examples

@perryts/postgres examples

Each .ts file in this directory is a standalone, runnable program — no test harness, no framework. Compile with Perry for a native binary, or run with node --import tsx / bun to use the driver under a JS host.

End-to-end smokes

These exercise the driver's public API against a real Postgres and print a short pass/fail line. They double as reference code for application authors.

File	What it does
perry-smoke.ts	Baseline: connect, auth, simple + extended queries, type codecs, error path, transaction
perry-smoke-tls.ts	sslmode=require + in-place TLS upgrade, SCRAM over TLS
perry-smoke-cancel.ts	conn.cancel() against a long-running pg_sleep(30)
perry-smoke-close.ts	conn.close() timing — FIN vs belt-and-braces timer
perry-smoke-notify.ts	NOTICE (RAISE NOTICE) + LISTEN / NOTIFY pub/sub
perry-smoke-codecs.ts	Round-trip every built-in scalar + array codec
perry-smoke-bulk.ts	1000-row × 20-column projection with timing

Running a smoke

# Node / Bun:
PGHOST=127.0.0.1 PGPORT=5432 PGUSER=you PGPASSWORD=pw PGDATABASE=db \
    node --import tsx examples/perry-smoke.ts

PATH=~/.bun/bin:$PATH PGHOST=… bun examples/perry-smoke.ts

# Perry-native (AOT-compiled binary):
/path/to/perry compile examples/perry-smoke.ts -o /tmp/perry-smoke
PGHOST=… /tmp/perry-smoke

All smokes honour the standard libpq environment variables (PGHOST / PGPORT / PGUSER / PGPASSWORD / PGDATABASE).

Microbenchmarks (diagnostic)

Not user-facing — these isolate specific hot paths that surfaced during Perry's perf work. Useful if you're investigating a regression.

File	Target
perry-smoke-bulk.ts	1000×20 end-to-end wall time
perry-bench-hidden-class.ts	10000×20 pure-JS dynamic-key obj build (no Postgres)
perry-bench-hidden-class-scaling.ts	Sweeps column count 5 → 80 to show O(N) vs O(N²) shape

See bench/ for the apples-to-apples matrix against pg, postgres.js, pg-native, and tokio-postgres.

Historical upstream-bug reproducers

Kept as a paper trail from the Perry fix chain — none of these should still reproduce a crash on Perry ≥ 0.5.87, but they're useful if a similar regression shows up.

File	Upstream issue
perry-bench-crash-repro.ts	PerryTS/perry#36 — module-level globals as GC roots
perry-bench-narrow.ts, perry-bench-narrow2.ts	Narrowing repros for the above
perry-tls-low-level.ts	Driver-free SSL upgrade test (from the TLS bring-up)
perry-driver-toplevel.ts	Minimal connect + query without the example harness
perry-net-sanity.ts	TCP round-trip against any postgres host
select-one.ts	Simplest possible SELECT 1