python memoryview 简要理解

最近要用到sklearn.neighbor.KDTree，它有一个attribute是data是指构造KDTree数据的memoryview。简单使用后发现memoryview有点像C里的指针

import numpy as np
from sklearn.neighbors import KDTree
from time import time

if __name__ == '__main__':
    pts = np.random.randn(10, 3)
    print("pts:", pts)
    # t1 = time()
    tree = KDTree(pts)
    # print("kdtree build time:", time() - t1)
    pts_rebuild_copy = np.array(tree.data)
    pts_rebuild_nocopy = np.array(tree.data, copy=False)
    pts_rebuild_copy[0] = [1, 1, 1]
    print("pts_1:", pts)
    pts_rebuild_nocopy[1] = [1, 1, 1]
    print("pts_2:", pts)

输出结果为

pts: [[ 0.06880616  0.96098154  1.17762945]
 [-1.17038063 -0.26971611 -1.63170026]
 [ 0.2531622   1.01200955  0.47513764]
 [ 1.14337371 -1.22593015 -1.64465175]
 [ 1.16018094 -1.24631525 -1.42440624]
 [-0.20000339 -1.83829021  1.12720473]
 [-1.47188812  0.57656556 -0.53477847]
 [-1.50508239 -1.02217681 -0.92497212]
 [ 0.14878263 -1.48558938  0.0131429 ]
 [-0.84247113 -0.64958466  1.60158359]]
pts_1: [[ 0.06880616  0.96098154  1.17762945]
 [-1.17038063 -0.26971611 -1.63170026]
 [ 0.2531622   1.01200955  0.47513764]
 [ 1.14337371 -1.22593015 -1.64465175]
 [ 1.16018094 -1.24631525 -1.42440624]
 [-0.20000339 -1.83829021  1.12720473]
 [-1.47188812  0.57656556 -0.53477847]
 [-1.50508239 -1.02217681 -0.92497212]
 [ 0.14878263 -1.48558938  0.0131429 ]
 [-0.84247113 -0.64958466  1.60158359]]
 pts_2: [[ 0.06880616  0.96098154  1.17762945]
 [ 1.          1.          1.        ]
 [ 0.2531622   1.01200955  0.47513764]
 [ 1.14337371 -1.22593015 -1.64465175]
 [ 1.16018094 -1.24631525 -1.42440624]
 [-0.20000339 -1.83829021  1.12720473]
 [-1.47188812  0.57656556 -0.53477847]
 [-1.50508239 -1.02217681 -0.92497212]
 [ 0.14878263 -1.48558938  0.0131429 ]
 [-0.84247113 -0.64958466  1.60158359]]

也就是说，tree.data就是指向内存中存放pts的地址，np.array(tree.data, copy=True)是对pts的deepcopy而np.array(tree.data, copy=False)是对pts的浅拷贝