检查Redis中的多个键

Question

So here's an example of what the keys I have set in redis:

SADD group:test subgroup:1
SADD subgroup:1 user:1
SADD subgroup:1 user:2
SADD group:test subgroup:2
SADD subgroup:2 user:3

So what I want to achieve is that I want to find out which subgroup has space in the key group:test and return that. Let's say for example that the maximum users per sub group is 2.

In that case, subgroup 2 would have space so it should return subgroup:2 .

If there isn't any space I'd like it to return the last subgroup +1. So in that case it would return subgroup:3 .

Currently, the way this is written means that I have to use smembers to get the members of the group:test set and loop through each key using scard to see which sub group has space. If the loop ends and there's no space in any key, then I just add 1 to the last subgroup and use that.

Users can leave a subgroup after they have joined. I used multi to do this, but it doesn't seem very efficient as it continues running scard even after it has found space in a key. So let's say in this example user:2 left subgroup:1 , it would still check subgroup:2 to see if it had space.

I'm more than willing to restructure the data if there's a way to make this work. The only thing I don't want to change is that the key for users needs to be stored within the subgroup set.

If there's a way to do this with a LUA script, I'd also be willing to do that.

I'm really looking for an efficient way to do this.

Can anyone help with this?

Answer 1

A Lua script isn't the way here - rewording the above requirement: we're looking for an efficient way to get the cardinality of Sets (eg subgroup:1 ) at any instant, and we're specifically interested in Set the one with the least members ("has space").

To do that, you keep another key in your database - lets call it counts . That key holds a Sorted Set data structure, where each element is the name of one of your sets and the score is the its current cardinality.

By "current" I mean that each SADD , SREM and DEL that your code performs on subgroups, needs to be coupled with the relevant ZADD on the counts key.

When you have all that in place, all that you need to do is query the key for the lowest-scored member:

ZRANGE counts 0 0

The cost in maintaining this extra structure is the complexity of updates whenever subgroup changes. Sorted Sets' operations are typically O(log(N)) so they're pretty efficient and wouldn't burden the server too much.