Beyond the core types, Redis offers specialized structures — HyperLogLog, bitmaps, geospatial, bitfields — and modules (RedisJSON, RediSearch, RedisTimeSeries, RedisBloom) that extend Redis for specialized use cases efficiently.
PFADD visitors:today
PFCOUNT visitors:today
PFMERGE visitors:week visitors:mon visitors:tue
HyperLogLog estimates cardinality (unique counts) using tiny fixed memory — ideal for counting unique visitors, IPs, or events at scale where exact counting would cost enormous memory.
# bitmaps: bit-per-item flags (very memory-efficient)
SETBIT active:2024-01-15 1000 1 # mark user 1000 active today
BITCOUNT active:2024-01-15 # how many users were active (count set bits)
BITOP AND active:both d1 d2 # users active BOTH days (bitwise AND)
# → track daily active users for millions in a few KB per day
Bitmaps store bit-per-item flags extremely compactly — perfect for daily-active-user tracking, feature flags per user, and presence at scale.
GEOADD places 13.361 38.115 "Palermo" # add a location (lng, lat)
GEOSEARCH places FROMMEMBER "Palermo" BYRADIUS 200 km ASC # find nearby
RedisJSON → store/query JSON documents natively (paths, partial updates)
RediSearch → full-text search and secondary indexing on Redis data
RedisTimeSeries→ time-series data (metrics, IoT) with downsampling/aggregation
RedisBloom → Bloom filters / Cuckoo filters (probabilistic membership)
RedisGraph (legacy), and others
→ Modules turn Redis into a multi-model platform (Redis Stack bundles popular ones).
Understanding Redis's advanced data structures and modules is valuable senior-level knowledge for solving specialized problems efficiently with Redis, expanding what Redis can do beyond core types.
The probabilistic/compact structures are particularly powerful for scale: HyperLogLog (estimating unique counts in tiny fixed memory ~12KB, ideal for counting unique visitors/IPs/events at scale where exact counting via sets would be prohibitively expensive — a dramatic memory win), bitmaps (bit-per-item flags storing things like daily-active-users for millions in a few KB, with efficient bitwise operations for analytics), and bitfields — these solve real large-scale problems (cardinality, presence tracking, flags) with remarkable memory efficiency that's hard to achieve otherwise. Geospatial support enables location queries directly in Redis.
The modules — RedisJSON (native JSON documents), RediSearch (full-text search and secondary indexing), RedisTimeSeries (time-series/metrics with aggregation), RedisBloom (probabilistic membership filters) — extend Redis into a multi-model platform (bundled in Redis Stack), letting it handle search, JSON, time-series, and probabilistic use cases that would otherwise need separate systems.
Understanding these lets you leverage Redis for specialized needs efficiently rather than reaching for additional infrastructure or accepting poor memory/performance.
Since specialized problems (large-scale unique counting, presence tracking, search, time-series, JSON) arise in real systems, and since Redis's advanced structures (HyperLogLog, bitmaps, geospatial) and modules (RedisJSON, RediSearch, RedisTimeSeries, RedisBloom) solve them efficiently, understanding advanced Redis data structures and modules is valuable senior-level knowledge that reveals Redis's full capabilities as a versatile, multi-model platform, enabling elegant, memory-efficient solutions to specialized problems and reflecting deep Redis expertise beyond the common caching use case.