Slice and index of array with data type

One 、 Slicing of arrays

Slice of the list

In understanding numpy Before the array slice of , Let's first look at list Slicing operation . obviously , Slicing will copy the selection of the original array and assign it to a new variable list_slice, In this way, when operating on slices , The original list value will not change .

list = [1 for x in range(10)]
print(list)
list_slice = list[3:5]
list_slice[1] = 12
print(list)
print(list_slice)
""" output: [1, 1, 1, 1, 1, 1, 1, 1, 1, 1] [1, 1, 1, 1, 1, 1, 1, 1, 1, 1] [1, 12] """

Automatic propagation of array slices

Perform slicing operations, such as arr[2:3]=1 when , This assignment will be automatically propagated to the original array . For example, the following operations ：

First create an all one array . Then on arr Slice and give another variable arr_slice. Then on arr_slice Assign a value to the second value of , Then the output arr. You can go to the original array arr It has also changed . This explanation , We are in the process of slicing , An array slice is just a reduced view of the original array . It means , Slicing does not copy data , Any numerical changes in the slice will be reflected in the original array .

arr_slice = arr[3:5]
arr_slice[1]=12
print(arr)
# output : [ 1. 1. 1. 1. 12. 1. 1. 1. 1. 1.]

Two 、 Boolean index

First, let's look at a failed case ：
You can see , The output result is false, Note: Boolean index operation cannot be directly performed on the list .

import numpy as np
name = ['wuqian', 'hua','wuqian', 'wanglihon']
date = np.random.randn(4,4)
name == 'wuqian'
#output：False

You should first convert the character list into array type .

import numpy as np
name = np.array(['wuqian', 'hua','wuqian', 'wanglihon'])
data = np.random.randn(4,4)
name == 'wuqian'  #  This operation is the same as arithmetic operation , Array comparison operations are also vectorized , Will be compared one by one , Then the output True or False
#  The output is a Boolean array 
""" output：array([ True, False, True, False]) """
#  Then the Boolean array can index the array 
data[name == 'wuqian']
"""  The output is data In the corresponding Boolean array of data True Location data . # output： array([[-0.05095068, -0.21921446, 0.30479235, 0.07790257], [-0.46580753, -0.3457661 , -0.40843969, -0.97014426]]) """

One point that needs great attention is
Use this index , The length of Boolean array must be consistent with the axis length of root index .
In addition, it can be used flexibly &、|、!= Boolean arithmetic symbols such as non, or greater than less than , Here's an example ：

huairen = (name == 'wuqian')|(name = 'hua')
name !='wuqian'
score = np.array([100, 99, 78, 12, 0, 65])
score = score.astype(np.unicode_)
score[score>'60'] = ' pass '
score
# output: array(['100', ' pass ', ' pass ', '12', '0', ' pass '], dtype='<U11')

3、 ... and 、Ndarray Data type of

1. Data type view

arr1 = np.array([1,3,4])
arr1.dtype
#  If no specific data type is specified , It's usually float64

Data type symbol
string_ Fixed length string type , One byte per character
unicode_ Fixed length unicode type , The number of bytes depends on the platform used
float64 Standard single precision floating point numbers
…

2. How to change the data type
You can specify the data type when creating data ：

arr = np.array([1,3,4,1],dtype = np.int64)

If the array has been established, it needs to be changed , You can do the following ：

arr = arr.astype(np.unicode)