一、簡介

"""
@Author   ：葉庭云
@公眾號    ：AI庭云君
@CSDN     ：https://yetingyun.blog.csdn.net/
"""

Python的資料型別集合：由不同元素組成的集合，集合中是一組無序排列的可 Hash 的值（不可變型別），可以作為字典的Key

Pandas中的DataFrame：DataFrame是一個表格型的資料結構，可以理解為帶有標簽的二維陣列，

常用的集合操作如下圖所示：

二、交集

• pandas的 merge 功能默認為 inner 連接，可以實作取交集
• 集合 set 可以直接用 & 取交集

import pandas as pd

print("CSDN葉庭云：https://yetingyun.blog.csdn.net/")
set1 = {"Python", "Go", "C++", "Java"}
set2 = {"Go", "C++", "JavaScript", "C"}
set1 & set2

df1 = pd.DataFrame([
        ['1', 'Python'],
        ['2', 'Go'],
        ['3', 'C++'],
        ['4', 'Java'],
    ], columns=['id','name'])


df2 = pd.DataFrame([
        ['2','Go'],
        ['3','C++'],
        ['5','JavaScript'],
        ['6','C'],
    ], columns=['id','name'])

pd.merge(df1, df2, on=['id','name'])

操作如下所示：

三、并集

• Pandas的 merge 方法里引數 how 的取值有 “left”, “right”, “inner”, “outer”，默認是inner，outer外連接可以實作取并集，另一種方法也可以df1.append(df2)后去重，保留第一次出現的也可以實作取并集，
• 集合 set 可以直接用 | 取并集

set1 = {"Python", "Go", "C++", "Java"}
set2 = {"Go", "C++", "JavaScript", "C"}
set1 | set2

print("CSDN葉庭云：https://yetingyun.blog.csdn.net/")

df1 = pd.DataFrame([
        ['1', 'Python'],
        ['2', 'Go'],
        ['3', 'C++'],
        ['4', 'Java'],
    ], columns=['id','name'])


df2 = pd.DataFrame([
        ['2','Go'],
        ['3','C++'],
        ['5','JavaScript'],
        ['6','C'],
    ], columns=['id','name'])

pd.merge(df1, df2,
         on=['id','name'],
         how='outer')
         
df3 = df1.append(df2)
df3.drop_duplicates(subset=['id'], keep="first")

四、差集

set1 = {"Python", "Go", "C++", "Java"}
set2 = {"Go", "C++", "JavaScript", "C"}
set1 - set2

print("CSDN葉庭云：https://yetingyun.blog.csdn.net/")
set1 = {"Python", "Go", "C++", "Java"}
set2 = {"Go", "C++", "JavaScript", "C"}
set2 - set1

# df1-df2
df1 = pd.DataFrame([
        ['1', 'Python'],
        ['2', 'Go'],
        ['3', 'C++'],
        ['4', 'Java'],
    ], columns=['id','name'])


df2 = pd.DataFrame([
        ['2','Go'],
        ['3','C++'],
        ['5','JavaScript'],
        ['6','C'],
    ], columns=['id','name'])

df1 = df1.append(df2)
df1 = df1.append(df2)
set_diff_df = df1.drop_duplicates(subset=df1.columns,
                                  keep=False)
set_diff_df

# df2-df1
df1 = pd.DataFrame([
        ['1', 'Python'],
        ['2', 'Go'],
        ['3', 'C++'],
        ['4', 'Java'],
    ], columns=['id','name'])

df2 = pd.DataFrame([
        ['2','Go'],
        ['3','C++'],
        ['5','JavaScript'],
        ['6','C'],
    ], columns=['id','name'])

print("CSDN葉庭云：https://yetingyun.blog.csdn.net/")
df2 = df2.append(df1)
df2 = df2.append(df1)
set_diff_df = df2.drop_duplicates(subset=df2.columns,
                                  keep=False)
set_diff_df

# df1-df2
df1 = pd.DataFrame([
        ['1', 'Python'],
        ['2', 'Go'],
        ['3', 'C++'],
        ['4', 'Java'],
    ], columns=['id','name'])


df2 = pd.DataFrame([
        ['2','Go'],
        ['3','C++'],
        ['5','JavaScript'],
        ['6','C'],
    ], columns=['id','name'])

pd.concat([df1, df2, df2]).drop_duplicates(keep=False)

# df2-df1
df1 = pd.DataFrame([
        ['1', 'Python'],
        ['2', 'Go'],
        ['3', 'C++'],
        ['4', 'Java'],
    ], columns=['id','name'])


df2 = pd.DataFrame([
        ['2','Go'],
        ['3','C++'],
        ['5','JavaScript'],
        ['6','C'],
    ], columns=['id','name'])

pd.concat([df2, df1, df1]).drop_duplicates(keep=False)

五、對稱差集

print("CSDN葉庭云：https://yetingyun.blog.csdn.net/")
set1 = {"Python", "Go", "C++", "Java"}
set2 = {"Go", "C++", "JavaScript", "C"}
set1 ^ set2    # 對稱差集

# 去重   不保留重復的：即可實作取對稱差集
df3 = df1.append(df2)

df3.drop_duplicates(subset=['id'], keep=False)

推薦學習：

• https://www.jianshu.com/p/877e2bc11d93
• https://www.runoob.com/python3/python3-set.html