问题
I have a few numpy arrays, lets say a, b, and c, and have created a mask to apply to all of them.
I am trying to mask them as such:
a = a[mask]
where mask is a bool array. It is worth noting that I have verified that
len(a) = len(b) = len(c) = len(mask)
And I am getting a rather cryptic sounding warning:
FutureWarning: in the future, boolean array-likes will be handled as a boolean array index
回答1:
False == 0, and True == 1. If your mask is a list, and not an ndarray, you can get some unexpected behaviour:
>>> a = np.array([1,2,3])
>>> mask_list = [True, False, True]
>>> a[mask_list]
__main__:1: FutureWarning: in the future, boolean array-likes will be handled as a boolean array index
array([2, 1, 2])
where this array is made up of a[1], a[0], and a[1], just like
>>> a[np.array([1,0,1])]
array([2, 1, 2])
On the other hand:
>>> mask_array = np.array(mask_list)
>>> mask_array
array([ True, False, True], dtype=bool)
>>> a[mask_array]
array([1, 3])
The warning is telling you that eventually a[mask_list] will give you the same as a[mask_array] (which is probably what you wanted it to give you in the first place.)
来源：https://stackoverflow.com/questions/38733897/numpy-boolean-array-warning