With Redis, relationships are typically represented by sets. A set can be used
to represent a one-way relationship, so you need one set per object to
represent a many-to-many relationship.
It is pretty useless to try to compare a relational database model to Redis
data structures. With Redis, everything is stored in a denormalized way. Example:
# Here are my categories
> hset category:1 name cinema ... more fields ...
> hset category:2 name music ... more fields ...
> hset category:3 name sports ... more fields ...
> hset category:4 name nature ... more fields ...
# Here are my users
> hset user:1 name Jack ... more fields ...
> hset user:2 name John ... more fields ...
> hset user:3 name Julia ... more fields ...
# Let's establish the many-to-many relationship
# Jack likes cinema and sports
# John likes music and nature
# Julia likes cinema, music and nature
# For each category, we keep a set of reference on the users
> sadd category:1:users 1 3
> sadd category:2:users 2 3
> sadd category:3:users 1
> sadd category:4:users 2 3
# For each user, we keep a set of reference on the categories
> sadd user:1:categories 1 3
> sadd user:2:categories 2 4
> sadd user:3:categories 1 2 4
Once we have this data structure, it is easy to query it using the set algebra:
# Categories of Julia
> smembers user:3:categories
1) "1"
2) "2"
3) "4"
# Users interested by music
> smembers category:2:users
1) "2"
2) "3"
# Users interested by both music and cinema
> sinter category:1:users category:2:users
1) "3"
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…