[SDK-503] Prevent users from uploading video annotations over the API limit (5000). (#1347)

Author: sfendell-labelbox

labelbox/schema/annotation_import.py

@@ -23,6 +23,7 @@
     from labelbox.types import Label
 
 NDJSON_MIME_TYPE = "application/x-ndjson"
+LABEL_LIMIT = 5000
 logger = logging.getLogger(__name__)
 
 
@@ -153,6 +154,9 @@ def _get_ndjson_from_objects(cls, objects: Union[List[Dict[str, Any]],
             )
 
         objects = serialize_labels(objects)
+        if len(objects) > LABEL_LIMIT:
+            raise ValueError(
+                f"Label count {len(objects)} exceeds limit of {(LABEL_LIMIT)}")
         cls._validate_data_rows(objects)
 
         data_str = parser.dumps(objects)
@@ -502,8 +506,8 @@ def create_from_objects(
         Returns:
             MALPredictionImport
         """
-        data = cls._get_ndjson_from_objects(predictions, 'annotations')
 
+        data = cls._get_ndjson_from_objects(predictions, 'annotations')
         if len(predictions) > 0 and isinstance(predictions[0], Dict):
             predictions_dicts = cast(List[Dict[str, Any]], predictions)
             has_confidence = LabelsConfidencePresenceChecker.check(

tests/integration/annotation_import/test_bulk_import_request.py

@@ -370,3 +370,69 @@ def test_nested_video_object_annotations(client,
     import_annotations.wait_until_done()
 
     assert import_annotations.errors == []
+
+
+def test_below_annotation_limit_on_single_data_row(
+        client, configured_project_with_one_data_row, video_data, rand_gen):
+    _, data_row_uids = video_data
+
+    def create_label():
+        return Label(data=VideoData(uid=data_row_uids[0]),
+                     annotations=[
+                         VideoObjectAnnotation(name="bbox",
+                                               keyframe=True,
+                                               frame=4,
+                                               segment_index=0,
+                                               value=Rectangle(
+                                                   start=Point(x=100, y=100),
+                                                   end=Point(x=105, y=105),
+                                               ))
+                     ])
+
+    configured_project_with_one_data_row.update(media_type=MediaType.Video)
+    configured_project_with_one_data_row.create_batch(
+        rand_gen(str),
+        data_row_uids,  # sample of data row objects
+        5  # priority between 1(Highest) - 5(lowest)
+    )
+    labels = [create_label() for _ in range(5000)]
+    import_annotations = MALPredictionImport.create_from_objects(
+        client=client,
+        project_id=configured_project_with_one_data_row.uid,
+        name=f"import {str(uuid.uuid4())}",
+        predictions=labels)
+    import_annotations.wait_until_done()
+
+    assert import_annotations.errors == []
+
+
+def test_above_annotation_limit_on_single_data_row(
+        client, configured_project_with_one_data_row, video_data, rand_gen):
+    _, data_row_uids = video_data
+
+    def create_label():
+        return Label(data=VideoData(uid=data_row_uids[0]),
+                     annotations=[
+                         VideoObjectAnnotation(name="bbox",
+                                               keyframe=True,
+                                               frame=4,
+                                               segment_index=0,
+                                               value=Rectangle(
+                                                   start=Point(x=100, y=100),
+                                                   end=Point(x=105, y=105),
+                                               ))
+                     ])
+
+    configured_project_with_one_data_row.update(media_type=MediaType.Video)
+    configured_project_with_one_data_row.create_batch(
+        rand_gen(str),
+        data_row_uids,  # sample of data row objects
+        5  # priority between 1(Highest) - 5(lowest)
+    )
+    labels = [create_label() for _ in range(5001)]
+    with pytest.raises(ValueError):
+        MALPredictionImport.create_from_objects(
+            client=client,
+            project_id=configured_project_with_one_data_row.uid,
+            name=f"import {str(uuid.uuid4())}",
+            predictions=labels)