60

I'd like to make a generic value -vs- time plot with python's matplotlib module. My times are in unix time but I'd like them to show up in a readable format on the plot's x-axis.

I have read answers about plotting with datetime objects but this method seems to remove hour/min/sec information and rails timestamps to the full day. Is there a way to generate these plots and show more granular labels?

Matt
  • 4,815
  • 5
  • 39
  • 40
  • Matplotlib certainly _doesn't_ remove the time information when plotting datetimes... What exactly are you doing? Do you just want to change the labeling of the axis? Are you just wanting one tick/label per plotted data point? – Joe Kington Nov 03 '10 at 20:01

1 Answers1

96

It is possible to call plt.plot(dates,values) with dates being a list of datetime.datetime objects. The plot will include xticks in a format like '%Y-%m-%d' and as you zoom in, automatically change to one that shows hours, minutes, seconds.

However, it sounds like you desire more control than this. Perhaps it is not showing the hours, minutes, seconds at the scale you wish.

In that case, you can set up your own date formatter:

ax=plt.gca()
xfmt = md.DateFormatter('%Y-%m-%d %H:%M:%S')
ax.xaxis.set_major_formatter(xfmt)

Unfortunately, if you pass datetime.datetime objects to plt.plot, the xticks automatically chosen by matplotlib seems to always have seconds equal to zero. For example, if you run

import matplotlib.pyplot as plt
import matplotlib.dates as md
import numpy as np
import datetime as dt
import time

n=20
duration=1000
now=time.mktime(time.localtime())
timestamps=np.linspace(now,now+duration,n)
dates=[dt.datetime.fromtimestamp(ts) for ts in timestamps]
values=np.sin((timestamps-now)/duration*2*np.pi)
plt.subplots_adjust(bottom=0.2)
plt.xticks( rotation=25 )
ax=plt.gca()
xfmt = md.DateFormatter('%Y-%m-%d %H:%M:%S')
ax.xaxis.set_major_formatter(xfmt)
plt.plot(dates,values)
plt.show()

alt text

then you get nicely formatted dates, but all the xtick seconds are zero.

So what's the solution?

If you convert your timestamps --> datetime.datetime objects --> matplotlib datenums yourself, and pass the datenums to plt.plot, then the seconds are preserved.

PS. By "matplotlib datenum" I mean the kind of number returned by matplotlib.dates.date2num.

import matplotlib.pyplot as plt
import matplotlib.dates as md
import numpy as np
import datetime as dt
import time

n=20
duration=1000
now=time.mktime(time.localtime())
timestamps=np.linspace(now,now+duration,n)
dates=[dt.datetime.fromtimestamp(ts) for ts in timestamps]
datenums=md.date2num(dates)
values=np.sin((timestamps-now)/duration*2*np.pi)
plt.subplots_adjust(bottom=0.2)
plt.xticks( rotation=25 )
ax=plt.gca()
xfmt = md.DateFormatter('%Y-%m-%d %H:%M:%S')
ax.xaxis.set_major_formatter(xfmt)
plt.plot(datenums,values)
plt.show()

alt text

unutbu
  • 842,883
  • 184
  • 1,785
  • 1,677
  • 4
    Nice example! For what it's worth it's not that the seconds are being lost by the conversion by `date2num`. It's that the tick locator is automatically choosing even second locations for the tick labels. – Joe Kington Nov 03 '10 at 20:23
  • @Joe: Yes, you are absolutely right. I started to realize that as I fumbled around editing this thing.... :) – unutbu Nov 03 '10 at 20:26
  • 1
    Do the dates need to be in string format or in numeric format? – eleijonmarck Mar 27 '15 at 15:37