How to Use Redis Sets for Powerful Intersection, Union, and Difference Statistics

Introduction

Redis is often used merely as a cache, but its native data structures—especially Set, Sorted Set, Bitmap, and HyperLogLog —are powerful tools for large‑scale statistical analysis in sign‑in systems, e‑commerce, and social networking applications.

Aggregation Statistics

Aggregating multiple sets allows you to compute intersections, unions, and differences, which are essential for tasks like finding common friends, daily new friends, or recent comments.

Intersection

To find common friends between two users, store each user's friend IDs in a set and use SINTERSTORE:

SINTERSTORE userid:new userid:20002 userid:20003

After execution, the key userid:new contains the intersection of userid:20002 and userid:20003.

Difference

For daily new‑friend counts, compute the difference between two consecutive day sets:

SDIFFSTORE user:new userid:20201102 userid:20201101

The resulting set user:new holds the friends added on 2020‑11‑02 that were not present on 2020‑11‑01.

Union

To count total new friends over two days, union the daily sets:

SUNIONSTORE userid:new userid:20201102 userid:20201101

The key userid:new now contains all friends added on either day.

Sorting Statistics

Ordered collections are needed for scenarios like displaying the latest comments first. Redis provides two ordered types:

List : preserves insertion order; pagination via LRANGE.

Sorted Set : orders by a user‑defined score; pagination via ZRANGEBYSCORE.

Lists are simple but limited to chronological order, while Sorted Sets allow flexible ranking based on any weight (e.g., timestamp, popularity).

Binary State Statistics

When a value is simply 0 or 1—such as sign‑in status—Redis Bitmap (implemented on top of String) offers extreme space efficiency. Each bit represents a user’s daily sign‑in.

Typical commands: SETBIT key offset 1 – mark a day as signed in. GETBIT key offset – check sign‑in status. BITCOUNT key – count total signed‑in days.

Key design example: userid:yyyyMM stores a month’s bitmap for a specific user.

SETBIT userid:10001:202011 1 1

GETBIT userid:10001:202011 1

BITCOUNT userid:10001:202011

For continuous‑day analysis (e.g., users who signed in for 20 consecutive days), perform bitwise AND across the daily bitmaps and then BITCOUNT the result.

Cardinality Statistics

Counting unique elements (UV) with a plain Set can be memory‑intensive for massive traffic. Redis HyperLogLog provides approximate cardinality with a fixed ~12 KB footprint, at the cost of a small error (~0.81%).

PFADD p1:uv 10001 10002 10003 10004

PFCOUNT p1:uv

When exact counts are required, fall back to a Set.

Overall Summary

Redis offers a rich set of data structures for statistical workloads: Set and Sorted Set support intersection and union; only Set supports difference. Bitmap can perform bitwise AND/OR/XOR across multiple keys, ideal for binary state tracking. List and Sorted Set both enable ordered retrieval, but Sorted Set provides flexible scoring.

For high‑cardinality, low‑precision counting, use HyperLogLog; for precise counting, use Set.

When dealing with very large datasets, consider offloading aggregation to a dedicated replica or performing it on the client side to avoid blocking the primary Redis instance.