[Python] Randomizing a user dataclass with pytest and hypothesis

I can manually define an Address builder strategy:

import attrs
from hypothesis import given
import hypothesis.strategies as st

@attrs.frozen(kw_only=True)
class Address:

street: str
city: str

AddressStrategy = st.builds(
Address,
street=st.text(),
city=st.text()
)

@given(AddressStrategy)
def test_proper_address(address):
assert len(address.city) < 4


When I run pytest, it indeed catches my bug:

address = Address(street='', city='0000') # <--- counterexample address found - good !

@given(AddressStrategy)
def test_proper_address(address):
> assert len(address.city) < 4
E AssertionError: assert 4 < 4
E + where 4 = len('0000')
E + where '0000' = Address(street='', city='0000').city

main.py:23: AssertionError


According to the docs, it seems like it should be possible to use an auto-generated address builder:

builds() will be used automatically for classes with type annotations on init ...
​

But when I try the following options, neither work:

• st.register_type_strategy(Address)
• @given(Address)

Continue reading...