v0.19.0

@kitcn/resend

Published @kitcn/resend@0.19.0.

kitcn

Minor Changes

• #336 aba33f5 Thanks @MikeyZhang75! - ## Breaking changes

  • Read .select().where({ id: { in: [...] } }) in the order the ids are given
    instead of creation order. Each page reads only the listed positions it
    visits rather than the complete ID list; missing or policy-filtered ids still
    count as reads. Add orderBy to keep creation order. Cursors issued by the
    previous behavior are not portable.

  // Before — creation order, and every page read the whole id list
  await ctx.orm.query.posts
    .select()
    .where({ id: { in: ids } })
    .map((row) => row)
    .paginate({ cursor, limit: 10 });
  
  // After — same call, rows come back in the order `ids` lists them
  await ctx.orm.query.posts
    .select()
    .where({ id: { in: ids } })
    .map((row) => row)
    .paginate({ cursor, limit: 10 });
  
  // After — creation order, which still reads every id in the list
  await ctx.orm.query.posts
    .select()
    .where({ id: { in: ids } })
    .orderBy({ createdAt: "asc" })
    .map((row) => row)
    .paginate({ cursor, limit: 10 });

  Patches

  • Fix .distinct({ fields }) on more than one field growing exponentially more
    expensive with each row returned, which made pages of about twenty rows fail
    to return at all.
  • Fix .distinct({ fields }) over .union(...).interleaveBy(...) slowing down
    sharply as the number of returned rows grows.
  • Support maxScan on paginated where: { id: { in: [...] } } reads, including
    reads ordered by a single indexed field.
  • Improve paginated reads to register fewer Convex queries per page, and to
    release the ones they opened when a page stops early.
  • Improve the per-row cost of every stream-backed read.