如何在numpy中创建一个布尔数组

Question

I have to deal with a large data (300 by 4 by 1400000 integer array) which is around

4byte * 300 * 4 * 1400000 = 6.72GB

However, this array only contains 0 or 1. So if I could use a boolean array, then the size would shrink to 6.25% of the original data.

6.72GB / 4bytes / 8bits/bytes * 2 = 420Mbits 

Is there any way to make use of boolean arrays in numpy?

EDIT: I don't know why he deleted his answer but this did exactly what I wanted.

arr = np.ones((300,2,1400000), dtype = np.bool)

This lead to 12.5% compression.

>>> arr = np.ones((300,2,1400000), dtype = np.bool)
>>> arr.nbytes
840000000

>>> arr = np.ones((300,2,1400000))
>>> arr.nbytes
6720000000

840000000/6720000000 = 12.5%

Answer 1

While there is a way to manipulate 'bit fields' in numpy these don't offer the conveniences of proper numpy arrays on a bit level.

That said, numpy does have "logical" or "boolean" arrays, ie arrays with dtype bool. These take only one byte per element and are proper arrays. When your array is created by a "logical array operation" like, say, b = (a > 0) it ( b ) will be automatically of bool type. You can obtain boolean arrays by the standard numpy ways a.astype(bool), array(..., dtype=bool) etc.

Answer 2

这将使您减少 75%，每个项目四个字节到每个项目一个字节。

bool_array = np.logical_and(int_array, True)