fixup! [r] Support for AnVIL duos_id (#6620)

src/azul/plugins/metadata/anvil/__init__.py

@@ -288,8 +288,10 @@ def manifest_config(self) -> ManifestConfig:
         # the fields listed here and those used in `self._field_mapping`.
         fields_to_omit_from_manifest = [
             ('contents', 'activities', 'activity_table'),
-            # duos_id is omitted from manifests since the field isn't included
-            # in the `files` index.
+            # We omit the `duos_id` field from manifests since there is only one
+            # DUOS bundle per dataset, and that bundle only contributes to outer
+            # entities of the `datasets` type, not to entities of the other
+            # types, such as files, which the manifest is generated from.
             ('contents', 'datasets', 'duos_id'),
             ('contents', 'files', 'uuid'),
             ('contents', 'files', 'version'),

src/azul/plugins/repository/tdr_anvil/__init__.py

@@ -136,10 +136,10 @@ class BundleType(Enum):
     request per *bundle* instead, potentially overloading the DUOS service. Our
     solution is to retrieve `description` only in a bundle of this dedicated
     DUOS type, once per snapshot, and merge it with the other dataset fields
-    during aggregation. As a result of this implemtation, the DUOS bundle is
-    only added to the `datasets` and `bundles` indices, making the `duos_id`
-    and `description` fields unavailable for file manifests as they pull values
-    from the `files` index.
+    during aggregation. As a result, `duos_id` cannot be included in file
+    manifests since there is only one DUOS bundle per dataset, and that bundle
+    only contributes to outer entities of the `datasets` type, not to entities
+    of the other types, such as files, which the manifest is generated from.
 
     All other bundles are replica bundles. Replica bundles consist of a batch of
     rows from an arbitrary BigQuery table, which may or may not be described by

test/service/test_manifest.py

@@ -1760,6 +1760,10 @@ def bundles(cls) -> list[SourcedBundleFQID]:
     def test_compact_manifest(self):
         response = self._get_manifest(ManifestFormat.compact, filters={})
         self.assertEqual(200, response.status_code)
+        # The `duos_id` field is absent from manifests since there is only one
+        # DUOS bundle per dataset, and that bundle only contributes to outer
+        # entities of the `datasets` type, not to entities of the other types,
+        # such as files, which the manifest is generated from.
         expected = [
             (
                 'bundles.bundle_uuid',