I am using a DRF serializer to update data in bulk. This is how I instantiate the serializer: # Order incoming data and DB instances data = sorted(data, key=lambda x: x["id"]) instances = WorksheetChecklistItem.objects.filter(id__in=(row["id"] for row in data)).order_by("id") # Serialize and save serializer = update.WorksheetChecklistItemSerializer(instance=instances, data=data, many=True) if not serializer.is_valid(): # ... more logic ... And this is the serializer: class WorksheetChecklistItemSerializer(serializers.ModelSerializer): class Meta: model = WorksheetChecklistItem fields = ["id", "value", "outcome"] def update(self, instance, validated_data): instance.outcome = validated_data.get("outcome", instance.outcome) instance.value = validated_data.get("value", instance.value) instance.done = instance.outcome is not None instance.save() return instance def validate(self, data): """Custom validation for the checklist item.""" instance = self.instance if not instance: raise serializers.ValidationError("Instance is required for validation") # Only update is allowed # Validate "must" condition if instance.must and not data.get("done"): raise serializers.ValidationError(f"Checklist item {instance.id} is required but not completed.") # Validate that value is a number if instance.check_type == WorksheetChecklistItem.CheckType.VALUE and not isinstance(data.get("value"), (int, float)): raise serializers.ValidationError(f"Checklist item {instance.id} requires a numeric value.") return data So I'm relying on the default ListSerializer class that is triggered when the serializer is instantiated with the many=True argument. My validation fails because the validate method does not have an "instance" argument like the update method does; and self.instance is the entire queryset, not only the "current" instance. So this instruction fails: if instance.must because "instance" is a queryset and doesn't have a "must" attribute. At the moment, my only solution is to get the object from the database using the id I have in the "data" argument; but that would result in a lot of queries. Is there any way to validate data vs instance? Continue reading...
