How can I convert a datetime into a floating point representation that is used by matplotlib's x-axis? matplotlib.dates.date2num
doesn't seem to work as per below...
import pandas as pd
import matplotlib as mpl
import datetime
idx = pd.DatetimeIndex(
[
'2021-12-21 10:25:00 00:00', '2021-12-21 10:30:00 00:00',
'2021-12-21 10:35:00 00:00', '2021-12-21 10:40:00 00:00',
'2021-12-21 10:45:00 00:00', '2021-12-21 10:50:00 00:00',
'2021-12-21 10:55:00 00:00', '2021-12-21 11:00:00 00:00',
'2021-12-21 11:05:00 00:00', '2021-12-21 11:10:00 00:00'
],
dtype='datetime64[ns, Europe/London]',
name='time',
freq=None
)
s = pd.Series(
[0.6, 1.8, 2.7, 1.8, 1.5, 2.1, 0.9, 1.8, 3.2, 4.4],
index=idx
)
fig, ax = plt.subplots()
s.plot(ax=ax)
target_x = datetime.datetime(2021, 12, 21, 18)
target_y = 2
ax.scatter(mpl.dates.date2num(target_x), target_y, marker='x')
The problem seems to be here:
>>> ax.get_xlim()
(27334705.0, 27334750.0)
whereas
>>> mpl.dates.date2num(target_x)
18982.75
References
- Converting between datetime, Timestamp and datetime64
- Python matplotlib.dates.date2num: converting numpy array to matplotlib datetimes
CodePudding user response:
It's Epochs / 60.
>>> ax.get_xticklabels()
[Text(27334705, 0, '10:25'),
Text(27334740, 0, '11:00'),
Text(27334750, 0, '11:10')]
So I noticed difference between the numbers in matplotlib representation is equal to difference in minutes between datetimes. And then you can confirm
>>> idx[7].timestamp() / 60 #'2021-12-21 11:00:00 00:00'
27334740.0
It checks out.
I don't know about the reasoning behind this, or a prettier method - just noticed this relation.
Code below draws the mark - had to add timezone information to your target datetime:
import pandas as pd
import matplotlib as mpl
import datetime
import pytz
idx = pd.DatetimeIndex(
[
'2021-12-21 10:25:00 00:00', '2021-12-21 10:30:00 00:00',
'2021-12-21 10:35:00 00:00', '2021-12-21 10:40:00 00:00',
'2021-12-21 10:45:00 00:00', '2021-12-21 10:50:00 00:00',
'2021-12-21 10:55:00 00:00', '2021-12-21 11:00:00 00:00',
'2021-12-21 11:05:00 00:00', '2021-12-21 11:10:00 00:00'
],
dtype='datetime64[ns, Europe/London]',
name='time',
freq=None
)
s = pd.Series(
[0.6, 1.8, 2.7, 1.8, 1.5, 2.1, 0.9, 1.8, 3.2, 4.4],
index=idx
)
fig, ax = plt.subplots()
s.plot(ax=ax)
target_x = datetime.datetime(2021, 12, 21, 11, 0, 0, 0, pytz.UTC)
target_y = 2
ax.scatter(int(target_x.timestamp()/60), target_y, marker='x')