object-model-conformance

object-model-conformance

Shared metadata (meta.json) + the scenarios every port's runtime object model must satisfy. Each port loads meta.json and runs these against its MetaObject / MetaField / ValueObject / ObjectClassRegistry + newInstance factory. Assertions are behavioral (type-kind, back-ref identity, field values, list contents, overflow) — not byte-identity. Resolve the corpus dir the same way the port's existing conformance runners resolve sibling corpora (repo-root walk).

Scenarios (Person = object.value with name:string, age:int, home:Address, tags:Tag[])

1. instantiate-value — newInstance(Person) returns the port's map-backed ValueObject (the unbound default for object.value); the instance's MetaObject back-reference is the Person MetaObject.
2. scalar-round-trip — set name="Ada", age=36; get name == "Ada", age == 36 (integer).
3. nested-object — newInstance(Address), set street="1 Main", city="Anytown"; set as Person.home; get Person.home → an Address-backed object with street/city as set; its back-reference is the Address MetaObject.
4. array-of-objects — build [Tag(label="a"), Tag(label="b")]; set Person.tags; get Person.tags → ordered list of 2; labels "a","b"; each element's back-reference is the Tag MetaObject.
5. overflow — on a ValueObject Person, set key "nickname" (not a declared field); get "nickname" round-trips (an instance may hold more than the declared field set).
6. bound-type — register a port-local test type (an "aware" type with name/age/home/tags) for FQN "com::example::om::Person" in the ObjectClassRegistry; newInstance(Person) returns THAT type (not a ValueObject); its back-reference is set; scalar/nested/array get/set behave identically to a ValueObject. Use a fresh/scoped registry so it does not leak into other scenarios.
7. no-binding-fallback — with nothing registered for Address, newInstance(Address) returns a ValueObject.