Namespace Conventions

A cache is one big shared pool of keys, so naming matters. The convention is colon-delimited namespaces that read from general to specific: user:42:profile. Good key names prevent collisions, make the cache easy to reason about, and let you find or clear related keys by their shared prefix.

A naming pattern

const key = (userId: number) => `user:${userId}:profile`;

await cache.set(key(42), JSON.stringify(profile), { EX: 300 });
const cached = await cache.get(key(42));

Mirror your entity hierarchy

Structure keys to reflect the relationships in your data model. If users own posts and posts have comments, that nesting should show up in the key: user:42:post:7:comments. A reader can look at any key and immediately know what entity it belongs to, where it sits in the hierarchy, and what other keys are related to it.

This also makes invalidation natural. When a user is deleted, all keys under user:42: are candidates for eviction. When a post is updated, user:42:post:7: keys are the scope to clear.

Why conventions help

• No collisions — user:42 and post:42 never clash, even though both end in 42
• Grouping — every key under user:42: belongs to one user and can be cleared together
• Versioning — bump a prefix (v2:user:42:profile) to invalidate a whole shape of data at once
• Shared caches — prefix by service name (auth:, catalog:) when multiple services share a cache instance so their keys never overlap

Pick one convention and apply it everywhere. A consistent scheme is what keeps a shared cache from turning into a junk drawer.

Exercise