What else apart from useOffset is controlling tick label offsets?

[michaelaye]

I was happy to have a new rcParams parameter implemented (#2400) to control the use of the ticker format parameter useOffset.
But I have since discovered another behavior of MPL that seems to ignore this control parameter, possibly to protect axis labels from overwriting each other. Could somebody point out to me what this behavior is called and how to avoid this?
The only difference below is that the second plot uses x-data factor 10 larger than in the first plot.

%matplotlib inline
fig, ax = plt.subplots(2)
ax = ax.flatten()
x = np.linspace(1e6, 1e6+100, 100)
y = np.linspace(5e5, 5e5+100, 100)
ax[0].ticklabel_format(useOffset=False)
ax[0].plot(x,y)
ax[1].ticklabel_format(useOffset=False)
ax[1].plot(x*10,y)

I'm posting this here and not in stackoverflow, because it could be a bug of an artist ignoring the useOffset=False instruction, as this looks pretty much like an offset to me.

[efiring]

What you are seeing is not an offset, it is a multiplier. To turn it off, use the style kwarg:

plt.ticklabel_format(useOffset=False, style='plain')