You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
However, it's not entirely clear whether property names can have annotations. The reason being that a JSON Pointer can't point to a property name. The pointer in this test technically points to the property value, not the property name.
Implementations could just use the standard pointer and users/tooling just need to know that it's a special case, but special cases make programmatic use of annotations problematic and implementations might choose to avoid that. Implementations could choose also use a custom pointer syntax or flag to indicate that the pointer points to the property name, but that wouldn't be interoperable.
Since there's no interoperable and accepted solution to this problem, I suggest we remove this test. That would mean the test suite would be ambiguous about whether property names can have annotations or how they are implemented if they do.
What do you think?
The text was updated successfully, but these errors were encountered:
Actually, I realized that this test wasn't testing what I thought it was. This is testing that propertyNames doesn't annotate the value. That's a good and important thing to test. There isn't actually a test for annotations on property names. This suite isn't even capable of expressing such a test because it relies on JSON Pointers. So, I don't think there's anything that needs to be addressed here after all.
In the new annotation test suite, we have the following test.
JSON-Schema-Test-Suite/annotations/tests/applicators.json
Lines 76 to 99 in 9f256c8
However, it's not entirely clear whether property names can have annotations. The reason being that a JSON Pointer can't point to a property name. The pointer in this test technically points to the property value, not the property name.
Implementations could just use the standard pointer and users/tooling just need to know that it's a special case, but special cases make programmatic use of annotations problematic and implementations might choose to avoid that. Implementations could choose also use a custom pointer syntax or flag to indicate that the pointer points to the property name, but that wouldn't be interoperable.
Since there's no interoperable and accepted solution to this problem, I suggest we remove this test. That would mean the test suite would be ambiguous about whether property names can have annotations or how they are implemented if they do.
What do you think?
The text was updated successfully, but these errors were encountered: