Fix: Dataset infos() can be broken if a transform not redefining infos() is stacked on the top

CHANGELOG.md

@@ -19,6 +19,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### Bug fixes
 - Create cache dir under only writable filesystem
   (<https://github.com/openvinotoolkit/datumaro/pull/1088>)
+- Fix: Dataset infos() can be broken if a transform not redefining infos() is stacked on the top
+  (<https://github.com/openvinotoolkit/datumaro/pull/1101>)
 
 ## 07/07/2023 - Release 1.4.0rc1
 ### New features

src/datumaro/components/transformer.py

@@ -48,6 +48,9 @@ def __len__(self):
     def media_type(self):
         return self._extractor.media_type()
 
+    def infos(self):
+        return self._extractor.infos()
+
 
 class ItemTransform(Transform):
     def transform_item(self, item: DatasetItem) -> Optional[DatasetItem]:

tests/unit/components/test_dataset_storage.py

@@ -75,29 +75,35 @@ def test_item_transform(self, fxt_stream_extractor: MagicMock):
         n_calls = 1
 
         self._test_loop(fxt_stream_extractor, storage, n_calls)
+        assert fxt_stream_extractor.__iter__.call_count == 1
 
         # Stack transform 1 level
         storage.transform(Rename, regex="|item_|rename_|")
         self._test_loop(fxt_stream_extractor, storage, n_calls, id_pattern="rename_{idx}")
+        assert fxt_stream_extractor.__iter__.call_count == 2
 
         # Stack transform 2 level
         storage.transform(Rename, regex="|rename_|renameagain_|")
         self._test_loop(fxt_stream_extractor, storage, n_calls, id_pattern="renameagain_{idx}")
+        assert fxt_stream_extractor.__iter__.call_count == 3
 
     def test_subset_transform(self, fxt_stream_extractor: MagicMock):
         storage = StreamDatasetStorage(source=fxt_stream_extractor)
 
         self._test_subsets(fxt_stream_extractor, storage)
+        assert fxt_stream_extractor.__iter__.call_count == 1
 
         # Stack transform 1 level
         storage.transform(RandomSplit, splits=[("train", 0.5), ("val", 0.5)], seed=3003)
         self._test_subsets(fxt_stream_extractor, storage, expect={"train", "val"})
+        assert fxt_stream_extractor.__iter__.call_count == 2
 
         # Stack transform 2 level
         storage.transform(
             MapSubsets, mapping={"train": DEFAULT_SUBSET_NAME, "val": DEFAULT_SUBSET_NAME}
         )
         self._test_subsets(fxt_stream_extractor, storage)
+        assert fxt_stream_extractor.__iter__.call_count == 3
 
     def test_info_transform(self, fxt_stream_extractor: MagicMock, fxt_infos: DatasetInfo):
         storage = StreamDatasetStorage(source=fxt_stream_extractor)
@@ -108,6 +114,7 @@ def test_info_transform(self, fxt_stream_extractor: MagicMock, fxt_infos: Datase
         storage.transform(ProjectInfos, dst_infos=dst_infos)
 
         assert storage.infos().get("new") == "info"
+        assert fxt_stream_extractor.__iter__.call_count == 0
 
     def test_categories_transform(
         self, fxt_stream_extractor: MagicMock, fxt_categories: CategoriesInfo
@@ -122,3 +129,44 @@ def test_categories_transform(
         actual = set(cat.name for cat in storage.categories()[AnnotationType.label])
         expect = set(mapping.values())
         assert actual == expect
+
+        assert fxt_stream_extractor.__iter__.call_count == 0
+
+    def test_mixed_transform(
+        self,
+        fxt_stream_extractor: MagicMock,
+        fxt_infos: DatasetInfo,
+        fxt_categories: CategoriesInfo,
+    ):
+        n_calls = 1
+        storage = StreamDatasetStorage(source=fxt_stream_extractor)
+
+        # Check extractor infos
+        assert storage.infos() == fxt_infos
+
+        dst_infos = {"new": "info"}
+        storage.transform(ProjectInfos, dst_infos=dst_infos)
+        assert fxt_stream_extractor.__iter__.call_count == 0
+
+        # Check extractor categories
+        assert storage.categories() == fxt_categories
+
+        mapping = {"car": "apple", "cat": "banana", "dog": "cinnamon"}
+        storage.transform(RemapLabels, mapping=mapping)
+        assert fxt_stream_extractor.__iter__.call_count == 0
+
+        # Stack Rename (ItemTransform) on the top
+        storage.transform(Rename, regex="|item_|rename_|")
+        assert fxt_stream_extractor.__iter__.call_count == 0
+
+        # Check ProjectInfos
+        assert storage.infos().get("new") == "info"
+
+        # Check RemapLabels
+        actual = set(cat.name for cat in storage.categories()[AnnotationType.label])
+        expect = set(mapping.values())
+        assert actual == expect
+
+        # Check Rename
+        self._test_loop(fxt_stream_extractor, storage, n_calls, id_pattern="rename_{idx}")
+        assert fxt_stream_extractor.__iter__.call_count == n_calls

tests/unit/test_dataset.py

@@ -58,7 +58,7 @@
 from datumaro.components.merge.intersect_merge import IntersectMerge
 from datumaro.components.progress_reporting import NullProgressReporter, ProgressReporter
 from datumaro.components.transformer import ItemTransform, Transform
-from datumaro.plugins.transforms import ProjectInfos
+from datumaro.plugins.transforms import ProjectInfos, RemapLabels
 
 from ..requirements import Requirements, mark_requirement
 
@@ -2275,15 +2275,21 @@ def test_dataset_infos_intersect_merge(
         fxt_test_case.assertEqual(dataset.infos(), infos)
 
     @mark_requirement(Requirements.DATUM_GENERAL_REQ)
+    @pytest.mark.parametrize("is_eager", [True, False])
     def test_dataset_infos_transform(
-        self, fxt_test_case, fxt_sample_dataset_factory, fxt_sample_infos
+        self, fxt_test_case, fxt_sample_dataset_factory, fxt_sample_infos, is_eager
     ):
         infos_1, infos_2, infos = fxt_sample_infos
+        with eager_mode(is_eager):
+            dataset = fxt_sample_dataset_factory(infos=infos_1)
 
-        dataset = fxt_sample_dataset_factory(infos=infos_1)
+            dataset.transform(ProjectInfos, dst_infos=infos_2, overwrite=False)
+            fxt_test_case.assertEqual(dataset.infos(), infos)
 
-        dataset.transform(ProjectInfos, dst_infos=infos_2, overwrite=False)
-        fxt_test_case.assertEqual(dataset.infos(), infos)
+            dataset.transform(ProjectInfos, dst_infos=infos_2, overwrite=True)
+            fxt_test_case.assertEqual(dataset.infos(), infos_2)
 
-        dataset.transform(ProjectInfos, dst_infos=infos_2, overwrite=True)
-        fxt_test_case.assertEqual(dataset.infos(), infos_2)
+            dataset.transform(
+                RemapLabels, mapping={"car": "apple", "cat": "banana", "dog": "cinnamon"}
+            )
+            fxt_test_case.assertEqual(dataset.infos(), infos_2)