Add test for "contains" with false "if" subschema

[marksparkza]

An intermediate solution to a problem I've been working on had contains failing when an if subschema failed.

My implementation only supports drafts 2019-09 and 2020-12, so I'm not 100% sure that this is a valid test for draft6 and draft7.

[karenetheridge]

These two commits (with slightly different messages) can be merged together into one.

[marksparkza]

Done!

[jdesrosiers]

How is this testing anything that isn't covered by https://github.com/json-schema-org/JSON-Schema-Test-Suite/blob/master/tests/draft2019-09/contains.json#L63-L78?

The if/else doesn't affect the behavior of contains in any way, so I'm not sure what value this test adds.

[karenetheridge]

@marksparkza 's implementation had a different behaviour:

An intermediate solution to a problem I've been working on had contains failing when an if subschema failed.

..so it sounds like perhaps his handling of 'if' was incorrect? perhaps the tests should be moved to if.json, as the subschema containing an 'if' should not be false if the 'if' keyword itself is false (and there is no 'else'). It's not relevant to 'contains' (but please correct me if I'm wrong @marksparkza )

[marksparkza]

@jdesrosiers

The if/else doesn't affect the behavior of contains in any way

The if/else shouldn't affect the behavior of contains in any way, but in my implementation it did. I was passing all tests, but had a hunch that I'd missed something. So I wrote this test, which my implementation failed.

By way of explanation:

In solving a larger problem (I was losing annotations and errors for array items, even though validation was being done correctly), I'd modified my contains implementation in such a way that it did not account for the behaviour of non-asserting keywords - not if specifically, though that's the only relevant case. I've since revised my implementation such that keyword non-assertion is dealt with transparently with respect to other keywords.

@karenetheridge

This was specifically a problem with the way my contains implementation processed its own subschema, so I think contains.json is the right place for the test. I think the more general problem of a subschema not being false if if is false is probably adequately covered by the first test ("ignore if without then or else") in if-then-else.json.

[jdesrosiers]

If it's exposing a real bug, that's good enough for me. Carry on.

[Julian]

The above sounds reasonable to me (if it was a real bug an implementation had, we generally I think merge these kinds of things).

Merging, thanks!