axhline affecting plot axis

Question

I'm experimenting with axhline and I'm finding unpredictable behavior. When I add an axhline sometimes it completely messes up my x-axis sometimes it does not.

Setup

import matplotlib.pyplot as plt
import pandas as pd

idx = pd.date_range('2016-01-01', '2016-03-31')
ts = pd.Series([0. for d in idx], index=idx)
t1 = ts['2016-01'] + 1
t2 = ts['2016-02'] + 2
t3 = ts['2016-03'] + 3

First test plot

Exactly as I wanted it.

Problem

Now let's add an axhline

ax = ts.plot()
ax.axhline(y=1.5)
t1.plot(ax=ax)
t2.plot(ax=ax)
t3.plot(ax=ax)
plt.ylim([-1, 4]);

Not at all what I was expecting!

However

If I add the axhline at the end.

ax = ts.plot()
t1.plot(ax=ax)
t2.plot(ax=ax)
t3.plot(ax=ax)
ax.axhline(y=1.5)
plt.ylim([-1, 4]);

No problems.

WHY!?

Why would the order in which I plot dictate the scale of the x-axis?

Versions

import matplotlib
print pd.__version__
print matplotlib.__version__

0.18.0
1.5.1

Answer 1

There is nothing wrong in the "Problem" plot. You just need to rescale x-axis so that the time scale starts in 2016. If you look at the "Problem" plot very closely then you will see there are three dots at the right end of the plot.

A quick way to fix it:

ax = ts.plot()
ax.axhline(y=1.5)
t1.plot(ax=ax)
t2.plot(ax=ax)
t3.plot(ax=ax)
plt.autoscale()
plt.ylim([-1, 4])
plt.show()

Seems like in pyplot if you create axhline first, you have to rescale before you do plt.show().