Snapshot-aware query result caching with Redis — any native support, or is Event Listener + Redis Smart Cache the right pattern?

[Vinoth721]

Hi all,

I'm running Trino against an Iceberg catalog and want to cache query results in Redis, with cache invalidation tied to the Iceberg snapshot_id of the tables involved — i.e. serve from cache as long as the underlying table hasn't changed, and automatically refresh the moment it has.

I've already built this as an external proxy (caches on query text + combined snapshot_id of all referenced tables, checks "<table>$snapshots" before serving), and it works. Before I invest further, two questions:

1. Is there any native support for this in Trino or Redis?

From what I've found so far, there isn't:

• Trino's Redis connector queries Redis as a data source — it doesn't cache Trino query results.
• Open-source Trino has no built-in result cache (I believe this is a Starburst Enterprise-only feature, not in OSS).
• Redis Smart Cache (Redis Inc.'s JDBC wrapper) caches results, but only on a TTL — no concept of Iceberg snapshots, so it can't auto-invalidate the instant a table changes.

Is that accurate, or is there a coordinator property, SPI hook, or plugin I'm missing?

2. If not — does an Event Listener + Redis Smart Cache combination make sense as a lower-effort alternative to a full custom proxy?

Idea: use Trino's [Event Listener SPI](https://trino.io/docs/current/develop/event-listener.html) to detect QueryCompletedEvents where a table was written to (INSERT/UPDATE/MERGE/DELETE), and on that event, delete/invalidate the relevant keys in Redis. Redis Smart Cache would handle the actual query-result caching/serving mechanics, so I'd only need to write the invalidation trigger rather than a full caching layer.

Has anyone tried this combination? Any gotchas with relying on Event Listener firing reliably as the invalidation trigger, versus checking snapshot_id directly like my current proxy does?

Thanks!