How to set the view_limits / range of an axis in Matplotlib - python-3.x

Question

I want some space before and after the last tick location on an axis. I would just use axis.set_xlim() for example but this interferes with my (custom) locator and reruns the tick generation. I found and overwritten the view_limits() method of the locator-classes but they don't seem to be called automatically and when called manually they don't have any impact on the resulting plot. I searched the docs and the source but haven't come up with a solution. Am I missing something?
For the greater picture I want to have a locator which gives me some space before and after the ticks and chooses tick points which are a multiple of 'base' like the MultipleLocator but scale the base automatically if the number of ticks exceeds a specified value. If there is another way to achieve this without subclassing a locator I am all ears :).
Here is my example code for the subclassed locator with overwritten view_limits-method:
import matplotlib.pyplot as plt
import numpy as np
from matplotlib.ticker import MaxNLocator
class MyLocator(MaxNLocator):
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
def view_limits(self, dmin, dmax):
bins = self.bin_boundaries(dmin, dmax)
step = bins[1] - bins[0]
result = np.array([bins[0] - step, bins[-1] + step])
print(result)
return result
a = 10.0
b = 99.0
t = np.arange(a, b, 0.1)
s = np.sin(0.1*np.pi*t)*np.exp(-t*0.01)
loc = MyLocator(9)
fig, ax = plt.subplots()
plt.plot(t, s)
ax.xaxis.set_major_locator(loc)
loc.autoscale() # results in [ 0. 110.] but doesnt change the plot
plt.show()

Answer 1

Not sure, if I understood completely what your problem is, but if you only want to add extra space, you can still use MaxNLocator and add that space manually like here:
import matplotlib.pyplot as plt
import numpy as np
from matplotlib.ticker import MaxNLocator
a = 10.0
b = 99.0
t = np.arange(a, b, 0.1)
s = np.sin(0.1*np.pi*t)*np.exp(-t*0.01)
loc = MaxNLocator(9)
fig, ax = plt.subplots()
plt.plot(t, s)
ax.xaxis.set_major_locator(loc)
ticks = ax.get_xticks()
newticks = np.zeros(len(ticks)+2)
newticks[0] = ticks[0]- (ticks[1]-ticks[0])
newticks[-1] = ticks[-1]+ (ticks[1]-ticks[0])
newticks[1:-1] = ticks
ax.set_xticks(newticks)
plt.show()

Answer 2

One slightly hacky solution to avoid ticks close to the plot's edges is the following:
class PaddedMaxNLocator(mp.ticker.MaxNLocator):
def __init__(self, *args, protected_width=0.25, **kwargs):
# `prune` edge ticks that might now become visible
super().__init__(*args, **kwargs, prune='both')
# Clamp to some reasonable range
self.protected_width = min(0.5, protected_width)
def tick_values(self, vmin, vmax):
diff = (vmax - vmin) * self.protected_width / 2
return super().tick_values(vmin + diff, vmax - diff)