Check Options

proof.check() and @sqlproof share the same core ideas: table sizes, run count, optional seed, optional setup, and a property function.

Decorator Form

@sqlproof(
    proof,
    sizes={"customers": 20, "orders": 100, "line_items": 500},
    runs=100,
    seed=1708891234,
    timeout_ms=5000,
)
def test_order_totals_non_negative(db):
    rows = db.query("SELECT total FROM orders")
    assert all(row["total"] >= 0 for row in rows)

Imperative Form

def check_totals(db):
    rows = db.query("SELECT total FROM orders")
    assert all(row["total"] >= 0 for row in rows)


proof.check(
    name="order totals are non-negative",
    sizes={"customers": 20, "orders": 100},
    property=check_totals,
    runs=100,
)

Fields

Field	Description
sizes	Per-table row counts for generated datasets
property	Callable that asserts the SQL property
setup	Optional callable run after data insertion
runs	Number of generated datasets to test
seed	Reproduce a specific generation and shrink trace
timeout_ms	Per-run timeout in milliseconds
commit	Use schema-isolation mode instead of rollback

SqlProofClient

The db object exposes convenience helpers:

rows = db.query("SELECT id, total FROM orders")
total = db.scalar("SELECT SUM(total) FROM orders")
affected = db.execute("UPDATE orders SET status = %s", "confirmed")
dataset = db.get_generated_data()

Queries use PyGreSQL internally. Prefer these helpers so tests stay independent of the driver.