1.0001e4. I'd suggest changing this to 1..10 + 1.0000e4, which is IMO much more legible.

The y axis doesn't show the tick labels, as expected, but the offset indicator ("+1.0001e4") is still there. setp(get_{x,y}ticklabels(), visible=False) is certainly a commonly used approach (e.g. by SubplotBase.label_outer()).
What I believe the correct solution, which is also set offsetText (which can be obtained by get_offset_text()) to invisible, is a bit obscure (and not mentioned in the docs).

[tacaswell]

If you reach in and change the visibility of the individual Text objects, there is not much that we can do. I would argue that this behavior is actually a feature so users can flip individual tick labels on/off without affecting anything else.

The easiest way to turn off the tick labels on an axis (if you are not using sharing any axis) is

import matplotlib.ticker as mticker
ax.yaxis.set_major_formatter(mticker.NullFormatter ())

which removes both the labels and the offset label.

In the case of label_outer things are a bit trickier as there may be shared axes around so setting the NullFormatter will remove all of the ticklabels.

Can you add few lines needed in label_outer to also hide the offset text and a test for this?

Also: http://matplotlib.org/api/axis_api.html#matplotlib.axis.Axis.get_offset_text

[anntzer]

Sure, I understand and don't expect the API of get_xticklabels to be changed; my point about lack of mention in the docs was that I was hoping for some "FAQ: How do I get rid of all ticklabels? Answer: Set ticklabels and offsetText to not visible" (or better, as you mention it, "use NullFormatter").
If you go through the examples, many more are using the "set ticklabels to not visible" approach than the "use NullFormatter" approach, which also leads to the former being more often mentioned on StackOverflow & others.
I can have a look at label_outer.