Brownie: Hypothesis given() is not resetting chain for each function call

Question

I have the following test code snippet in my brownie project:

from hypothesis import settings, given, strategies as st
from brownie import *
import brownie

st_buy_amount_eth = st.integers(min_value=int(Wei("1 gwei")), max_value=int(Wei("10 ether")))

@given(buy_amount=st_buy_amount_eth)
@settings(max_examples=20)
def test_buy_lgt_atleast_for(lgt, accounts, buy_amount):
    print(web3.eth.blockNumber)
    token_amount = buy_amount_to_tokens(buy_amount)  # hard-coded function to verify the math
    if token_amount >= 1:
        tx = lgt.buyLgtAtLeastFor(1, 99999999999, accounts[3], {'from': accounts[2], 'value': buy_amount})
        assert tx.return_value == token_amount
        assert lgt.balanceOf(accounts[3]) == token_amount
    else:
        with brownie.reverts("dev: not enough eth to buy tokens"):
            liquidLgt.buyLgtAtLeast(1, 99999999999, {'from': accounts[2], 'value': buy_amount})

When running the tests with brownie test -s the print output I get is one block higher for each example with hypothesis and I get inconsistent results errors from hypothesis.

Is there a way to set it up so hypothesis will properly reset the chain for each example? That would be awesome, I really hope to get hypothesis to work with brownie!

Answer 1

You need to import @given from Brownie, not directly from hypothesis:

from brownie.test import given

This is required because Brownie handles isolation via a fixture, but function scoped fixtures only execute once per test (not once per run). From the hypothesis documentation:

... each fixture will run once for the whole function, not once per example.

Brownie solves this by wrapping the test in a function that performs the chain rewind. You can see it handled here in the source code.