numpy之简单，复杂分片索引

接上篇文章继续。
本章主要说明python中有高级索引分片功能，可以直接使用系统函数，也可以使用下标索引。

1. 普通索引获取ndarray对象中的部分元素：

• 使用系统函数slice(start,stop,step)

>>> a = np.arange(1,10)
>>> a
array([1, 2, 3, 4, 5, 6, 7, 8, 9])
>>> s = slice(2,8,1)
>>> print(a[s])
[3 4 5 6 7 8]
>>> s = slice(2,8,2)
>>> print(a[s])
[3 5 7]
>>> 

• 使用下标作为标识,一维的数据：

>>> a = np.arange(1,10)
>>> a[1:5:1]
array([2, 3, 4, 5])
>>> a[1:5:2]
array([2, 4])
>>> a[3]
4
>>> a[3:]
array([4, 5, 6, 7, 8, 9])
>>> a[3::2]
array([4, 6, 8])

• 二维数据（每个列表作为一个元素）

>>> a = np.arange(20).reshape(4,5)
>>> a
array([[ 0,  1,  2,  3,  4],
       [ 5,  6,  7,  8,  9],
       [10, 11, 12, 13, 14],
       [15, 16, 17, 18, 19]])
>>> a[::2]
array([[ 0,  1,  2,  3,  4],
       [10, 11, 12, 13, 14]])
>>> a[1:1]
array([], shape=(0, 5), dtype=int64)
>>> a[1:2]
array([[5, 6, 7, 8, 9]])

• 使用… 标识一个维度下的所有记录都选中：

>>> a = np.arange(20).reshape(4,5)
>>> a
array([[ 0,  1,  2,  3,  4],
       [ 5,  6,  7,  8,  9],
       [10, 11, 12, 13, 14],
       [15, 16, 17, 18, 19]])
>>> a[...,2]
array([ 2,  7, 12, 17])
>>> a[2,...]
array([10, 11, 12, 13, 14])

2. 高级索引获取某个位置的元素：

• 使用列表对应ndarray对象中的元素

>>> a
array([[ 0,  1,  2,  3,  4],
       [ 5,  6,  7,  8,  9],
       [10, 11, 12, 13, 14],
       [15, 16, 17, 18, 19]])
>>> a[[1,2,3],[2,3,4]]
array([ 7, 13, 19])
>>> 

• 使用ndarray对象获取元素

>>> a
array([[ 0,  1,  2,  3,  4,  5],
       [ 6,  7,  8,  9, 10, 11],
       [12, 13, 14, 15, 16, 17],
       [18, 19, 20, 21, 22, 23],
       [24, 25, 26, 27, 28, 29]])
>>> i = np.array([[1,2],[3,4]],dtype='int')
>>> j = np.array([[2,1],[4,3]],dtype='int')
>>> a[i,j]
array([[ 8, 13],
       [22, 27]])

• 使用普通索引使用的参数类型

>>> i = a[1:4,2:5]
>>> j = a[1:3,4:6]
>>> k = a[1,3,4,...]
>>> i
array([[ 8,  9, 10],
       [14, 15, 16],
       [20, 21, 22]])
>>> j
array([[10, 11],
       [16, 17]])
>>> k
array([[ 6,  7,  8,  9, 10, 11],
       [18, 19, 20, 21, 22, 23],
       [24, 25, 26, 27, 28, 29]])

• 布尔值类型的索引

>>> k[k>4]
array([ 6,  7,  8,  9, 10, 11, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28,
       29])

• 指定下标索引，包括正序和倒叙索引,多个索引数组（使用np.ix_）

>>> k
array([[ 6,  7,  8,  9, 10, 11],
       [18, 19, 20, 21, 22, 23],
       [24, 25, 26, 27, 28, 29]])
>>> k[[1,2]]
array([[18, 19, 20, 21, 22, 23],
       [24, 25, 26, 27, 28, 29]])
>>> k[[-1,-2]]
array([[24, 25, 26, 27, 28, 29],
       [18, 19, 20, 21, 22, 23]])
>>> 
>>> k[np.ix_([1,2],[3,4])]
array([[21, 22],
       [27, 28]])