Your example is nt quite complete, so maybe I'm doing something wrong, but for

from matplotlib import pyplot as plt
import datetime

n, x = plt.hist([datetime.datetime(2018,1,1),  datetime.datetime(2018, 2, 1),
                 datetime.datetime(2018,3, 1)])

on 2.1.2 I get

Traceback (most recent call last):
  File "testHist.py", line 5, in <module>
    datetime.datetime(2018,3, 1)])
ValueError: too many values to unpack (expected 2)

So I don't think it worked on 2.1.2 either.

Thanks for the fast feedback!
Arghs, forgot to import the datetime in the example (updated issue now).
The value error ist the result of plt.hist() returning 3 instead of 2 variables, namely n, bins and patches.

Code variant with assigning output to variables:

from matplotlib import pyplot as plt
from datetime import datetime
n, x, p = plt.hist([datetime(2018,1,1),  datetime(2018, 2, 1),  datetime(2018,3, 1)])
print(n)  

Output in 2.12 is:
[1. 0. 0. 0. 0. 1. 0. 0. 0. 1.]

Output in 2.2.3 is:
[array([1., 0., 0., 0., 0., 0., 0., 0., 0., 0.]), array([0., 0., 0., 0., 0., 1., 0., 0., 0., 0.]), array([0., 0., 0., 0., 0., 0., 0., 0., 0., 1.])]

@jklymak have i been able to clarify your questions? Greetings!

This bisects to c02a84d "Correctly convert units for a stacked histogram" @dstansby #9654

Obviously this works for floats, so the issue is that your data is not floats. I guess this used to work before, but whether or not it really should have is debatable. What should

plt.hist(['a', 'b', 'c'])

return? It looks like the reason datetime worked at all is that it has a toordinal method.

I guess this used to work before, but whether or not it really should have is debatable.

To be able to create a histogram from a list of datetimes seems pretty beneficial to me.

from datetime import datetime
from matplotlib import pyplot as plt
plt.figure(figsize=(12, 6))
dates = [datetime(2010, 6, 15)] * 2 + \
        [datetime(2011, 6, 15)] * 3 + \
        [datetime(2012, 6, 15)] * 7 + \
        [datetime(2013, 6, 15)] * 4 + \
        [datetime(2015, 6, 15)] * 1
plt.hist(dates, bins=5, rwidth=0.8)
plt.xlabel('Year')
plt.ylabel('Number of events')

Old matplotlib version

Newer version

In contrast to dates, i haven't that often tried to create histograms of strings, .... 😕

This just caused plotting a 70MB file to explode to over 50GB of memory for me, quite surprising that I can't plot a histogram of datetimes.

Edit: The fix:
mpl_data = mdates.date2num(data)

Also try histtype='step' if it's slow to render

This should be fixed on master...

Awesome! I'm on a pretty old version, so I'll leave the explicit conversion for now