Performance improvement when rendering relationships with ModelSerializer (#461)

Author: timcbaoth

AUTHORS

@@ -15,4 +15,5 @@ Oliver Sauder <[email protected]>
 Raphael Cohen <[email protected]>
 Roberto Barreda <[email protected]>
 santiavenda <[email protected]>
+Tim Selman <[email protected]>
 Yaniv Peer <[email protected]>

CHANGELOG.md

@@ -7,6 +7,8 @@
 * Add optional [jsonapi-style](http://jsonapi.org/format/) sort filter backend. See [usage docs](docs/usage.md#filter-backends)
 * For naming consistency, renamed new `JsonApi`-prefix pagination classes to `JSONAPI`-prefix.
   * Deprecates `JsonApiPageNumberPagination` and `JsonApiLimitOffsetPagination`
+* Performance improvement when rendering relationships with `ModelSerializer`
+
 
 v2.5.0 - Released July 11, 2018
 

example/tests/test_serializers.py

@@ -2,24 +2,39 @@
 from django.test import TestCase
 from django.urls import reverse
 from django.utils import timezone
-
-from rest_framework_json_api.serializers import ResourceIdentifierObjectSerializer
+from rest_framework.request import Request
+from rest_framework.test import APIRequestFactory
+
+from rest_framework_json_api.serializers import (
+    DateField,
+    ModelSerializer,
+    ResourceIdentifierObjectSerializer
+)
 from rest_framework_json_api.utils import format_resource_type
 
 from example.models import Author, Blog, Entry
+from example.serializers import BlogSerializer
+
+try:
+    from unittest import mock
+except ImportError:
+    import mock
 
+request_factory = APIRequestFactory()
 pytestmark = pytest.mark.django_db
 
 
 class TestResourceIdentifierObjectSerializer(TestCase):
     def setUp(self):
         self.blog = Blog.objects.create(name='Some Blog', tagline="It's a blog")
+        now = timezone.now()
+
         self.entry = Entry.objects.create(
             blog=self.blog,
             headline='headline',
             body_text='body_text',
-            pub_date=timezone.now(),
-            mod_date=timezone.now(),
+            pub_date=now.date(),
+            mod_date=now.date(),
             n_comments=0,
             n_pingbacks=0,
             rating=3
@@ -30,6 +45,59 @@ def setUp(self):
                 Author.objects.create(name=name, email='{}@example.org'.format(name))
             )
 
+    def test_forward_relationship_not_loaded_when_not_included(self):
+        to_representation_method = 'example.serializers.BlogSerializer.to_representation'
+        with mock.patch(to_representation_method) as mocked_serializer:
+            class EntrySerializer(ModelSerializer):
+                blog = BlogSerializer()
+
+                class Meta:
+                    model = Entry
+                    fields = '__all__'
+
+            request_without_includes = Request(request_factory.get('/'))
+            serializer = EntrySerializer(context={'request': request_without_includes})
+            serializer.to_representation(self.entry)
+
+            mocked_serializer.assert_not_called()
+
+    def test_forward_relationship_optimization_correct_representation(self):
+        class EntrySerializer(ModelSerializer):
+            blog = BlogSerializer()
+
+            class Meta:
+                model = Entry
+                fields = '__all__'
+
+        request_without_includes = Request(request_factory.get('/'))
+        serializer = EntrySerializer(context={'request': request_without_includes})
+        result = serializer.to_representation(self.entry)
+
+        # Remove non deterministic fields
+        result.pop('created_at')
+        result.pop('modified_at')
+
+        expected = dict(
+            [
+                ('id', 1),
+                ('blog', dict([('type', 'blogs'), ('id', 1)])),
+                ('headline', 'headline'),
+                ('body_text', 'body_text'),
+                ('pub_date', DateField().to_representation(self.entry.pub_date)),
+                ('mod_date', DateField().to_representation(self.entry.mod_date)),
+                ('n_comments', 0),
+                ('n_pingbacks', 0),
+                ('rating', 3),
+                ('authors',
+                    [
+                        dict([('type', 'authors'), ('id', '1')]),
+                        dict([('type', 'authors'), ('id', '2')]),
+                        dict([('type', 'authors'), ('id', '3')]),
+                        dict([('type', 'authors'), ('id', '4')]),
+                        dict([('type', 'authors'), ('id', '5')])])])
+
+        self.assertDictEqual(expected, result)
+
     def test_data_in_correct_format_when_instantiated_with_blog_object(self):
         serializer = ResourceIdentifierObjectSerializer(instance=self.blog)
 

rest_framework_json_api/serializers.py

@@ -182,35 +182,41 @@ def to_representation(self, instance):
 
         for field in readable_fields:
             try:
-
-                if isinstance(field, ModelSerializer) and hasattr(field, field.source + "_id"):
-                    attribute = getattr(instance, field.source + "_id")
-                    if attribute is None:
-                        ret[field.field_name] = None
-                        continue
-                    resource_type = get_resource_type_from_instance(field)
-                    if resource_type:
-                        ret[field.field_name] = OrderedDict([("type", resource_type),
-                                                             ("id", attribute)])
-                        continue
-
-                attribute = field.get_attribute(instance)
+                field_representation = self._get_field_representation(field, instance)
+                ret[field.field_name] = field_representation
             except SkipField:
                 continue
 
-            # We skip `to_representation` for `None` values so that fields do
-            # not have to explicitly deal with that case.
-            #
-            # For related fields with `use_pk_only_optimization` we need to
-            # resolve the pk value.
-            check_for_none = attribute.pk if isinstance(attribute, PKOnlyObject) else attribute
-            if check_for_none is None:
-                ret[field.field_name] = None
-            else:
-                ret[field.field_name] = field.to_representation(attribute)
-
         return ret
 
+    def _get_field_representation(self, field, instance):
+        request = self.context.get('request')
+        is_included = field.source in get_included_resources(request)
+        if not is_included and \
+                isinstance(field, ModelSerializer) and \
+                hasattr(instance, field.source + '_id'):
+            attribute = getattr(instance, field.source + '_id')
+
+            if attribute is None:
+                return None
+
+            resource_type = get_resource_type_from_serializer(field)
+            if resource_type:
+                return OrderedDict([('type', resource_type), ('id', attribute)])
+
+        attribute = field.get_attribute(instance)
+
+        # We skip `to_representation` for `None` values so that fields do
+        # not have to explicitly deal with that case.
+        #
+        # For related fields with `use_pk_only_optimization` we need to
+        # resolve the pk value.
+        check_for_none = attribute.pk if isinstance(attribute, PKOnlyObject) else attribute
+        if check_for_none is None:
+            return None
+        else:
+            return field.to_representation(attribute)
+
 
 class PolymorphicSerializerMetaclass(SerializerMetaclass):
     """