mirror of
https://github.com/multica-ai/multica.git
synced 2026-07-29 06:28:23 +02:00
e8575139f3
Cursor-paginated useInfiniteQuery refetches every loaded page with its cached pageParam on invalidate. After a row is archived, page 0 (no cursor) shifts up to include items previously on page 1, while page 1 still uses a cursor pointing at the now-archived row — so the same item appears on both pages. The old client-side deduplicateInboxItems masked this; deleting it surfaced the bug. Mutations whose effect the client can predict EXACTLY apply optimistically and skip the list invalidate (only unread-count gets invalidated, single-query immune to the cross-page issue): - markRead / archive / markAllRead / archiveAll / archiveAllRead archiveCompletedInbox is the lone exception — its server-side filter on issue.status can't be enumerated client-side, so it still invalidates the list and inherits the rare cross-page risk on deeply paginated caches. Same trade-off as timeline today. Adds optimistic onMutate to archiveAll / archiveAllRead which previously fell through to invalidate-only. Also: - rename filterAllItems → removeMatchingItems with explicit doc; the predicate semantics are *removal* (true → drop), opposite of Array.filter, easy to invert by mistake. - 069 migration comment now echoes 068's CONCURRENTLY trade-off and sets a "revisit when inbox_item exceeds ~1M rows" threshold rather than handwaving "pre-production scale". Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>