If I make a list of time deltas, the average is larger than if I average the microsecond values from these deltas. Why is that?

如果我列一个时间增量的列表，平均值比我对这些增量的微秒值求平均时要大。为什么会这样呢？

import time
import numpy as np
l = []
l_m = []
for _ in range(5):
    start = datetime.datetime.now()
    time.sleep(1)
    l.append((datetime.datetime.now() - start))
    l_m.append((datetime.datetime.now() - start).microseconds / 1000000)
print(l)
print(np.mean(l), np.mean(l_m))

gives

赠送

[datetime.timedelta(seconds=1, microseconds=846), datetime.timedelta(seconds=1, microseconds=1017), datetime.timedelta(seconds=1, microseconds=1010), datetime.timedelta(seconds=1, microseconds=1013), datetime.timedelta(seconds=1, microseconds=887)]
0:00:01.000955 0.9639999999999999

This is Python 3.8.10 on Linux.

这是Linux上的Python3.8.10。

Reading the docs, I see that ‘timedelta’ objects can be added. It can also be divided by a scalar so the average of the ‘timedelta’ is correct.

阅读文档后，我发现‘time Delta’对象可以添加。它还可以被标量除以，因此‘time Delta’的平均值是正确的。

Thanks to the comments, I see that the ‘microsecond’ attribute is just the microsecond portion of the ‘timedelta’ and does not include other components such as the seconds values.

多亏了这些评论，我看到‘micro秒’属性只是‘time Delta’的微秒部分，不包括其他组件，如秒值。