redis-core

Match the data type to your access pattern, not the data shape. Use Strings for counters, Hashes for objects with independently updated fields, Sorted Sets for leaderboards, Sets for membership checks, Lists for queues, JSON for nested data, Streams for event logs, and Vector Sets for similarity search.

Use a Hash when you need atomic per-field reads and updates on a flat object. Use Redis JSON when your data is nested or hierarchical and you need path-level updates, nested arrays, or RQE indexing.

Use lowercase, colon-separated segments in a stable hierarchy like `user:1001:profile`. Keep keys short but readable, avoid spaces and mixed casing, never use full URLs as keys, and add tenant prefixes like `tenant:42:user:7:cart` for multi-tenant applications.

Updating one field in a serialized string requires fetching, parsing, mutating, and rewriting the entire object, which is slow and not atomic. A Redis Hash lets you update individual fields atomically with HSET without touching other fields.

Use a Sorted Set for leaderboards and rankings. It stores members with scores and supports ZADD for updates, ZRANGE for ranked retrieval, and ZRANK for position lookups, all ordered by score.