Fixes #35988: Add abstract, source, and deprecation fields to native ontology classes

docs/decisions/2026-10-21-ontology-definition-status-is-derived-from-releases.md

@@ -1,7 +1,7 @@
 # Ontology definition status is derived from releases
 
 - **Status:** Accepted
-- **Revisions:** v1 2026-10-21 (initial); v2 2026-10-23 (clarify publication response and working-change counts)
+- **Revisions:** v1 2026-10-21 (initial); v2 2026-10-23 (clarify publication response and working-change counts); v3 2026-10-25 (separate deprecation counts at stack integration)
 - **Deciders:** Harsha
 - **Guard:** `OntologyDefinitionStatusResolverTest`, `OntologyDerivedStatusStorageTest`, `IndependentOntologyStatusIT`
 - **Related:** ADR:2026-10-14-ontology-reviews-bind-workflow-tasks-to-revisions, ADR:2026-10-14-ontology-releases-seal-canonical-sql-members, #35981
@@ -59,3 +59,18 @@ reads for the governed-content comparison.
 
 Clarify the existing response names and stage vocabulary, the complete working-change count, and the
 writes that affect derived status. This amendment changes no API or stored data.
+
+## Amendment v3, 2026-10-25
+
+Release change summaries partition the default-base diff rows into added, changed, removed and
+deprecated counts. Newly deprecating an existing concept or property counts only as deprecated,
+even when its other governed fields also change. Adding an already deprecated definition counts
+as added. Editing a definition that was already deprecated, or clearing its flag, counts as
+changed. Removing a deprecated definition counts as removed.
+
+New captures always store all four counts, including zero. The schema leaves `deprecated` optional
+with no default. A historical summary without this bucket has an unavailable total; readers must
+not infer zero. Existing captures are immutable and receive no backfill.
+
+`OntologyReleaseComparisonTest.summaryPartitionsDeprecationsAndOtherChangesWithoutDoubleCounting`
+and `unchangedAndNewClosuresHaveAnExplicitZeroDeprecationCount` guard the partition and explicit zero.

docs/decisions/2026-10-22-native-classes-record-abstraction-source-and-deprecation.md

@@ -0,0 +1,52 @@
+# Native classes record abstraction, source, and deprecation as standard RDF annotations
+
+- **Status:** Accepted
+- **Revisions:** v1 2026-10-22 (initial)
+- **Deciders:** Harsha
+- **Guard:** `OntologyReleaseProjectionTest`, `NativeOntologyRdfImporterTest`, `OntologyStructureValidatorTest`, `OntologyReleaseComparisonTest`, `IndependentOntologyClassModelIT`, `NativeOntologyMemoryIT`
+- **Related:** #35988, ADR:2026-10-14-ontology-definitions-have-independent-entity-storage
+
+## Context
+
+Ontology Studio authors classes that have no direct instances, cite the standard their definition
+comes from, and retire in favor of a successor. Native concepts and properties had no fields for
+these facts, so a release export could not state them, and a release diff showed a retirement as
+an ordinary change. The Studio also lists a new class as one change row with its parents and
+property count, which the flat working diff cannot supply.
+
+## Decision
+
+- Concepts carry `abstract`, `definitionSource`, `deprecated`, and `replacedBy`. Properties carry
+  `deprecated` and `replacedBy`. `abstract` and `deprecated` default to false. The class layer stays
+  on the ontology.
+- `replacedBy` names another definition of the same kind in the same ontology, never the definition
+  itself. The reference index records it with the role `REPLACED_BY`, so deleting the replacement is
+  refused like any other dangling reference.
+- Releases project the fields as `om:abstract true`, `dct:source <iri>`, `owl:deprecated true`, and
+  `dct:isReplacedBy <iri>`, where `om:` is `https://open-metadata.org/ontology/`. OWL has no
+  abstract-class construct, so `om:abstract` is an annotation property. False and absent values
+  project nothing. Native import reads the same four predicates and requires `xsd:boolean` values
+  and IRIs.
+- A data enumeration needs at least two allowed values. An empty list means no enumeration. The
+  schema states this as `not: {minItems: 1, maxItems: 1}`, because `minItems: 2` would reject the
+  empty default. `OntologyPropertyRepository` enforces it on every write, with the message
+  `An enumeration needs at least two allowed values`. A one-value enumeration stored before this
+  rule stays editable until its values change.
+- A member whose `deprecated` turned true diffs as `DEPRECATED` in release and working diffs.
+  Clearing the flag, or editing a deprecated member, is `CHANGED`.
+- `GET /ontologies/{id}/working/diff?groupBy=concept` returns `groups` and an empty `data`. A group
+  names a concept, its own change type or `CHANGED` when only its axioms or properties changed, the
+  labels of its named `SUBCLASS_OF` parents, the count of properties whose domain includes it, and
+  its member changes. An axiom joins its subject. A property joins the first concept in its domain
+  that this ontology declares. A change that belongs to no concept, such as the ontology root or an
+  axiom about an external subject, forms a group without a concept, so the groups cover every
+  working change. Groups sort by label and then IRI, and concept-less groups come last. Paging
+  counts groups. The cursor is bound to the working and activation revisions like the plain cursor,
+  and uses its own prefix, so neither cursor is accepted by the other view.
+
+## Consequences
+
+Other OWL tools can read the exported deprecation and source. Releases captured before this change
+compare unchanged, because both sides of a comparison read the new fields with their defaults.
+Clients that read grouped changes must handle a group without a concept. Grouping reads the same
+working content and active release as the plain diff and adds no queries.

openmetadata-integration-tests/src/test/java/org/openmetadata/it/tests/BaseOntologyApprovalIT.java

@@ -11,6 +11,7 @@
 import org.openmetadata.it.bootstrap.SharedEntities;
 import org.openmetadata.it.util.SdkClients;
 import org.openmetadata.it.util.TestNamespace;
+import org.openmetadata.schema.api.data.ActivateOntologyRelease;
 import org.openmetadata.schema.api.data.CaptureOntologyRelease;
 import org.openmetadata.schema.api.data.CreateOntology;
 import org.openmetadata.schema.api.data.CreateOntologyConcept;
@@ -24,6 +25,7 @@
 import org.openmetadata.schema.type.TaskCategory;
 import org.openmetadata.schema.type.TaskEntityType;
 import org.openmetadata.schema.type.TaskResolutionType;
+import org.openmetadata.service.Entity;
 
 abstract class BaseOntologyApprovalIT {
   @AfterEach
@@ -54,6 +56,22 @@ protected static OntologyRelease capture(Ontology ontology) {
             .withExpectedWorkingRevision(service.state(ontology.getId()).getWorkingRevision()));
   }
 
+  protected static OntologyRelease activate(TestNamespace ns, Ontology ontology) {
+    final OntologyRelease release = capture(ontology);
+    final var task = submit(ns, Entity.ONTOLOGY_RELEASE, release.getFullyQualifiedName());
+    SdkClients.user1Client()
+        .tasks()
+        .resolve(task.getId().toString(), decision(TaskResolutionType.Approved));
+    SdkClients.adminClient()
+        .ontologies()
+        .activate(
+            ontology.getId(),
+            new ActivateOntologyRelease()
+                .withReleaseId(release.getId())
+                .withExpectedActivationRevision(0L));
+    return release;
+  }
+
   protected static Ontology createOntology(TestNamespace ns, String name) {
     Ontology ontology =
         SdkClients.adminClient().ontologies().create(request(ns).withName(ns.prefix(name)));