Cache
rspc-cache is a server-side caching prototype. It is separate from TanStack Query's frontend cache.
Do not use this prototype to cache production data. In the checked implementation, every request uses the same hard-coded cache key. Different procedures, inputs, and users can collide. A collision can return the wrong value or panic on a type mismatch.
How the prototype works
The implementation provides:
CacheState::builder(Memory::new()).mount()to install a store through router setup.cache()to wrap a procedure.cache_ttl(seconds)to set a time to live while the resolver runs.
Only successful results with a TTL are stored. The middleware requires cloneable inputs and results. It expects cache state to be mounted and can panic if that state is missing.
Cache in an application
Until cache keys and isolation are implemented, put caching in application code. Include every value that affects a response in the key, including the user or tenant where needed. Define expiration and invalidation rules, and do not cache authorization decisions indefinitely.
For client-side caching and refetching, use TanStack Query.
Last updated on