Benchmark Guide

This page tracks reproducible performance snapshots for the public SQLQueryHelperjs runtime engines.

Scope

The benchmark suite is designed to measure representative hot paths instead of every exported method in isolation.

Current benchmark groups:

• query builder SQL generation
• DML roundtrips with returning support
• schema inspection
• entity generation
• planReflect(...)
• no-op reflect(...)

These scenarios cover the paths that matter most when comparing SQLite, PostgreSQL, and MySQL in this project.

How To Run

Use:

npm run bench
Copy

Environment variables:

• POSTGRES_BENCH_CONNECTION_STRING or POSTGRES_TEST_CONNECTION_STRING
• MYSQL_BENCH_CONNECTION_STRING or MYSQL_TEST_CONNECTION_STRING
• MYSQL_BENCH_SCHEMA or MYSQL_TEST_SCHEMA

SQLite uses a temporary local file under benchmarks/.tmp/.

The benchmark runner writes:

• benchmarks/latest-results.json
• benchmarks/latest-results.md
• benchmarks/history/<version>/results.json
• benchmarks/history/<version>/results.md
• benchmarks/history/<version>/comparison.json when a previous versioned baseline exists
• benchmarks/history/<version>/comparison.md when a previous versioned baseline exists

Latest Snapshot

Latest captured snapshot:

• generated: 2026-04-22T16:59:34.621Z
• host: Windows 11 (10.0.26100), Node v22.12.0
• CPU: AMD Ryzen 5 5600GT with Radeon Graphics

Scenario	SQLite avg ms/op	PostgreSQL avg ms/op	MySQL avg ms/op
build.select.toSQL	0.005	0.007	0.007
build.insert.toSQL	0.005	0.006	0.016
runtime.dml.roundtrip	12.688	7.743	22.281
schema.inspect.singleTable	0.381	19.203	10.374
schema.generate.singleTable	0.389	20.759	9.929
schema.planReflect.singleEntity	0.404	38.915	23.656
schema.reflect.noop.singleEntity	0.382	38.231	29.173

Quick read:

• SQLite is best for tight local schema and builder loops.
• PostgreSQL leads the live DML roundtrip benchmark in this snapshot.
• MySQL pays extra overhead in DML when returning(...) uses follow-up selection, but remains stronger than PostgreSQL in this snapshot's schema inspection/generation timings.

The source artifact for this snapshot is benchmarks/latest-results.md.

Reading The Numbers

• Build benchmarks (toSQL) show pure builder overhead without network cost.
• Runtime DML benchmarks show end-to-end reflector cost for insert/update/select/delete.
• Schema benchmarks show metadata and planning costs after the target schema is already aligned.
• Cross-engine comparisons are directional unless you pin hardware, Docker limits, server tuning, and dataset size.

History And Regressions

The runner now keeps the latest point-in-time snapshot and also archives versioned snapshots by package release.

When a previous versioned baseline exists, the runner also emits comparison artifacts for release review.

Use benchmark-history.md as the operating guide for reading and operating that workflow.

Coverage Notes

This page tracks the implemented runtime engines:

• SQLite
• PostgreSQL
• MySQL

SQL Server remains a documentation split target, not a benchmarked runtime engine yet.