gajus/slonik @slonik/sql-tag@49.10.8 on GitHub

8a9878a Thanks @gajus! - Reuse a single parser instance for sql.unsafe
sql.unsafe constructed a fresh z.unknown() schema on every tagged-template evaluation, i.e. once per query for the most commonly used tag. Zod schema construction is not cheap: in a CPU profile of a steady-state query loop it accounted for roughly 10% of active CPU, spread across init, defineLazy and schema-class construction, and it was a significant contributor to garbage collection.
The schema carries no per-query state and zod schemas are immutable, so a single module-level instance now serves every token. parser is only ever read — it is surfaced to interceptors as resultParser — so sharing the instance is not observable.
Measured against PostgreSQL 17 over 60,000 queries at a pool size of 10: CPU per query fell from ~48.6 µs to ~39.1 µs (about 20%), and the garbage-collection share of active CPU fell from 5.8% to 1.2%.
Updated dependencies []:
- @slonik/errors@49.10.8
- @slonik/types@49.10.8