How do you filter a nested serializer in Django Rest Framework?

While all the above answers work, I find the use of Django's Prefetch object the easiest way of all.

Say a Restaurant obj has a lot of MenuItems, some of which are is_removed == True, and you only want those that are not removed.

In RestaurantViewSet, do something like

from django.db.models import Prefetch

queryset = Restaurant.objects.prefetch_related(
    Prefetch('menu_items', queryset=MenuItem.objects.filter(is_removed=False), to_attr='filtered_menu_items')
)

In RestaurantSerializer, do something like

class RestaurantSerializer(serializers.ModelSerializer):
    menu_items = MenuItemSerializer(source='filtered_menu_items', many=True, read_only=True)

Tested many solutions from SO and other places.

Found only one working solution for Django 2.0 + DRF 3.7.7.

Define a method in model which has nested class. Craft a filter that will fit your needs.

class Channel(models.Model):
    name = models.CharField(max_length=40)
    number = models.IntegerField(unique=True)
    active = models.BooleanField(default=True)

    def current_epg(self):
        return Epg.objects.filter(channel=self, end__gt=datetime.now()).order_by("end")[:6]


class Epg(models.Model):
    start = models.DateTimeField()
    end = models.DateTimeField(db_index=True)
    title = models.CharField(max_length=300)
    description = models.CharField(max_length=800)
    channel = models.ForeignKey(Channel, related_name='onair', on_delete=models.CASCADE)

.

class EpgSerializer(serializers.ModelSerializer):
    class Meta:
        model = Epg
        fields = ('channel', 'start', 'end', 'title', 'description',)


class ChannelSerializer(serializers.ModelSerializer):
    onair = EpgSerializer(many=True, read_only=True, source="current_epg")

    class Meta:
        model = Channel
        fields = ('number', 'name', 'onair',)

Pay attention to source="current_epg" and you'll get the point.

You can subclass the ListSerializer and overwrite the to_representation method.

By default the to_representation method calls data.all() on the nested queryset. So you effectively need to make data = data.filter(**your_filters) before the method is called. Then you need to add your subclassed ListSerializer as the list_serializer_class on the meta of the nested serializer.

1. subclass ListSerializer, overwriting to_representation and then calling super
2. add subclassed ListSerializer as the meta list_serializer_class on the nested Serializer

Here is the relevant code for your sample.

class FilteredListSerializer(serializers.ListSerializer):

    def to_representation(self, data):
        data = data.filter(user=self.context['request'].user, edition__hide=False)
        return super(FilteredListSerializer, self).to_representation(data)


class EditionSerializer(serializers.ModelSerializer):

    class Meta:
        list_serializer_class = FilteredListSerializer
        model = Edition


class QuestionnaireSerializer(serializers.ModelSerializer):
    edition = EditionSerializer(read_only=True)
    company = serializers.StringRelatedField(read_only=True)

    class Meta:
        model = Questionnaire

I find it easier, and more straight forward, to use a SerializerMethodField on the serializer field you want to filter.

So you would do something like this.

class CarTypesSerializer(serializers.ModelSerializer):

    class Meta:
        model = CarType
        fields = '__all__'


class CarSerializer(serializers.ModelSerializer):

    car_types = serializers.SerializerMethodField()

    class Meta:
        model = Car
        fields = '__all__'

    def get_car_types(self, instance):
        # Filter using the Car model instance and the CarType's related_name
        # (which in this case defaults to car_types_set)
        car_types_instances = instance.car_types_set.filter(brand="Toyota")
        return CarTypesSerializer(car_types_instances, many=True).data

This saves you from having to create many overrides of the serializers.ListSerializer if you need different filtering criteria for different serializers.

It also has the extra benefit of seeing exactly what the filter does within the serializer instead of diving into a subclass definition.

Of course the downside is if you have a serializer with many nested objects that all need to be filtered in some way. It could cause the serializer code to greatly increase. It's up to you how you would like to filter.

Hope this helps!