matplotlib plot (not scatter) color based on extra variable

Question

Good day,

I am trying to plot two arrays (timesin and locations) as a scatter of points. However, because timesin is a datetime object (of which I want time only), I find that I can only plot it properly using pyplot.plot(), not pyplot.scatter(). The issue arrises when I want to color the points on this plot with a third variable, idx. I know pyplot.scatter() is capable of doing this quite easily, but I don't know how to do it with pyplot.plot().

My excerpt of code:

import os
import tempfile
from datetime import datetime
import numpy as np
os.environ['MPLCONFIGDIR'] = tempfile.mkdtemp()

import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
from matplotlib.backends.backend_pdf import PdfPages

pp = PdfPages('output.pdf')
names = ['WestHall', 'Elevator', 'EastHall', 'MathLounge']
locations = np.arange(4)+1
plt.scatter(timesin, locations, c=idx, marker="o")
plt.xlabel("Time of day")
plt.ylabel("Location")
plt.yticks(np.arange(4)+1, names)
plt.gcf().autofmt_xdate()
pp.savefig()
plt.close()
pp.close()

When I try this, I get an error, because it tries to interpret idx as rgba:

ValueError: to_rgba: Invalid rgba arg "[...]"
number in rbg sequence outside 0-1 range

How do I get it to interpret idx as conditional coloring without using pyplot.scatter()?

Thanks

Update:

As suggested by Hun, I actually can use pyplot.scatter() in this context by converting the datetime objects to numbers using matplotlibs dates library. Thus, figuring out how to use pyplot.plot() for conditional coloring was unnecessary.

You marked Hun's answer as the answer, but this doesn't answer your question. You should change your question to claim that Hun have answered it. — Alejandro, Apr 10 '16 at 20:34

score 2 · Accepted Answer · answered Apr 10 '16 at 19:41

2

It would be easier if you use plt.scatter(). But you need to convert the datetime to something scatter() can understand. There is a way to do it.

>>> dt # datetime numpy array
array(['2005-02-01', '2005-02-02', '2005-02-03', '2005-02-04'], dtype='datetime64[D]')

>>> dt.tolist() # need to be converted to list
[datetime.date(2005, 2, 1), datetime.date(2005, 2, 2), datetime.date(2005, 2, 3), datetime.date(2005, 2, 4)]

# convert the list to internal time information for matplotlib. But this is float.
>>> dt1 = matplotlib.dates.date2num(dt.tolist())
array([ 731978.,  731979.,  731980.,  731981.])

With this dt1 you can use plt.scatter()

answered Apr 10 '16 at 19:41

Hun

3,707
2
15
15

Do they plot as numbers or as time? – Nathan Goedeke Apr 10 '16 at 19:49
I will be plotted as number. Since you know what those number means you can use x-axis label formatter to display datetime. You can use num2date() matplotlib function. Extra work but I don't see any other way of doing it for your case. – Hun Apr 10 '16 at 19:54
It seems date2num() and num2date() are made for this kind of purpose. – Hun Apr 10 '16 at 19:56
Would you mind adding an example of setting the x-axis by num2date() please? – Nathan Goedeke Apr 10 '16 at 20:08
Have you tried to look up the document on your own? – Hun Apr 10 '16 at 20:11
Well, I was able to figure it out. Thank you so much for your help! – Nathan Goedeke Apr 10 '16 at 20:29

Alejandro · Answer 2 · 2016-04-10T20:25:20.030

0

I think that it is not possible to do this at once with matplotlib.pyplot.plot. However, here is my solution:

import matplotlib.pyplot as plt
import numpy as np
from matplotlib import cm

def scatterplot(x,y,prop):
    prop = cm.jet((prop-np.min(prop))/(np.max(prop)-np.min(prop)))
    ax = plt.gca()
    for i in xrange(len(x)):
        ax.plot(x[i],y[i],color=prop[i], marker='o')
    return

x = np.random.rand(100)
y = np.random.rand(100)
prop = -20+40*np.random.rand(100)

fig = plt.figure(1,figsize=(5,5))
ax  = fig.add_subplot(111)
scatterplot(x,y,prop)
plt.show()

which produces:

The only drawback of this approach is that if you have several particles, the process of walk through all of them can be relatively slow.

EDIT (In response to @Nathan Goedeke's comment:

I tried the same implementation but using a datetime object:

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

x = np.array([dt.datetime(2016, 10, 19, 10, 0, 0),
              dt.datetime(2016, 10, 19, 10, 0, 1),
              dt.datetime(2016, 10, 19, 10, 0, 2),
              dt.datetime(2016, 10, 19, 10, 0, 3)])

fig = plt.figure()
y = np.array([1, 3, 4, 2])
prop = np.array([2.,5.,3.,1.])
scatterplot(x,y,prop)
plt.show()

and it works as well.

edited Apr 10 '16 at 20:25

answered Apr 10 '16 at 19:23

Alejandro

3,263
2
22
38

Because my data to plot are datetime objects, I get an error when I try to implement this: `ValueError: ordinal must be >= 1` – Nathan Goedeke Apr 10 '16 at 19:43
Could you please write a minimum working example? I can not test it with your example... – Alejandro Apr 10 '16 at 19:47
You said that pyplot.plot worked for you, so I don't see why now you complain about this implementation, which actually uses pyplot.plot. – Alejandro Apr 10 '16 at 19:49
My example is the same as the code in my question only with the addition of `timesin` being an array of datetime objects. – Nathan Goedeke Apr 10 '16 at 19:53
I know pyplot.plot works, but for some reason when I tried your implementation I got a long traceback ending with `File "/usr/lib/pymodules/python2.7/matplotlib/dates.py", line 203, in _from_ordinalf dt = datetime.datetime.fromordinal(ix) ValueError: ordinal must be >= 1` – Nathan Goedeke Apr 10 '16 at 19:55
I have tested it with datetime objects and edited my answer. It works as well, so your error seems to depend on your particular implementation within dates.py rather than a problem with my solution. – Alejandro Apr 10 '16 at 20:26
Thanks, but I found it was easier to convert to numbers and use Hun's suggestion – Nathan Goedeke Apr 10 '16 at 20:30

matplotlib plot (not scatter) color based on extra variable

2 Answers2