FK Distribution Strategies

By default, SqlProof picks parent rows uniformly when generating FK references. Distribution strategies let you simulate realistic skew or adversarial boundary cases.

Built-In Strategies

proof.customize(
    "orders",
    fk_distribution={"customer_id": "uniform"},
)

proof.customize(
    "orders",
    fk_distribution={"customer_id": "zipf"},
)

proof.customize(
    "line_items",
    fk_distribution={"product_id": "adversarial"},
)

Strategy	Behavior
uniform	Each parent row has equal probability
zipf	Early parents are referenced more often
adversarial	Only first, middle, and last parents are used
single	All children point to one parent

Custom Strategy

from hypothesis import strategies as st


def hottest_parent(parent_pks, ctx):
    return st.just(parent_pks[0])


proof.customize(
    "orders",
    fk_distribution={"customer_id": hottest_parent},
)

The callable returns a strategy, not a value, so Hypothesis can still shrink counterexamples.