how to plot time on y-axis in '%H:%M' format in matplotlib?

Question

i would like to plot the times from a datetime64 series, where the y-axis is formatted as '%H:%M, showing only 00:00, 01:00, 02:00, etc.

this is what the plot looks like without customizing the y-axis formatting.

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from matplotlib.dates import DateFormatter
from matplotlib.dates import HourLocator

df = pd.DataFrame(data=dict(a=pd.date_range('1/1/2011',periods=1440000,freq='1min')))
df = df.iloc[np.arange(0,1440*100,1440)+np.random.randint(1,300,100)]

plt.plot(df.index,df['a'].dt.time)
plt.show()

After reading around on the topic on SO, I attempted the following but without success.

ax = plt.subplot()
ax.yaxis.set_major_locator(HourLocator())
ax.yaxis.set_major_formatter(DateFormatter('%H:%M'))
plt.plot(df.index,df['a'].dt.time)
plt.show()

ValueError: DateFormatter found a value of x=0, which is an illegal date.  This usually occurs because you have not informed the axis that it is plotting dates, e.g., with ax.xaxis_date()

Could anyone advise me?

Solution

For that to work you need to pass datetime objects (and I mean datetime, not datetime64). You can convert all timestamps to the same date and then use .tolist() to get the actual datetime objects.

y = df['a'].apply(lambda x: x.replace(year=1967, month=6, day=25)).tolist()
ax = plt.subplot()
ax.plot(df.index, y)
ax.yaxis.set_major_locator(HourLocator())
ax.yaxis.set_major_formatter(DateFormatter('%H:%M'))

Comments

[themachinist]

thanks but these lines dont appear to work. also, why should it be ax.axis when time is on the y-axis?

[tacaswell]

Importing pandas implicitly installs converters to deal with datetime64 correctly (by laundering them through the pandas dt64 boxed datatype).

[themachinist]

thanks @tcaswell, but pandas (0.17.0) was imported in my example above. any suggestions on why it didnt work?

[Stop harming Monica]

@tcaswell Not sure what you mean, I had to add the .tolist() call to make HourLocator work. I noticed that when plotting .df.time matplotlib used pandas stuff for ticks and labels, but I did not find an easy way to customize them to met the requirements.