Laravel Redis/Memcached缓存标签行为差异及设计原因咨询

Great question—this is a common point of confusion with Laravel's cache tag behavior for distributed stores like Redis and Memcached. Let's break down the "why" and the workaround clearly.

The Core Problem #

As you noted, when you associate a cache item with multiple tag combinations (like tag1+tag2, tag2+tag3, tag2+tag4 for the same tag-test1 key), you’d expect flushing tag2 to clear all items linked to that tag. But Laravel’s default implementation doesn’t do this—you’d have to explicitly flush each tag combination individually to invalidate those caches.

Why Laravel Uses This Design #

The short answer: performance prioritization for read-heavy applications.

Laravel’s default tag handling for Redis/Memcached uses a combination-based tracking system:

• When you store an item with Cache::tags(['tagA', 'tagB'])->put('key', 'value'), Laravel generates a unique key for that specific tag combination + item key, then adds that unique key to separate sets for tagA and tagB.
• Reads are fast because Laravel just looks up the precomputed unique key for the tag combination—no cross-referencing multiple tag sets during fetch operations.
• The tradeoff? Flushing a single tag only clears items that were explicitly stored with that standalone tag (or the exact same tag combination). Items stored with tag2 plus other tags aren’t in the standalone tag2 set, so they don’t get cleared when you run Cache::tags(['tag2'])->flush().

Tracking every individual tag association for items (instead of combinations) would add significant overhead to write operations—especially in high-traffic apps where you’re constantly adding/updating cached items with multiple tags. Laravel’s default design opts for fast reads over flexible flushes.

A Workaround for Expected Behavior #

If you need the "flush all items linked to any occurrence of a tag" behavior you described, there’s a community-maintained package that reimplements Laravel’s cache tags to support this. It tracks each item against every individual tag it’s associated with, rather than just tag combinations.

With this package, your example code:

Cache::tags(['tag1', 'tag2'])->get('tag-test1');
Cache::tags(['tag2', 'tag3'])->get('tag-test1');
Cache::tags(['tag2', 'tag4'])->get('tag-test1');
Cache::tags(['tag2'])->flush();

...would behave exactly as you expect: running Cache::tags(['tag2'])->flush() clears all three instances of tag-test1 without needing extra lines of code.

内容的提问来源于stack exchange，提问作者matastic