How to check a particular error was returned by a previous step?

[will-hart]

In some cases I want to assert that a particular error was returned when I perform a step. For instance in a toy example I would love to have:

Scenario: Divide by n
    Given I start with 6
    When I divide by 6
    Then I have 1
Scenario: Divide by 0
    Given I start with 6
    When I divide by 0
    Then I get a "divide by 0" error

How would I implement an expected failure? Ideally the When I divide by (\d+) step would be the same for both scenarios, but if I return an Err(_) result from the When step then the test fails.

Should I be caching the result of the When step in the World? e.g.

pub struct MySpecialInteger(u64);

#[derive(Debug, Default, World)]
pub struct MyWorld{
    number: MySpecialInteger,
    result: Result<u64, MyNumberError>
}

#[when(r"I divide by (\d+)"])
fn divide_by_n(world: &mut MyWorld, num: u64) {
    world.result = world.number.0 / num; // pretend this produces a Result
}

// pseudocode
#[then(r"I get a (\s+) error"])
fn divide_by_n(world: &mut MyWorld, expected_message: &str) {
    assert!(world.result.message.contains(expected_message));
}

I could also have a then step which is Then dividing by (\d+) produces error xyz, but this seems like mixing steps and assertions.

[ilslv]

I could also have a then step which is Then dividing by (\d+) produces error xyz, but this seems like mixing steps and assertions.

You are absolutely right, the recommended way is to store whole Results on when steps and assert them on thens.