开启吐槽模式，已经被整过很多次了，numpy和pandas的boolean index里如果要使用逻辑操作符，必须用括号括起来。好蛋疼，好蛋疼！
The operators are: | for or, & for and, and ~ for not. These must be grouped by using parentheses.

In [1]: a = 1

In [2]: b = 2

In [3]: a > 0 and b < 2
Out[3]: False

In [4]: (a > 0) and (b < 2)
Out[4]: False

In [5]: import numpy as np

In [6]: a = np.arange(4)

In [7]: b = a**2

In [8]: a > 2 and b < 5
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
----> 1 a > 2 and b < 5

ValueError: The truth value of an array with more than one element is ambiguous.
 Use a.any() or a.all()

In [9]: (a > 2) and (b < 5)
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
----> 1 (a > 2) and (b < 5)

ValueError: The truth value of an array with more than one element is ambiguous.
 Use a.any() or a.all()

In [10]: a > 2 & b < 5
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
----> 1 a > 2 & b < 5

ValueError: The truth value of an array with more than one element is ambiguous.
 Use a.any() or a.all()

In [11]: (a > 2) & (b < 5)
Out[11]: array([False, False, False, False], dtype=bool)

只有最后一种方式才是`numpy`的正确打开方式！