decentraland / comms-gatekeeper / 25224471412

85%

main: 85%

Build Type

Pull #261

github

Committed by LautaroPetaccio

Commit Message

refactor: use @dcl/memory-cache-component for the webhook refresh cooldown

replaces the local Map<string, number> + manual prune helper inside
roomMetadataSync with an injected ICacheStorageComponent. the cache
entry's existence is the cooldown signal; ttl handles eviction; the
lru cap (10k) handles bounded memory automatically.

upsides:
- no more manual gc (pruneStaleRefreshEntries gone) and no timestamp
  arithmetic — the api is now `cache.get` / `cache.set(key, true, ttl)`.
- ttl is in seconds, matching the component's contract.
- the cache is injected, so swapping to a redis-backed implementation
  later (e.g., for cluster-wide cooldown) is a one-line change in
  components.ts with no consumer updates.

trade-off worth knowing: the get/set pair is now async, so two
concurrent webhooks for the same room could theoretically both pass
the gate before either set lands. in practice this is a microsecond
window on an in-memory cache; worst case is one duplicate refresh,
which is negligible compared to the ~200x reduction we already get
on busy scenes.

cache keys are namespaced (`room-metadata-sync:webhook-refresh:<room>`)
so the cache instance can be shared with future consumers without key
collisions.

Pull Request Pull Request #261: refactor: extract room metadata sync and add scene admins to metadata

Coverage Stats

1061 of 1362 branches covered (77.9%)

Branch coverage included in aggregate %.

82 of 83 new or added lines in 9 files covered. (98.8%)

7 existing lines in 3 files now uncovered.

2999 of 3391 relevant lines covered (88.44%)

43.31 hits per line