Python Matplotlib wont show rolling averages over existing subplot

Question

I've created subplots which contain 2 bars sitting side-by-side. I now want to add 7 day rolling averages as lines over the top of the bars.

I can't get the plot to work, whichever I define last seems to occupy the figure. I'd like all 4 of these plots to be on the same figure. How can I achieve this?

date_index = pd.date_range(df.Created.min(), df.Created.max(), freq='D')

fig = plt.figure()
ax2 = fig.add_subplot(111)
ax22 = ax2.twinx()

s1 = CDdf.resample('D', on='Created').size().fillna(0).reindex(date_index, fill_value=0)
s2 = CDdf.groupby('Created')['Machine Count'].first().fillna(0).reindex(date_index, fill_value=0)
s = (s1/s2).fillna(0).reindex(date_index, fill_value=0)

s1.plot(kind='bar', ax=ax2, position=0, label='Sales Total', width=0.25)
s.plot(kind='bar', ax=ax22, color='red', position=1, label='Adjusted For Machine Count', width=0.25)

s1rolling = s1.rolling(window=7,center=False).mean().fillna(0).reindex(date_index, fill_value=0)
plt.plot(s1rolling, color='blue', label='_nolegend_')

srolling = s.rolling(window=7,center=False).mean().fillna(0).reindex(date_index, fill_value=0)
plt.plot(srolling, color='red', label='_nolegend_')

ax2.set_ylim(0,s1.max()*1.1)
ax22.set_ylim(0,s1.max()*1.1)
plt.legend(loc='upper left')
plt.ylabel('Frequency')
plt.title('Items Deposited Per Day')

ticklabels = s.index.strftime('%Y-%m-%d')
ax2.xaxis.set_major_formatter(ticker.FixedFormatter(ticklabels))

plt.show()

Data sample:

    Created Quoted Price    Machine Count
6   2017-10-06  0.99    3
454 2017-10-21  0.43    11
534 2017-10-21  0.98    11
487 2017-10-21  0.05    11
530 2017-10-21  0.05    11
482 2017-10-21  0.06    11
503 2017-10-21  0.05    11
416 2017-10-21  0.24    11
532 2017-10-21  0.07    11
469 2017-10-21  0.52    11
459 2017-10-21  0.05    11
515 2017-10-21  1.82    11
411 2017-10-21  0.32    11
539 2017-10-21  0.05    11
508 2017-10-21  0.23    11
1057    2017-10-28  0.07    11
1037    2017-10-28  0.06    11
1042    2017-10-28  0.17    11
1048    2017-10-28  0.34    11
1028    2017-10-28  0.09    11
1053    2017-10-28  0.50    11
1055    2017-10-28  1.33    11
1149    2017-10-29  0.25    11
1142    2017-10-29  0.12    11
1160    2017-10-29  0.05    11

Answer 1

Usually the datetimes utilities of pandas and matplotlib are incompatible. If you use a matplotlib.dates object on a date axis created with pandas then this will in most cases fail.

Here is a solution where pandas is used for plotting and matplotlib for formatting (see comments):

import matplotlib.pyplot as plt  # version 2.1.0
import matplotlib.ticker as ticker
import pandas as pd  # version 0.21.0

df = pd.read_csv('data.csv', delim_whitespace=True, index_col=0, parse_dates=['Created'])
date_index = pd.date_range(df.Created.min(), df.Created.max(), freq='D')

_, ax = plt.subplots()

s1 = df.resample('D', on='Created').size().fillna(0).reindex(date_index, fill_value=0)
s2 = df.groupby('Created')['MachineCount'].first().fillna(0).reindex(date_index, fill_value=0)
s = (s1 / s2).fillna(0).reindex(date_index, fill_value=0)

s1rolling = s1.rolling(window=7, center=False).mean().fillna(0).reindex(date_index, fill_value=0)
s2rolling = s2.rolling(window=7, center=False).mean().fillna(0).reindex(date_index, fill_value=0)
srolling = s.rolling(window=7, center=False).mean().fillna(0).reindex(date_index, fill_value=0)

# Plot with pandas without date axis (i.e. use_index=False).
s1.plot(kind='bar', color='C0', position=0, label='Sales Total', width=0.25, use_index=False)
s.plot(kind='bar', color='C1', position=1, label='Adjusted For Machine Count', width=0.25, use_index=False)

# Plot with pandas without date axis (i.e. use_index=False).
s1rolling.plot(kind='line', color='C0', label='_nolegend_', use_index=False)
srolling.plot(kind='line', color='C1',  label='_nolegend_', use_index=False)

plt.ylim(0, s1.max() * 1.1)
plt.legend(loc='upper left')
plt.ylabel('Frequency')
plt.title('Items Deposited Per Day')

# Format date axis with matplotlib.
ticklabels = s1.index.strftime('%Y-%m-%d')
ax.xaxis.set_major_formatter(ticker.FixedFormatter(ticklabels))
plt.xticks(rotation=90)

plt.tight_layout()
plt.show()

I hope this will help you.