Django REST framework (DRF) serializer speedup
Django REST framework serializers speedup
Simple short-living cache implementation for DRF serializers. It solves two issues with DRF serializers:
Serializer.fieldsproperty can be extremely slow in some cases. Such cases include complex
ModelSerializer, serializer hierarchies with repeated serializers and recursive serializers.
fieldsis cached and computed once per class for each serialization run. This leads to limitation -
fieldsshould not be dynamically adjusted inside of serializer, which inherits from
- By default serializers will re-compute object representation event if they encounter same object with same serializer twice. This can turn into an issue, if multiple object include same dependent hard-to-serialize objects. Cache avoid re-computing instance representation if it encounters same instances multiple times.
Cache life-time and invalidation
Cache is designed to be simple and non-error-prone.
Cache is created when top-level serializer
and is cleaned up once this method is finished. Thus there is no need
for timeouts or complex cache invalidation. No data is shared between
Thus, following cases will work fine:
user = User.objects.get(name='Lee') data_1 = UserSerializer(user).data # OrderedDict([('name', 'Lee')]) user.name = 'Bob' user.save() data_2 = UserSerializer(user).data # OrderedDict([('name', 'Bob')])
Usage should be pretty simple - inject
from drf_serializer_cache import SerializerCacheMixin from rest_framework import serializers class UserSerializer(SerializerCacheMixin, serializers.ModelSerializer): class Meta: model = User fields = ('id', 'name')
Too often cache cleanup
Cache lives in serializer hierarchy root, but it's life-time is defined
by nearest tree node, which inherits from
Ideal situation is when root inherits from
SerializerCacheMixin uses custom
which also inherits from
If you use custom list as root of serializer hierarchy - it's recommended
SerializerCacheMixin as it's base class.
Too many hierarchies
By default serializers build nice hierarchy by calling
bind on their
child serializers. This case can be broken for
if it uses some serializer without calling
bind on it:
from drf_serializer_cache import SerializerCacheMixin from rest_framework import serializers class ResultSerializer(SerializerCacheMixin, serializers.Serializer): results = serializers.SerializerMethodField() def get_results(self, instance): # recursive serializer serializer = self.__class__(instance.results, many=True) serializer.bind('*', self) # bind call is essential for efficient cache ! return serializer.data