Pandas matplotlib plotting, irregularities in time series labels between bar graph and line graph

Question

When creating bar graphs and line graphs using matplotlib via pandas I've come across some inconsistent behaviour. For example:

import matplotlib.pyplot as plt
import pandas as pd
from pandas_datareader import data

test_df = data.get_data_yahoo('AAPL', start='2015-10-01')
test_df['Adj Close'].plot()

Plots as expected with sensible x axis labels:

However if you then try to plot something from the same dataframe as a bar graph:

test_df['Volume'].plot(kind='bar')

The x axis tick labels are no longer automatically sensible.

Is this intended behaviour of pandas/matplotlib? And how can one easily rectify the x axis tick labels on bar graphs to be similar to the one in the line graph above?

Answer 1

You could tell matplotlib to show every Nth label:

# show every Nth label
locs, labels = plt.xticks()
N = 10
plt.xticks(locs[::N], test_df.index[::N].strftime('%Y-%m-%d'))

---

import matplotlib.pyplot as plt
import pandas as pd
from pandas_datareader import data

test_df = data.get_data_yahoo('AAPL', start='2015-10-01')
fig, ax = plt.subplots(nrows=2)
test_df['Adj Close'].plot(ax=ax[0])
test_df['Volume'].plot(kind='bar', ax=ax[1])

# show every Nth label
locs, labels = plt.xticks()
N = 10
plt.xticks(locs[::N], test_df.index[::N].strftime('%Y-%m-%d'))

# autorotate the xlabels
fig.autofmt_xdate()
plt.show()

yields

---

Another option is to use matplotlib directly:

import matplotlib.pyplot as plt
import pandas as pd
from pandas_datareader import data
import matplotlib.dates as mdates

df = data.get_data_yahoo('AAPL', start='2015-10-01')
fig, ax = plt.subplots(nrows=2, sharex=True)

ax[0].plot(df.index, df['Adj Close'])
ax[0].set_ylabel('price per share')

ax[1].bar(df.index, df['Volume']/10**6)
ax[1].xaxis.set_major_locator(mdates.MonthLocator(bymonthday=-1))
xfmt = mdates.DateFormatter('%B %d, %Y')
ax[1].xaxis.set_major_formatter(xfmt)
ax[1].set_ylabel('Volume (millions)')

# autorotate the xlabels
fig.autofmt_xdate()
plt.show()