v0.17.4

@kitcn/resend

Published @kitcn/resend@0.17.4.

kitcn

Patch Changes

• #331 23c0c99 Thanks @MikeyZhang75! - ## Patches

  • Improve mutation latency on tables with triggers, aggregateIndex() or
    rankIndex(). Writes no longer re-read the document after patching or
    replacing it, and no longer read it at all when no hook consumes it.
  • Improve update() and delete() cascade latency: set null, set default
    and cascade-update fan-out now apply their patches concurrently with a
    bounded pool instead of one round trip at a time. Tables with hooks keep
    their strict write order.
  • Improve read throughput on row-level-security tables. Policy using /
    withCheck callbacks run once per query execution or write-free mutation
    decision batch instead of once per returned row. Multi-row insert and delete
    re-resolve stateful policies after each write.
  • Fix row-level-security visibility when one query object is awaited more than
    once. Every await re-resolves the table's policies, so a policy that reads
    the database sees writes made between the two awaits.
  • Improve CPU cost of every ORM read and every returning() row by reshaping
    documents in a single pass.
  • Improve returning({ _count }) on multi-row mutations: the aggregate-index
    readiness check now runs once per index instead of once per row.
  • Improve cascade delete throughput on schemas with many triggered or
    aggregate-indexed tables by removing per-row table-name probing.
  • Improve mutations that issue concurrent writes to a hooked table: the
    internal write lock now hands off to one waiter instead of waking all of
    them.
  • Fix kitcn codegen dropping nested cRPC HTTP routes when parsing projects
    through generated server placeholders.