在 matplotlib 中绘制 y=times（作为数据）与 x=dates：如何格式化所有日期的 y 轴？

Question

All the SO answers relating to "plotting times vs. dates" did not help with my problem:
Given two series of datetime data, I want to plot the time part versus the date part.

Using the code below, I get the desired formatting, but only for the first point (of both series).
I can't figure out how to define the y-axis limits (or format) that would work for all dates.

Thanks in advance for pointing out my mistake(s)!

import numpy as np
import pandas as pd
from matplotlib import pyplot as plt
import matplotlib.dates as mdates

data = np.array(
    [( 0, '2021-04-14T07:45:00.000000000', '2021-04-14T22:42:00.000000000'),
     ( 1, '2021-04-15T06:37:00.000000000', '2021-04-15T23:20:00.000000000'),
     ( 2, '2021-04-16T06:45:00.000000000', '2021-04-16T22:45:00.000000000'),
     ( 3, '2021-04-17T06:35:00.000000000', '2021-04-17T23:01:00.000000000'),
     ( 4, '2021-04-18T06:30:00.000000000', '2021-04-18T22:50:00.000000000'),
     ( 5, '2021-04-19T06:14:00.000000000', '2021-04-19T23:05:00.000000000'),
     ( 6, '2021-04-20T07:10:00.000000000', '2021-04-21T00:00:00.000000000'),
     ( 7, '2021-04-21T06:37:00.000000000', '2021-04-21T22:30:00.000000000'),
     ( 8, '2021-04-22T07:25:00.000000000', '2021-04-22T23:40:00.000000000'),
     ( 9, '2021-04-23T06:24:00.000000000', '2021-04-23T23:45:00.000000000')],
     dtype=[('index', '<i8'), ('up_dt', '<M8[ns]'), ('down_dt', '<M8[ns]')])

# the actual data comes from a csv, so I use pd for further manipulations
df = pd.DataFrame.from_records(data)


fig, ax1 = plt.subplots(figsize=(9,7))

line1 = ax1.plot(df.up_dt.dt.date, df.up_dt,
                 label='rise time',
                 marker='^', linewidth=0, color='b')
line2 = ax1.plot(df.down_dt.dt.date, df.down_dt,
                 label='bed time',
                 marker='v', linewidth=0, color='b')

# x_axis ranges & format:
ax1.set_xlim([df.up_dt.min()-pd.DateOffset(days=1),
              df.up_dt.max()+pd.DateOffset(days=1)]);
ax1.xaxis.set_major_locator(mdates.DayLocator(interval=1))
ax1.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m-%d'))

# y_axis ranges & format:
d1 = str(df.up_dt.min().date())+' 00:00'
d2 = str(df.up_dt.min().date())+' 23:59'
y_time = pd.date_range(start=d1, end=d2,freq='H')
ax1.set_ylim([y_time.min(), y_time.max()])

ax1.yaxis.set_major_locator(mdates.HourLocator(byhour=range(24), interval=2))
ax1.yaxis.set_major_formatter(mdates.DateFormatter('%H:%M'))

ax1.set_ylabel('Clock Hours')
fig.autofmt_xdate()

VER = """python: 3.7.6 [Windows 10, MSC v.1916 64 bit (AMD64)]
pandas: 1.2.3; numpy: 1.19.2; matplotlib: 3.3.4"""

plt.title(F'Missing data if y-limit set with "df.<series1>.min()":\n{VER}')

plt.show();

Output:

Answer 1

If you take out the line where you set the y limits, you see that the y values are datetimes, not just times. So you have a couple of choices:

1) Set all the dates the same date for the time column.

You are hiding the date on the y axis, so this is the quickest replacement in your code. The snippet below only shows the added or changed lines from your code block.

import datetime
...
# make a new column with one date and the time from up_dt
df["up_time"] = df.up_dt.apply(lambda d: datetime.datetime(2021, 1, 1, d.hour, d.minute, d.second, d.microsecond))
...
# plot using the new time column for the y values
line1 = ax1.plot(df.up_dt.dt.date, df.up_time,
                 label='rise time',
                 marker='^', linewidth=0, color='b')
...
# use the new time column when finding the y limits
d1 = str(df.up_time.min().date())+' 00:00'
d2 = str(df.up_time.min().date())+' 23:59'

2) Use decimal representation of time

If you want to strip off just the time portion and not use a fake date for a place-holder, you need to convert it to a number, because time objects are not treated numerically by matplotlib, or for the standard datetime package. Using the pendulum package we can convert a time to decimal representation, here I am converting to hours since midnight. You can replace the tick labels with the clock representation strings.

import pendulum as pend
...
# make separate up and time columns
df["up_date"] = df.up_dt.apply(lambda d: d.date())
df["up_time_hr"] = df.up_dt.apply(lambda d: (pend.parse(d.isoformat()).time() - pend.time(0)).seconds/3600)

# plot time vs date using the new columns
fig, ax1 = plt.subplots(figsize=(9,7))
line1 = ax1.plot(df.up_date, df.up_time_hr)

# example of setting tick labels
ax1.set_yticks([ 6.5, 7, 7.5])
ax1.set_yticklabels(["6:30", "7:00", "7:30"])