在对数组进行操作时,为了满足格式和计算的要求通常会改变其形状。
numpy.ndarray.shape
表示数组的维度,返回一个元组,这个元组的长度就是维度的数目,即ndim
属性(秩)。
【例】通过修改 shape 属性来改变数组的形状。
import numpy as np
x = np.array([1, 2, 9, 4, 5, 6, 7, 8])
print(x.shape) # (8,)
x.shape = [2, 4]
print(x)
# [[1 2 9 4]
# [5 6 7 8]]
numpy.ndarray.flat
将数组转换为一维的迭代器,可以用for访问数组每一个元素。
【例】
import numpy as np
x = np.array([[11, 12, 13, 14, 15],
[16, 17, 18, 19, 20],
[21, 22, 23, 24, 25],
[26, 27, 28, 29, 30],
[31, 32, 33, 34, 35]])
y = x.flat
print(y)
# <numpy.flatiter object at 0x0000020F9BA10C60>
for i in y:
print(i, end=' ')
# 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
y[3] = 0
print(end='\n')
print(x)
# [[11 12 13 0 15]
# [16 17 18 19 20]
# [21 22 23 24 25]
# [26 27 28 29 30]
# [31 32 33 34 35]]
numpy.ndarray.flatten([order='C'])
将数组的副本转换为一维数组,并返回。- order:’C’ — 按行,’F’ — 按列,’A’ — 原顺序,’k’ — 元素在内存中的出现顺序。(简记)
- order:{‘C / F,’A,K},可选使用此索引顺序读取a的元素。’C’意味着以行大的C风格顺序对元素进行索引,最后一个轴索引会更改F表示以列大的Fortran样式顺序索引元素,其中第一个索引变化最快,最后一个索引变化最快。请注意,’C’和’F’选项不考虑基础数组的内存布局,仅引用轴索引的顺序.A’表示如果a为Fortran,则以类似Fortran的索引顺序读取元素在内存中连续,否则类似C的顺序。“ K”表示按照步序在内存中的顺序读取元素,但步幅为负时反转数据除外。默认情况下,使用Cindex顺序。
【例】flatten()
函数返回的是拷贝。
import numpy as np
x = np.array([[11, 12, 13, 14, 15],
[16, 17, 18, 19, 20],
[21, 22, 23, 24, 25],
[26, 27, 28, 29, 30],
[31, 32, 33, 34, 35]])
y = x.flatten()
print(y)
# [11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
# 35]
y[3] = 0
print(x)
# [[11 12 13 14 15]
# [16 17 18 19 20]
# [21 22 23 24 25]
# [26 27 28 29 30]
# [31 32 33 34 35]]
x = np.array([[11, 12, 13, 14, 15],
[16, 17, 18, 19, 20],
[21, 22, 23, 24, 25],
[26, 27, 28, 29, 30],
[31, 32, 33, 34, 35]])
y = x.flatten(order='F')
print(y)
# [11 16 21 26 31 12 17 22 27 32 13 18 23 28 33 14 19 24 29 34 15 20 25 30
# 35]
y[3] = 0
print(x)
# [[11 12 13 14 15]
# [16 17 18 19 20]
# [21 22 23 24 25]
# [26 27 28 29 30]
# [31 32 33 34 35]]
numpy.ravel(a, order='C')
Return a contiguous flattened array.
【例】ravel()
返回的是视图。
import numpy as np
x = np.array([[11, 12, 13, 14, 15],
[16, 17, 18, 19, 20],
[21, 22, 23, 24, 25],
[26, 27, 28, 29, 30],
[31, 32, 33, 34, 35]])
y = np.ravel(x)
print(y)
# [11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
# 35]
y[3] = 0
print(x)
# [[11 12 13 0 15]
# [16 17 18 19 20]
# [21 22 23 24 25]
# [26 27 28 29 30]
# [31 32 33 34 35]]
【例】order=F 就是拷贝
x = np.array([[11, 12, 13, 14, 15],
[16, 17, 18, 19, 20],
[21, 22, 23, 24, 25],
[26, 27, 28, 29, 30],
[31, 32, 33, 34, 35]])
y = np.ravel(x, order='F')
print(y)
# [11 16 21 26 31 12 17 22 27 32 13 18 23 28 33 14 19 24 29 34 15 20 25 30
# 35]
y[3] = 0
print(x)
# [[11 12 13 14 15]
# [16 17 18 19 20]
# [21 22 23 24 25]
# [26 27 28 29 30]
# [31 32 33 34 35]]
numpy.reshape(a, newshape[, order='C'])
在不更改数据的情况下为数组赋予新的形状。
【例】reshape()
函数当参数newshape = [rows,-1]
时,将根据行数自动确定列数。
import numpy as np
x = np.arange(12)
y = np.reshape(x, [3, 4])
print(y.dtype) # int32
print(y)
# [[ 0 1 2 3]
# [ 4 5 6 7]
# [ 8 9 10 11]]
y = np.reshape(x, [3, -1])
print(y)
# [[ 0 1 2 3]
# [ 4 5 6 7]
# [ 8 9 10 11]]
y = np.reshape(x,[-1,3])
print(y)
# [[ 0 1 2]
# [ 3 4 5]
# [ 6 7 8]
# [ 9 10 11]]
y[0, 1] = 10
print(x)
# [ 0 10 2 3 4 5 6 7 8 9 10 11](改变x去reshape后y中的值,x对应元素也改变)
【例】reshape()
函数当参数newshape = -1
时,表示将数组降为一维。
import numpy as np
x = np.random.randint(12, size=[2, 2, 3])
print(x)
# [[[11 9 1]
# [ 1 10 3]]
#
# [[ 0 6 1]
# [ 4 11 3]]]
y = np.reshape(x, -1)
print(y)
# [11 9 1 1 10 3 0 6 1 4 11 3]