In depth understanding of pandas sorting mechanism

author ：Peter edit ：Peter

Hello everyone , I am a Peter~

In a previous article , Describes in detail how to use pandas Built in functions for sort_values To sort the data . This article explains how to use custom methods to achieve sorting ：

• Mapping relationship implementation
• CategoricalDtype Type implementation

<!--MORE-->

Analog data

First simulate a simple data ：

import pandas as pd
import numpy as np
df = pd.DataFrame({
"nick":["aaa","bbb","aba","abc","cac","ccc"], # nickname
"math":[100,120,130,111,100,128], # mathematics
"english":[140,80,120,90,125,116], # English
"size":["S","M","L","XS","XL","L"] # Clothing size
})
df

sort_values

DataFrame.sort_values(by,
axis=0,
ascending=True,
inplace=False,
kind='quicksort',
na_position='last', # last,first; The default is last
ignore_index=False,
key=None)

The specific interpretation of the parameter is ：

• by： Indicates what field or index to sort by , It can be one or more
• axis： Is the sorting on the horizontal or vertical axis , The default is the vertical axis axis=0
• ascending： Whether the sorting result is in ascending or descending order , The default is ascending
• inplace： Indicates whether the sorting result is directly modified in place on the original data or generated a new DatFrame
• kind： Indicates the algorithm using sorting , Quick line up quicksort,, Merger mergesort, Heap sort heapsort, Stable sequencing stable , The default is ： Quick line up quicksort
• na_position： Location handling of missing values , The default is last , Another option is to first
• ignore_index： Whether the index of the newly generated data frame is rearranged , Default False（ Use the index of the original data ）
• key： Functions used before sorting

Here are a few simple examples to review sort_values Use ：

Single field sorting

adopt nick Field sorting , Strings are based on letters ASCII code ; The default is ascending from small to large . Same first letter , Compare the second , Reason by analogy ：

Arrange in ascending order according to the size of the values ：

You can change the sorting method to descending ：

Sorting multiple fields

Sorting multiple fields at the same time , Default is also ascending . When the value of the first field is the same , Then arrange them in ascending order according to the second field

Assign different sorting methods to different fields ：

Then compare the two different ways completely ：

Above is sort_values Methods .

Custom sort

Use sort_values Methods are used to sort by the size of the built-in alphabetic or numeric data , When you encounter the following situations , How to operate ？

When we according to the size of the clothes size Sort by , And what you get is ：

Obviously, this sort of sorting is not what we expected , In our cognition ：

• XS： Very small
• S： Small
• M： secondary
• L： Big
• XL： Super large

How to solve this problem ？ There are two ways ：

Method 1： By mapping

1、 Find each size The value size corresponding to the order of

2、 Generate new fields order

3、 We are right. order Sort

Method 2： Use CategoricalDtype

CategoricalDtype Is a type of categorical data with a category and order , Can create our custom sort data type . Official website address ：

https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.CategoricalDtype.html

1、 Specify a classified data type CategoricalDtype

category_size = pd.CategoricalDtype(
['XS', 'S', 'M', 'L', 'XL'],
ordered=True)
category_size

2、 take size The field is set to the above CategoricalDtype type

3、 We're right size Use sort_values We can achieve our goal , And the above map The effect of mapping is the same

And by looking at df Data type of , We also see size The type is category：