Sorted Sets

Introduction

Sorted sets (ZSET) combine a unique member with a double score — Redis maintains skip-list + hash index for O(log N) rank, range, and score queries. This is the type behind leaderboards, priority queues, and time-series indexes.

ZADD updates score; ZREVRANGE returns top-N; ZRANK gives position. No other in-memory store matches this ergonomics for real-time ranking at scale.

Scores can be timestamps, points, or composite values — members are typically entity IDs.

Understanding the topic

Key concepts

• ZADD key score member — insert or update score.
• ZREVRANGE key 0 9 WITHSCORES — top 10 descending.
• ZRANGE BYSCORE — filter by score band.
• ZINCRBY — atomic score increment.
• ZRANK/ZREVRANK — position of member.
• ZREM, ZCARD — delete and count.

flowchart TB
  ZADD --> SkipList
  SkipList --> ZREVRANGE[Top N query]
  SkipList --> ZRANK[Rank lookup]

Step-by-step explanation

1. Each member appears once; score determines order.
2. Skip list enables log-time rank and range.
3. Same member re-ZADD updates score in place.
4. Lex ordering available with special score encoding.
5. Union/inter store ops merge multiple zsets.

Syntax reference

Common commands

• ZINCRBY — game points, vote counts.
• ZREVRANGE 0 9 — top 10 leaderboard.
• Score as Unix timestamp — delayed job queue.

ZADD leaderboard 9850 user:1001
ZINCRBY leaderboard 50 user:1001
ZREVRANGE leaderboard 0 9 WITHSCORES
ZRANK leaderboard user:1001

Informative example

Leaderboard service with Spring Redis ZSet ops:

@Service
public class Leaderboard {
    private final StringRedisTemplate redis;

    public Leaderboard(StringRedisTemplate redis) {
        this.redis = redis;
    }

    public void addScore(String board, String userId, double delta) {
        redis.opsForZSet().incrementScore("lb:" + board, userId, delta);
    }

    public Set<ZSetOperations.TypedTuple<String>> top10(String board) {
        return redis.opsForZSet().reverseRangeWithScores("lb:" + board, 0, 9);
    }
}

Trim board with ZREMRANGEBYRANK if millions of players — keep top 10K only.

Real-world use

Real-world use cases

• Gaming leaderboard — ZINCRBY on match end.
• Twitter timeline rank — score = timestamp.
• Priority task queue — score = priority or run-at time.
• Rate limit sliding window — ZADD timestamp member.
• Trending topics — increment score per mention.

Best practices

• Trim large zsets — ZREMRANGEBYRANK remove low ranks.
• Use member IDs not heavy JSON as members.
• Paginate with ZREVRANGE start stop not full scan.
• Composite scores need documented encoding.
• Replicate read-heavy boards to replicas.
• Cache top-N in app for sub-ms if needed.

Common mistakes

• Millions of members never trimmed.
• Using member JSON with large payload.
• ZRANGE entire zset in API hot path.
• Float score equality bugs in range queries.

Advanced interview questions

Summary

ZSET = Redis superpower for ranking.