`SeeAnyOf` and `SeeAllOf` have a few inconsistencies.

[perrygoy]

With the recent merge of #64, @bandophahita noticed that there are a couple ways that SeeAllOf and SeeAnyOf are inconsistent:

• When asserting no tests, SeeAllOf() will pass, while SeeAnyOf() will fail.
• When executing tests, SeeAllOf will short-circuit at the first failure, while SeeAnyOf will execute all tests.

I think that the desired behavior is:

• When asserting no tests, both should pass.
• When executing tests, all tests should be executed (no short-circuiting).

[bandophahita]

When executing tests, all tests should be executed (no short-circuiting).

Consider this code:

actor.shall(See(q1, r1))
actor.shall(See(q2, r2))

If the first test fails, pytest will never execute the 2nd test unless they were to wrap it in a try/except. Shouldn't SeeAllOf mimic that same behavior?

[perrygoy]

If the first test fails, pytest will never execute the 2nd test unless they were to wrap it in a try/except. Shouldn't SeeAllOf mimic that same behavior?

Maybe that's one reason to use SeeAllOf instead of a series of Sees, so they can perform each test every time to see the result?

[perrygoy]

After more consideration, i think (personally) that SeeAnyOf() and SeeAllOf() should actually raise an exception for not having any tests. I can only imagine someone put them in like that by mistake, and either a test would be passing when it is missing its assertion or there's just an unnecessary line in their log.

Either way, it doesn't make sense to me that someone would want that intentionally—i think raising an exception to call out the mistake is the best UX.

[perrygoy]

After more consideration, and more conversation with @bandophahita, we've decided to change course on this:

• SeeAllOf and SeeAnyOf should both short-circuit, mimicking Python behavior.
• SeeAllOf and SeeAnyOf should both accept 0 tests and pass.
• SeeAllOf and SeeAnyOf should both accept 1 test.

[perrygoy]

Merged and fixed!