Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Discontinuous timeseries plot with dates on x-axis

I got data for several months, but in between some months are missing. This looks quite strange if I plot the whole dataset in one plot (lots of empty space in between). I wrote the small example script to show how it works (based on: Python/Matplotlib - Is there a way to make a discontinuous axis?)

The problem: I can't get the x-axis use the same date formatting! Either ax or ax2 is correct, but never both of them. Do you have any idea?

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

def getDates(startdate, enddate):
    days  = (enddate + datetime.timedelta(days=1) - startdate).days
    dates = [ startdate + datetime.timedelta(days=x) for x in range(0,days) ]
    return dates

dates1 = getDates(datetime.datetime(2013,1,1), datetime.datetime(2013,1,31))
dates2 = getDates(datetime.datetime(2013,3,1), datetime.datetime(2013,3,31))
dates = dates1+dates2
data = np.arange(len(dates))

Locator = mpl.dates.DayLocator(interval=5)
Formatter = mpl.dates.DateFormatter('%d-%m-%y')

fig,(ax,ax2) = plt.subplots(1,2,sharey=True)
fig.subplots_adjust(wspace=0.05)
fig.set_size_inches(10,3)
ax.plot(dates, data)
ax2.plot(dates, data)
ax.legend(loc=1)
ax.set_ylim( 0, 61 )
ax.set_xlim( datetime.datetime(2013,1,1), datetime.datetime(2013,1,31) )
ax2.set_xlim( datetime.datetime(2013,3,1), datetime.datetime(2013,3,31) )
labels = ax.get_xticklabels()
for label in labels: label.set_rotation(30)
labels = ax2.get_xticklabels()
for label in labels: label.set_rotation(30) 
ax.spines['right'].set_visible(False)
ax2.spines['left'].set_visible(False)
ax.tick_params(right='off')
ax2.tick_params(left='off')
ax2.yaxis.tick_right()
ax.xaxis.set_major_locator(Locator)
ax.xaxis.set_major_formatter(Formatter)
ax2.xaxis.set_major_locator(Locator)
ax2.xaxis.set_major_formatter(Formatter)
plt.savefig("test.png", bbox_inches='tight')

Result: Result

like image 748
HyperCube Avatar asked Sep 24 '13 13:09

HyperCube


1 Answers

You have found an interesting detail about the internals of matplotlib. The locator object you pass into set_major_locator is the object used by the axes to figure out where to put it's ticks both axes were using the same locater object. As part of the draw the locator generates a list of where the ticks should be based on the limits of the axes which when it gets done for the second axes means no ticks are visible in the first axes. You just need to pass in distinct (separate instantiations) locator objects, done here with copy.

import datetime
import copy

def getDates(startdate, enddate):
    days  = (enddate + datetime.timedelta(days=1) - startdate).days
    dates = [ startdate + datetime.timedelta(days=x) for x in range(0, days) ]
    return dates

dates1 = getDates(datetime.datetime(2013, 1, 1), datetime.datetime(2013, 1, 31))
dates2 = getDates(datetime.datetime(2013, 3, 1), datetime.datetime(2013, 3, 31))
dates = dates1+dates2
data = np.arange(len(dates))

Locator = mpl.dates.DayLocator(interval=5)
Formatter = mpl.dates.DateFormatter('%d-%m-%y')

fig, (ax, ax2) = plt.subplots(1, 2, sharey=True, tight_layout=True)
fig.subplots_adjust(wspace=0.05)
fig.set_size_inches(10, 3, forward=True)

ax.plot(dates, data)
ax2.plot(dates, data)

ax.legend(loc=1)
ax.set_ylim(0, 61)
ax.set_xlim(datetime.datetime(2013, 1, 1), datetime.datetime(2013, 1, 31))
ax2.set_xlim(datetime.datetime(2013, 3, 1), datetime.datetime(2013, 3, 31))

labels = ax.get_xticklabels()
for label in labels:
    label.set_rotation(30)
labels = ax2.get_xticklabels()
for label in labels:
    label.set_rotation(30)

ax.spines['right'].set_visible(False)
ax2.spines['left'].set_visible(False)
ax.tick_params(right='off')
ax2.tick_params(left='off')
ax2.yaxis.tick_right()


# note the copy here
ax.xaxis.set_major_locator(copy.copy(Locator))
ax.xaxis.set_major_formatter(copy.copy(Formatter))
ax2.xaxis.set_major_locator(copy.copy(Locator))
ax2.xaxis.set_major_formatter(copy.copy(Formatter))

enter image description here

like image 97
tacaswell Avatar answered Nov 03 '22 14:11

tacaswell