[Python] How to tell hypothesis.strategies to choose not-None for optional fields

I am a consumer of a class that I don't own:

from dataclasses import dataclass
@dataclass
class Child:
f1: int
f2: int | None
@dataclass
class Parent:
child: Child


The actual class is much much deeper and wider.

To run manual tests (not pytest, not CI), I want to create an instance of Parent with random values.

from typing import TypeVar
import hypothesis
import hypothesis.strategies
T = TypeVar("T")
def generate(cls: type[T], seed: int) -> T:
objects = []

@hypothesis.seed(seed)
@hypothesis.given(hypothesis.strategies.from_type(cls))
@hypothesis.settings(max_examples=10)
def f(o):
objects.append(o)

f()

# Firts example tends to be a trivial one so return the last one instead
return objects[-1]

print(generate(Parent))
print(generate(Parent))


output:

Parent(child=Child(f1=1234, f2=None))
Parent(child=Child(f1=-10, f2=4567))


I don't care about the values in the instance, except I do not want Nones for optional fields (I want Child.f2 to be a random int, always). I don't want to maintain code to instantiate one because the class often changes.

How can I update the snippet to tell hypothesis to never choose None?

Continue reading...