Tried to optimize annotation fetching

cvat/apps/dataset_manager/task.py

@@ -12,7 +12,7 @@
 from django.db.models.query import Prefetch
 from django.utils import timezone
 
-from cvat.apps.engine import models, serializers
+from cvat.apps.engine import models
 from cvat.apps.engine.plugins import plugin_decorator
 from cvat.apps.profiler import silk_profile
 
@@ -42,6 +42,13 @@ def values(cls):
     def __str__(self):
         return self.value
 
+def _remove_none(dictionary):
+    keys = list(dictionary.keys())
+    for field in keys:
+        if dictionary[field] is None:
+            del dictionary[field]
+    return dictionary
+
 def _merge_table_rows(rows, keys_for_merge, field_id):
     # It is necessary to keep a stable order of original rows
     # (e.g. for tracked boxes). Otherwise prev_box.frame can be bigger
@@ -313,6 +320,7 @@ def _create(self, data):
             self._set_updated_date()
             self.db_job.save()
 
+
     def create(self, data):
         self._create(data)
 
@@ -366,6 +374,7 @@ def _extend_attributes(attributeval_set, default_attribute_values):
                 attributeval_set.append(dotdict([
                     ('spec_id', db_attr.spec_id),
                     ('value', db_attr.value),
+                    ('id', None),
                 ]))
 
     def _init_tags_from_db(self):
@@ -399,8 +408,15 @@ def _init_tags_from_db(self):
             self._extend_attributes(db_tag.labeledimageattributeval_set,
                 self.db_attributes[db_tag.label_id]["all"].values())
 
-        serializer = serializers.LabeledImageSerializer(db_tags, many=True)
-        self.ir_data.tags = serializer.data
+        def convert_tag(tag):
+            tag['attributes'] = tag['labeledshapeattributeval_set']
+            del tag['labeledshapeattributeval_set']
+
+            _remove_none(tag)
+            for attr in tag['attributes']:
+                del attr['id']
+
+        self.ir_data.tags = [convert_tag(value) for value in db_tags]
 
     def _init_shapes_from_db(self):
         db_shapes = self.db_job.labeledshape_set.prefetch_related(
@@ -453,8 +469,20 @@ def _init_shapes_from_db(self):
         for shape_id, shape_elements in elements.items():
             shapes[shape_id].elements = shape_elements
 
-        serializer = serializers.LabeledShapeSerializer(list(shapes.values()), many=True)
-        self.ir_data.shapes = serializer.data
+        def convert_shape(shape):
+            shape['attributes'] = shape['labeledshapeattributeval_set']
+            del shape['labeledshapeattributeval_set']
+            _remove_none(shape)
+            for attr in shape['attributes']:
+                del attr['id']
+
+            if 'elements' in shape:
+                for element in shape['elements']:
+                    convert_shape(element)
+
+            return shape
+
+        self.ir_data.shapes = [convert_shape(value) for value in shapes.values()]
 
     def _init_tracks_from_db(self):
         db_tracks = self.db_job.labeledtrack_set.prefetch_related(
@@ -512,6 +540,7 @@ def _init_tracks_from_db(self):
         tracks = {}
         elements = {}
         for db_track in db_tracks:
+            db_track['elements'] = []
             db_track["trackedshape_set"] = _merge_table_rows(db_track["trackedshape_set"], {
                 'trackedshapeattributeval_set': [
                     'trackedshapeattributeval__value',
@@ -546,8 +575,28 @@ def _init_tracks_from_db(self):
         for track_id, track_elements in elements.items():
             tracks[track_id].elements = track_elements
 
-        serializer = serializers.LabeledTrackSerializer(list(tracks.values()), many=True)
-        self.ir_data.tracks = serializer.data
+        def convert_track(track):
+            track['shapes'] = track['trackedshape_set']
+            del track['trackedshape_set']
+            track['attributes'] = track['labeledtrackattributeval_set']
+            del track['labeledtrackattributeval_set']
+            _remove_none(track)
+            for attr in track['attributes']:
+                del attr['id']
+
+            for shape in track['shapes']:
+                _remove_none(shape)
+                shape['attributes'] = shape['trackedshapeattributeval_set']
+                for attr in shape['attributes']:
+                    del attr['id']
+
+            if 'elements' in track:
+                for element in track['elements']:
+                    convert_track(element)
+
+            return track
+
+        self.ir_data.tracks = [convert_track(value) for value in tracks.values()]
 
     def _init_version_from_db(self):
         self.ir_data.version = 0 # FIXME: should be removed in the future

cvat/apps/engine/backup.py

@@ -386,8 +386,7 @@ def serialize_annotations():
             for db_job_id in db_job_ids:
                 annotations = dm.task.get_job_data(db_job_id)
                 annotations_serializer = LabeledDataSerializer(data=annotations)
-                annotations_serializer.is_valid(raise_exception=True)
-                job_annotations.append(self._prepare_annotations(annotations_serializer.data, self._label_mapping))
+                job_annotations.append(self._prepare_annotations(annotations_serializer.initial_data, self._label_mapping))
 
             return job_annotations
 

cvat/apps/engine/mixins.py

@@ -279,8 +279,7 @@ def export_annotations(self, request, db_obj, export_func, callback, get_data=No
 
         data = get_data(self._object.pk)
         serializer = LabeledDataSerializer(data=data)
-        if serializer.is_valid(raise_exception=True):
-            return Response(serializer.data)
+        return Response(serializer.initial_data)
 
     def import_annotations(self, request, db_obj, import_func, rq_func, rq_id):
         is_tus_request = request.headers.get('Upload-Length', None) is not None or \