Table Customization

customize() lets you tune generated data while keeping the rest of the schema generation automatic.

from hypothesis import strategies as st

proof.customize(
    "products",
    price=st.decimals(min_value="0.01", max_value="9999.99", places=2),
    sku=st.from_regex(r"^[A-Z]{2}-\d{4}$", fullmatch=True),
)

FK Distribution Strategies

Control how child rows pick parent rows:

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

Strategy	Behavior
uniform	Equal probability per parent
zipf	A few parents receive most child rows
adversarial	First, middle, and last parents only
single	All children point to one parent
custom callable	Return any Hypothesis strategy per FK

Custom FK distribution callables receive the available parent primary keys and a draw context, and must return a Hypothesis strategy so values still shrink.