python - 溢出错误 : int too big to convert when formatting date on pandas series plot

Question

我正在尝试绘制 Pandas 系列，但是当我尝试格式化 x 轴日期时遇到错误。
(在评论中发现了一个相关的 issue，但它似乎是在比我使用的更旧版本的 Pandas 中解决的。所以，这似乎是一个新问题。)
考虑以下 Pandas 系列的图:

import pandas as pd

d = {pd.Timestamp('2021-03-15 08:30:00'): -65.926651,
     pd.Timestamp('2021-03-15 08:30:05'): -42.115551,
     pd.Timestamp('2021-03-15 08:30:10'): -24.699627,
     pd.Timestamp('2021-03-15 08:30:15'): -12.010081,
     pd.Timestamp('2021-03-15 08:30:20'): -2.781321}

s = pd.Series(d)

ax = s.plot()

我试图使用以下方法格式化绘图上的 x 轴日期:

from matplotlib.dates import DateFormatter
format_str: str = '%H:%M:%S'
format_: DateFormatter = DateFormatter(format_str)
ax.xaxis.set_major_formatter(format_)

这会导致以下错误:

      Traceback (most recent call last):
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/matplotlib/backends/backend_macosx.py", line 61, in _draw
    self.figure.draw(renderer)
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/matplotlib/artist.py", line 41, in draw_wrapper
    return draw(artist, renderer, *args, **kwargs)
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/matplotlib/figure.py", line 1863, in draw
    mimage._draw_list_compositing_images(
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/matplotlib/image.py", line 131, in _draw_list_compositing_images
    a.draw(renderer)
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/matplotlib/artist.py", line 41, in draw_wrapper
    return draw(artist, renderer, *args, **kwargs)
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/matplotlib/cbook/deprecation.py", line 411, in wrapper
    return func(*inner_args, **inner_kwargs)
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/matplotlib/axes/_base.py", line 2747, in draw
    mimage._draw_list_compositing_images(renderer, self, artists)
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/matplotlib/image.py", line 131, in _draw_list_compositing_images
    a.draw(renderer)
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/matplotlib/artist.py", line 41, in draw_wrapper
    return draw(artist, renderer, *args, **kwargs)
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/matplotlib/axis.py", line 1164, in draw
    ticks_to_draw = self._update_ticks()
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/matplotlib/axis.py", line 1022, in _update_ticks
    major_labels = self.major.formatter.format_ticks(major_locs)
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/matplotlib/ticker.py", line 250, in format_ticks
    return [self(value, i) for i, value in enumerate(values)]
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/matplotlib/ticker.py", line 250, in <listcomp>
    return [self(value, i) for i, value in enumerate(values)]
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/matplotlib/dates.py", line 605, in __call__
    return num2date(x, self.tz).strftime(self.fmt)
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/matplotlib/dates.py", line 511, in num2date
    return _from_ordinalf_np_vectorized(x, tz).tolist()
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/numpy/lib/function_base.py", line 2108, in __call__
    return self._vectorize_call(func=func, args=vargs)
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/numpy/lib/function_base.py", line 2192, in _vectorize_call
    outputs = ufunc(*inputs)
  File "/Users/me/VirtualEnvironments/my_venv/lib/python3.9/site-packages/matplotlib/dates.py", line 331, in _from_ordinalf
    np.timedelta64(int(np.round(x * MUSECONDS_PER_DAY)), 'us'))
OverflowError: int too big to convert

有趣的是，如果我向时间戳添加小数偏移量，则一切正常:

s.index += pd.DateOffset(seconds=0.5)

当我检查 x在 np.timedelta64调用，仅当我向时间戳添加小数部分时，它才对应于自 unix 纪元(1970 年 1 月 1 日)开始以来的天数。如果没有小数部分，则结果整数很大，似乎与自 1970 年 1 月 1 日以来的天数没有明显关系。
这里有什么问题？

Answer 1

由于提供的数据超出了 DateFormatter 处理的数字范围，因此发生错误。
请引用official reference .
例如，第一个时间序列的实际数据如下所示

s.index[0].value
1615797000000000000

这需要转换为 matplotlib 可以处理的数字。

s.index = mdates.date2num(s.index)
s
18701.354167   -65.926651
18701.354225   -42.115551
18701.354282   -24.699627
18701.354340   -12.010081
18701.354398    -2.781321
dtype: float64

更新(我在 3.6.3，所以我正在修复它。)

ax = s.plot(style='o-')
import matplotlib.dates as mdates
format_str = '%H:%M:%S'
format_ = mdates.DateFormatter(format_str)
ax.xaxis.set_major_formatter(format_)