Changelog

[0.1.3] — 2026-04-03

Bug: CacheStore.set() incorrectly deleted parsed key arrays immediately after caching them, forcing redundant re-parsing on every subsequent invalidation
Bug: Duplicate keyArrays.delete(key) call in startGcTimer cleanup
Test: Timestamp race condition in security test for corrupted cache keys
Test: Set Vitest environment to jsdom globally so DOM-dependent tests run correctly

matchesKey() now uses reference equality fast-path (===) before falling back to JSON.stringify comparison — significantly faster for string/number key segments

Added main, module, import export condition, sideEffects: false, engines, keywords, homepage, repository, bugs, and author fields to package.json
Included README.md, LICENSE, and CHANGELOG.md in published package files
Version bumped to 0.1.3

cache.dehydrate() — serialize cache to a JSON-safe snapshot for SSR
cache.rehydrate(state) — seed client cache from server snapshot; prevents loading flash
DehydratedState type exported from public API
Global onError hook on CacheConfig — fires once after all retries exhausted (queries pass key array, mutations pass [])
retry now accepts (failureCount: number, error: Error) => boolean in addition to number — allows skipping retries on 401/404
timeout on CacheConfig and QueryConfig — aborts fetch after N ms; each retry attempt gets a fresh timeout
onEvent hook on CacheConfig — typed event bus for all cache lifecycle events
CacheEvent discriminated union exported from public API: fetch:start, fetch:success, fetch:error, invalidate, set, gc, rehydrate
Per-query retry and timeout overrides on QueryConfig

cache.mutate() — reactive mutation with optimistic update support (onMutate, onSuccess, onError, onSettled)
cache.invalidate(key) — prefix-based cache invalidation
cache.prefetch(config) — populate cache without creating a reactive result; deduplicates concurrent prefetches
cache.getQueryData(key) / cache.setQueryData(key, data) — direct cache read/write
cache.cancelQuery(key) — abort in-flight request
Reactive keys — key: () => ['todos', userId] re-runs query when dependencies change
select transform — cache.query<Todo[], ActiveTodo[]>({ select: ... })
keepPreviousData — show stale data during key transitions
refetchOnReconnect — refetch stale queries on network reconnect
Exponential backoff on retry (capped at 30s)
gcTime — prune inactive cache entries after configurable idle period (default 5 min)
Request deduplication for concurrent identical queries