Significantly reduced the number of queries to DB from api/jobs, api/tasks,...

Significantly reduced the number of queries to DB from api/jobs, api/tasks, and api/projects endpoints. (#5304)

CHANGELOG.md

@@ -34,6 +34,7 @@ from online detectors & interactors) (<https://github.com/opencv/cvat/pull/4543>
 - The `--https` option of CLI (<https://github.com/opencv/cvat/pull/4910>)
 
 ### Fixed
+- Significantly optimized access to DB for api/jobs, api/tasks, and api/projects.
 - Removed a possibly duplicated encodeURI() calls in `server-proxy.ts` to prevent doubly encoding
 non-ascii paths while adding files from "Connected file share" (issue #4428)
 - Removed unnecessary volumes defined in docker-compose.serverless.yml

cvat/apps/engine/models.py

@@ -364,6 +364,10 @@ class Task(models.Model):
     class Meta:
         default_permissions = ()
 
+    def get_labels(self):
+        project = self.project
+        return project.label_set if project else self.label_set
+
     def get_dirname(self):
         return os.path.join(settings.TASKS_ROOT, str(self.id))
 

cvat/apps/engine/serializers.py

@@ -454,7 +454,7 @@ class StorageSerializer(serializers.ModelSerializer):
         fields = ('id', 'location', 'cloud_storage_id')
 
 class TaskReadSerializer(serializers.ModelSerializer):
-    labels = LabelSerializer(many=True, source='label_set', partial=True, required=False)
+    labels = LabelSerializer(many=True, source='get_labels')
     segments = SegmentSerializer(many=True, source='segment_set', read_only=True)
     data_chunk_size = serializers.ReadOnlyField(source='data.chunk_size', required=False)
     data_compressed_chunk_type = serializers.ReadOnlyField(source='data.compressed_chunk_type', required=False)
@@ -483,11 +483,6 @@ class TaskReadSerializer(serializers.ModelSerializer):
             'overlap': { 'allow_null': True },
         }
 
-    def to_representation(self, instance):
-        response = super().to_representation(instance)
-        if instance.project_id:
-            response["labels"] = LabelSerializer(many=True).to_representation(instance.project.label_set)
-        return response
 
 class TaskWriteSerializer(WriteOnceMixin, serializers.ModelSerializer):
     labels = LabelSerializer(many=True, source='label_set', partial=True, required=False)

cvat/apps/engine/views.py

@@ -308,9 +308,10 @@ class ProjectViewSet(viewsets.GenericViewSet, mixins.ListModelMixin,
     mixins.RetrieveModelMixin, CreateModelMixin, DestroyModelMixin,
     PartialUpdateModelMixin, UploadMixin, AnnotationMixin, SerializeMixin
 ):
-    queryset = models.Project.objects.prefetch_related(Prefetch('label_set',
-        queryset=models.Label.objects.order_by('id')
-    ))
+    queryset = models.Project.objects.select_related('assignee', 'owner',
+        'target_storage', 'source_storage').prefetch_related(
+        'tasks', 'label_set__sublabels__attributespec_set',
+        'label_set__attributespec_set')
 
     # NOTE: The search_fields attribute should be a list of names of text
     # type fields on the model,such as CharField or TextField
@@ -752,10 +753,12 @@ class TaskViewSet(viewsets.GenericViewSet, mixins.ListModelMixin,
     mixins.RetrieveModelMixin, CreateModelMixin, DestroyModelMixin,
     PartialUpdateModelMixin, UploadMixin, AnnotationMixin, SerializeMixin
 ):
-    queryset = Task.objects.prefetch_related(
-            Prefetch('label_set', queryset=models.Label.objects.order_by('id')),
-            "label_set__attributespec_set",
-            "segment_set__job_set")
+    queryset = Task.objects.all().select_related('data', 'assignee', 'owner',
+        'target_storage', 'source_storage').prefetch_related(
+        'segment_set__job_set__assignee', 'label_set__attributespec_set',
+        'project__label_set__attributespec_set',
+        'label_set__sublabels__attributespec_set',
+        'project__label_set__sublabels__attributespec_set')
     lookup_fields = {'project_name': 'project__name', 'owner': 'owner__username', 'assignee': 'assignee__username'}
     search_fields = ('project_name', 'name', 'owner', 'status', 'assignee', 'subset', 'mode', 'dimension')
     filter_fields = list(search_fields) + ['id', 'project_id', 'updated_date']
@@ -1318,10 +1321,16 @@ class TaskViewSet(viewsets.GenericViewSet, mixins.ListModelMixin,
             '200': JobReadSerializer, # check JobWriteSerializer.to_representation
         })
 )
+
 class JobViewSet(viewsets.GenericViewSet, mixins.ListModelMixin,
     mixins.RetrieveModelMixin, PartialUpdateModelMixin, UploadMixin, AnnotationMixin
 ):
-    queryset = Job.objects.all()
+    queryset = Job.objects.all().select_related('segment__task__data').prefetch_related(
+        'segment__task__label_set', 'segment__task__project__label_set',
+        'segment__task__label_set__sublabels__attributespec_set',
+        'segment__task__project__label_set__sublabels__attributespec_set',
+        'segment__task__label_set__attributespec_set',
+        'segment__task__project__label_set__attributespec_set')
     iam_organization_field = 'segment__task__organization'
     search_fields = ('task_name', 'project_name', 'assignee', 'state', 'stage')
     filter_fields = list(search_fields) + ['id', 'task_id', 'project_id', 'updated_date']