gajus/slonik @slonik/dataloaders@49.10.1

on GitHub

Patch Changes

• #810 cf370a9 Thanks @mikeroelens! - Fix O(n²) record matching in createNodeByIdLoaderClass

  The node-by-id DataLoader batches all keys into a single query, but then
  matched the returned rows back to the requested keys using
  Array.prototype.find inside a map, making the in-memory join O(n²) in the
  batch size. For large batches this added significant synchronous, event-loop
  blocking CPU time after the query had already returned (e.g. ~800ms for a
  10,000-key batch).

  The rows are now indexed into a Map once and looked up in O(1) per key —
  mirroring the grouping approach already used in batchQueries. Behaviour is
  unchanged, including first-match-wins semantics for duplicate key values.