convert: Brainstorm-E split migrators (treatment→manipulation, ontology_table_row→observations)

.github/workflows/test-code.yml

@@ -35,16 +35,33 @@ jobs:
       - name: Check out repository
         uses: actions/checkout@v4
 
-      - name: Check out did-schema (sibling, for V_delta validation)
+      - name: Check out did-schema (sibling, for V_epsilon validation)
         uses: actions/checkout@v4
         with:
           repository: Waltham-Data-Science/did-schema
-          ref: main
+          # Track the matching did-schema E branch: the Brainstorm-E classes
+          # (observation property classes, shape library, session_relative_
+          # reference, subject_statement, etc.) live on this branch and are
+          # not yet on did-schema main. Revert to `main` once DID-schema#62
+          # merges. Falls back to main off-PR (e.g. direct pushes to V2).
+          ref: ${{ github.head_ref || 'main' }}
           path: did-schema
 
+      - name: Assemble V_epsilon schema set (stable + draft)
+        # V_epsilon replaces V_delta as the migration/validation target.
+        # The schema cache loads a single flat dir of *.json, but the
+        # Brainstorm-E observation/manipulation classes live in
+        # V_epsilon/draft while base/element/etc. live in V_epsilon/stable
+        # (class names are disjoint across tiers), so combine both tiers
+        # into one directory for DID_SCHEMA_PATH.
+        run: |
+          mkdir -p "${GITHUB_WORKSPACE}/epsilon-schemas"
+          cp did-schema/schemas/V_epsilon/stable/*.json "${GITHUB_WORKSPACE}/epsilon-schemas/"
+          cp did-schema/schemas/V_epsilon/draft/*.json  "${GITHUB_WORKSPACE}/epsilon-schemas/"
+
       - name: Export DID_SCHEMA_PATH
         run: |
-          echo "DID_SCHEMA_PATH=${GITHUB_WORKSPACE}/did-schema/schemas/V_delta/stable" \
+          echo "DID_SCHEMA_PATH=${GITHUB_WORKSPACE}/epsilon-schemas" \
             >> "$GITHUB_ENV"
 
       - name: Set up MATLAB

src/did/+did/database.m

@@ -1211,6 +1211,19 @@ function validate_doc_vs_schema(database_obj, docProps, schemaStruct, all_ids)
             for i = 1 : numel(superFullNames)
                 [~,superNames{i}] = fileparts(superFullNames{i}); % keep compatibility with Matlab 2019a
             end
+            % did2-form documents carry superclass names directly in
+            % `.class_name` (e.g. {class_name:'base'}) and have no
+            % `.definition` path. Fall back to those names so the legacy
+            % validator recognises did2-shaped documents during the
+            % v1->v2 transition (otherwise superNames is empty and every
+            % such doc fails the superclasses check spuriously).
+            if isempty(superNames)
+                try
+                    superNames = {classProps.superclasses.class_name};
+                catch
+                    superNames = {};
+                end
+            end
             if ~iscell(superNames), superNames = {superNames}; end
             superNames = unique(superNames);
             schemaFields = fieldnames(schemaStruct);
@@ -1238,8 +1251,13 @@ function validate_doc_vs_schema(database_obj, docProps, schemaStruct, all_ids)
                         assert(areSame,'DID:Database:ValidationSuperClasses', ...
                             'Dissimilar superclasses defined/found for %s ("%s" <=> "%s")', ...
                             doc_name, expectedStr, superNamesStr);
-                        % Recursively validate all superNames against this doc:
-                        for idx = 1 : numel(superNames)
+                        % Recursively validate all superclasses against this
+                        % doc. Bound by superFullNames (the `.definition`
+                        % paths): did2-form docs expose names but no paths, so
+                        % superFullNames is empty and the deeper recursion is
+                        % skipped -- the concrete class is still validated, and
+                        % the superclasses-name check above already passed.
+                        for idx = 1 : numel(superFullNames)
                             % First get the superClass' definition struct
                             defStruct = database_obj.get_document_schema(superFullNames{idx});
                             % Extract validation file from definition