Redis counts new and retained users

Foreword

set (unordered, unique) sets provide methods like intersection, union, and difference.
You can use the set collection to count new users and retain users.

Statistics added

Assume that the system initially has five users, ID: 1,2,3,4,5
Set A to record users (cumulative users, that is, the users currently owned by the system)

sadd A 1sadd A 2sadd A 3sadd A 4sadd A 5

Set B to record users (users who have logged in that day).
On the first day here, suppose 3 and 5 are logged in, and 6 is newly registered. When the system is successfully registered, it will log in to the system by default, so it will also be written in B.

sadd B 3sadd B 5sadd B 6

The new user on the first day is the difference between B and A
By the way, make a new user record and use sdiffstore

# 6sdiff B A# Save B A difference to user_new:2022-07-28sdiffstore user_new: 2022-07-28 B A

The newly added user above is 6.In the end, 6 needs to be added to the cumulative user, set provides a method to obtain and save the union

# The first A stands for save there# Save A B set union overwrite to A# A has 1, 2, 3, 4, 5, 6 in itsunionstore A A B

The next day, let's say 3, 6 are logged in and 7, 8 are registered.This record is written to C, and B is kept as a historical record (B, C generally have names in the project similar to user: 2022-08-01).It can be processed according to the method of the first day (adding users: C,A difference, save C,A union to A)

sadd C 3sadd C 6sadd C 7sadd C 8

Statistics of retained users

Statistics of users who logged in on the first day and logged in on the second day
Seek the intersection of C and B, and save it to user:keep

# Save the intersection of C and B to user:keepsinterstore user:keep C B

Defects

Because the intersection, union and difference are more complicated.When the amount of data is relatively large, it will be very time-consuming.Since redis is a single process, it will cause the instance to block.
Solution 1: In master-slave mode, you can choose a slave library to execute.Due to read only from the library.As a result, sdiffstore, sinterstore, and sunionstore cannot be used.But you can use (sdiff,sinter,sunion) to get the result first, and then save it to the main library.
Solution 2: read it out, hand it over to the client program to disjoin, differ, and union