I'm trying to make a time series plot, using Pandas and matplotlib, with ticked weekdays (Monday) on x axis. However, I've encountered a strange behavior which I believe is caused by (wrong?) handling of datetime indexing in pandas.Series.plot function.
From the documentation for pandas.Series.plot (also in pandas.DataFrame.plot):
use_indexbool, defaultTrue. Use index as ticks for x axis.
However, although the index is of DatetimeIndex type, it looks like it is not considered as such (probably handled as string?).
Consider the following code which tries to create the same plot using three different approaches:
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import matplotlib as mpl
idx = pd.date_range(start='2020-10-02', periods=35)
Y = np.cumsum(np.random.uniform(0, 10, idx.size))
ts = pd.Series(Y, idx)
fig, axes = plt.subplots(figsize=(10, 6), nrows=3, constrained_layout=True)
ts.plot(kind='bar', ax=axes[0]) # use_index=True is used by defualt
axes[0].xaxis.set_major_locator(mpl.dates.WeekdayLocator(interval=1, byweekday=mpl.dates.MONDAY))
ts.plot(kind='bar', ax=axes[1])
axes[1].xaxis.set_major_locator(mpl.dates.WeekdayLocator(interval=1, byweekday=mpl.dates.MONDAY))
axes[1].xaxis.set_major_formatter(mpl.dates.DateFormatter('%Y-%m-%d'))
axes[2].bar(ts.index, ts.values)
axes[2].xaxis.set_major_locator(mpl.dates.WeekdayLocator(interval=1, byweekday=mpl.dates.MONDAY))
axes[2].xaxis.set_major_formatter(mpl.dates.DateFormatter('%Y-%m-%d'))
The results is
I expected all three plots would be more-or-less the same. And they are, excluding the ticks of x axis (which should mark Monday's dates). As i see it, the DatetimeIndex is not really recognized correctly in pandas.Series.plot. The dates on the first plot are correct, but the ticked dates are not Mondays. What is even more annoying, if a new format is applied on the ticks (second plot) even the dates are wrong!
The behavior of bar(ts.index, ts.values) is correct and marks Mondays as expected.
