GraphQL resolvers can be treated as pure functions, allowing their results to be memoized and cached based on query arguments.
Time-based invalidation is simple but can serve stale data, whereas event-based invalidation keeps data synced but is more complex to implement.
A cache can use fast in-memory storage for small, frequent data or a shared Redis instance for larger datasets across multiple nodes.
By tagging cache entries with references to the underlying data entities, you can precisely invalidate them when a write event occurs.
A Fastify server is configured with the Mercurius plugin to serve a GraphQL schema with user and country data from an in-memory database.
The Mercurius Cache plugin is configured with a time-to-live policy for a query, using onHit and onMiss events to verify its effectiveness.
An invalidation function is added to a mutation resolver, which uses references to clear the relevant user data from the cache upon update.
Mercurius Cache also supports custom key generation, request deduplication, programmatic invalidation with wildcards, and garbage collection for references.
Common GraphQL pitfalls are discussed, such as neglecting to use DataLoaders to solve the N+1 query problem.
Hubert Burda Media
München, Germany
46:47Ruwan Xaviour Fernando
28:12Gregor Bauer
47:48William Lyon
40:45Josh Goldberg
41:06David Leitner
34:28Uri Goldshtein
42:08Matteo Collina
30:27Thijs Feryn
Jobs that call for the skills explored in this talk.
Okeiro
Paris, France
nono
nono
Jobriver Hr Service
Berlin, Germany
&why GmbH
Berlin, Germany
SMG Swiss Marketplace Group
Sachseln, Switzerland
ITproposal B.V.
Amsterdam, Netherlands
