Geospatial Data

Introduction

Redis geospatial commands (GEOADD, GEORADIUS, GEOSEARCH) store lat/lon as sorted-set members scored by geohash — Uber-style driver dispatch queries nearby entities in milliseconds without scanning every pin in the city.

Under the hood GEO is a ZSET; GEORADIUS uses geohash indexing for radius search. Distances returned in m, km, mi, or ft with WITHDIST and WITHCOORD options.

App-side haversine over millions of points fails at scale — push location indexing to Redis or a dedicated geo service.

Understanding the topic

Key concepts

• GEOADD key lon lat member — add location (note lon before lat).
• GEORADIUS key lon lat radius unit — nearby search.
• GEOSEARCH — Redis 6.2+ unified search API.
• GEODIST — distance between two members.
• Stored as ZSET — scores are geohash-encoded.
• GEOPOS — retrieve coordinates of members.

flowchart LR
  Mobile -->|lat lon| GEOADD
  GEOADD --> ZSet[Geohash ZSET]
  Dispatch -->|GEORADIUS| ZSet
  ZSet -->|driver IDs| Matcher

Step-by-step explanation

1. Lat/lon converted to 52-bit geohash score in ZSET.
2. Radius search checks geohash cells overlapping circle.
3. Results sorted by distance when requested.
4. Updates via GEOADD replace member position.
5. TTL applies to entire geo key like any key.

Syntax reference

Common commands

• Longitude first — common mistake is lat/lon swap.
• GEORADIUS COUNT 10 — limit results.
• Shard keys by city/region.

GEOADD drivers:nyc -73.985 40.758 driver:42
GEORADIUS drivers:nyc -73.985 40.758 2 km WITHDIST WITHCOORD
GEODIST drivers:nyc driver:42 driver:99 km

Informative example

Update driver location and find nearest five in Spring:

@Service
public class DriverGeo {
    private final StringRedisTemplate redis;

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

    public void updateLocation(String driverId, double lon, double lat) {
        redis.opsForGeo().add("drivers:sf", new Point(lon, lat), driverId);
    }

    public GeoResults<RedisGeoCommands.GeoLocation<String>> nearby(double lon, double lat) {
        return redis.opsForGeo().radius("drivers:sf",
            new Circle(new Point(lon, lat), new Distance(3, Metrics.KILOMETERS)),
            RedisGeoCommands.GeoRadiusCommandArgs.newGeoRadiusArgs().limit(5));
    }
}

Expire stale drivers with separate TTL key or periodic ZREM. Active pins only — not historical trail (use streams for that).

Real-world use

Real-world use cases

• Ride-share driver matching — GEORADIUS 3km.
• Store locator — nearest branches.
• Delivery fleet ETA input — distance sort.
• IoT asset tracking — last known position.
• Geofence entry — periodic radius check.

Best practices

• Partition by metro: drivers:{cityCode}.
• Remove inactive drivers — ZREM or key TTL.
• Limit GEORADIUS COUNT to cap work.
• Validate coordinates before GEOADD.
• Use read replicas for heavy read radius traffic.
• Combine with Kafka for location event pipeline.

Common mistakes

• Lat/lon argument order swapped.
• Single global key for all drivers worldwide.
• No cleanup of offline drivers — stale results.
• Storing full trip history in geo key.

Advanced interview questions

Summary

GEO = ZSET + geohash for radius search.