The default format to display time series xtick labels in time series plots seems to have been changed between matplotlib 1.4 and 1.5. When using time stamps in a datetime format with a 1 second resolution and plotting a minute worth of data, the major xtick labels are now formatted as "HH:MM:SS.mmmmm" as opposed to "HH:MM:SS" in matplotlib 1.4. This new behaviour doesn't match the significany of the timestamps and results in overlapping xtick labels as well.

A simple example to illustrate the issue:

import numpy as np
import matplotlib.pyplot as plt
import datetime

# generating time stamps with a resolution of 1s
x = np.array([datetime.datetime(2016, 5, 4, 12, 0, i) for i in range(60)])
# generating some random y-values
y = np.random.randn(x.shape[0]).cumsum()

# plot
plt.plot(x, y)
plt.show()

This results in the following figure:

Of course, the xtick label format can be manually adjusted:

import matplotlib.dates as dates

# create plot
plt.plot(x, y)
# get reference to x-axis
xax = plt.gca().get_xaxis()
# format major xtick label
xax.set_major_formatter(dates.DateFormatter('%H:%M:%S'))
# show plot
plt.show()

But I found the old behavior more elegant.

I'm using matplotlib 1.5.1 on Python 3.5.1 32-bit as part of the Anaconda3 4.0.0 Windows x86 distrubtion. The OS is Windows 7 64 bit.