CODEWITHSUNDEEP
hashredis
Yusuf Kayikci
Yusuf Kayikci

Reputation: 332

Redis hash usage as table

I want to use redis like Nosql database and I have some idea like below.

Assume that I have 3 table

1 - user
2 - post
3 - comment

I create hash for each table like below

hset user _usr_100 {"id":"_usr_100","name":"john","username"="jhn","age":25}
hset user _usr_101 {"id":"_usr_101","name":"adam","username"="adm","age":26}
hset user _usr_102 {"id":"_usr_102","name":"eric","username"="erc","age":27}

hset post _post_100 {"id":"_post_100","title":"title","content":"testpost","userid"="_usr_100"}
hset post _post_101 {"id":"_post_101","title":"title","content":"testpost","userid"="_usr_101"}
hset post _post_102 {"id":"_post_102","title":"title","content":"testpost","userid"="_usr_102"}

hset comment _comment_100 {"id":"_comment_100","content":"testpost","userid"="_usr_100","postid":"_post_100"}    
hset comment _comment_101 {"id":"_comment_101","content":"testpost","userid"="_usr_101","postid":"_post_101"}    
hset comment _comment_102 {"id":"_comment_102","content":"testpost","userid"="_usr_102","postid":"_post_102"}

When I want get user(_user_100) from redis 

hget user _usr_100
{"id":"_usr_100","name":"john","username"="jhn","age":25}

When I want get users

hgetall user
{"id":"_usr_100","name":"john","username"="jhn","age":25}
{"id":"_usr_101","name":"adam","username"="adm","age":26}
{"id":"_usr_102","name":"eric","username"="erc","age":27}

Afer deserialize json string one by pne and fill them in list , I have List so I can do some operation (search,groupby,order,pagination ...) and I can do same thing for another hashes(post,comment)

I can delete,update user with;

hdel user _usr_101 // deleted _usr_101
hset user _usr_100 {"id":"_usr_100","name":"john","username"="jhn","age":26} //updated age
hset user _usr_103 {"id":"_usr_103","name":"max","username"="max","age":15} //new user

hgetall user
{"id":"_usr_100","name":"john","username"="jhn","age":26}
{"id":"_usr_102","name":"eric","username"="erc","age":27}
{"id":"_usr_103","name":"max","username"="max","age":15}

What can be disadvantage of this usage?Can you suggest another idea about hash to use redis like nosql tables.

Upvotes: 1

Views: 783

Answers (1)

Ersoy
Ersoy

Reputation: 9624

Depending on your business rules/model, this option "may" work but it may not be the best/near the best solution for your domain. Using key/value store in the need of mostly relational domain cause you to make tradeoffs which may be disadvantage for you.

When your user class has new fields and this fields needed to be queried then you need to create more "space" to reduce the "time". You keep denormalizing your data to just achieve a single query. You will try to implement your relational database in the key/value store world. When you just need to update your user 101 with a simple statement;

UPDATE users SET username = 'mynewusername' where id = 101;

In your case you will need to find all related keys/fields through all hash/set/lists and update them for the data integrity. Keeping age as a field may be a bad idea, you will need to use birthday or and if your business needs to fetch list of users's whose birthday is today then you need to create new keys, duplicate most of your data, migrate all your existing users to there to just get the today's birthdays. It's better to keep that in mind, you need to query by day and month to get birthdays - which means that you have to keep users in separate sets such as users:birthday:01:01, users:birthday:02:05, users:birthday:11:08 to fetch them. If the users wants to update their birthday(depending on the business) then you need to manually move users between those sets while updating the other sets too.

Adding active/passive to users will be another pain. I am not sure whether you need to get all users, you may need to paginate them and while using hash - it will be hard, You will need another another sorted set/list to gain that.

Same goes for comments of posts of the users, last 25 comments of the user, most recent comments of the users who have the most posts or searching through posts of users etc etc. Your product manager will come with the idea, let's add tag to each post and you will need to relate this into your data model with new data structures.

These are relational data, it is better to keep them relational. When you start modeling your data in non-relational database all the elasticity rdbms provide you will be gone and it will be replaced with complexity on both data and application layer.

A single postgresql may boost you far better than redis in this problem. Redis has excellent features to solve problems but user/post/comment is not one of them.

This post may provide some insights too

Upvotes: 1

Related Questions