My aim is to build endpoint which will surve to create objects of model with GenericForeignKey. Since model also includes ContentType, the actual type of model which we will reference is not known before object creation.
I will provide an example:
I have a 'Like' model which can reference a set of other models like 'Book', 'Author'.
class Like(models.Model):
created = models.DateTimeField()
content_type = models.ForeignKey(ContentType)
object_id = models.PositiveIntegerField()
content_object = GenericForeignKey('content_type', 'object_id')
Serializer may look like this:
class LikeSerializer(serializers.ModelSerializer):
class Meta:
model = models.Like
fields = ('id', 'created', )
What I want to achieve is to determine type of Like based on keys passed in request. The problem is that DRF do not pass those keys from request if they were not expilictly specified in Serializer fields. For example, POST request body contains:
{
"book":2
}
I want to do next
def restore_object(self, attrs, instance=None)
if attrs.get('book', None) is not None:
# create Like instance with Book contenttype
elif attrs.get('author', None) is not None:
# create Like instance with Author contenttype
In this case first if clause will be executed.
As you can see, The type determined based on key passed in request, without specifying special Field.
Is there any way to achieve this?
Thanks
You might try instantiating your serializer whenever your view is called by wrapping it in a function (you make a serializer factory):
def like_serializer_factory(type_of_like):
if type_of_like == 'book':
class LikeSerializer(serializers.ModelSerializer):
class Meta:
model = models.Like
fields = ('id', 'created', )
def restore_object(self, attrs, instance=None):
# create Like instance with Book contenttype
elif type_of_like == 'author':
class LikeSerializer(serializers.ModelSerializer):
class Meta:
model = models.Like
fields = ('id', 'created', )
def restore_object(self, attrs, instance=None):
# create Like instance with Author contenttype
return LikeSerializer
Then override this method in your view:
def get_serializer_class(self):
return like_serializer_factory(type_of_like)
Solution 1
Basically there is a method you can add on GenericAPIView class called get_context_serializer
By default your view, request and format class are passed to your serializer
DRF code for get_context_serializer
def get_serializer_context(self):
"""
Extra context provided to the serializer class.
"""
return {
'request': self.request,
'format': self.format_kwarg,
'view': self
}
you can override that on your view like this
def get_serializer_context(self):
data = super().get_serializer_context()
# Get the book from post and add to context
data['book'] = self.request.POST.get('book')
return data
And use this on your serializer class
def restore_object(self, attrs, instance=None):
# Get book from context to use
book = self.context.get('book', None)
author = attrs.get('author', None)
if book is not None:
# create Like instance with Book contenttype
pass
elif author is not None:
# create Like instance with Author contenttype
pass
Solution 2
Add a field on your serializer
class LikeSerializer(serializers.ModelSerializer):
# New field and should be write only, else it will be
# return as a serializer data
book = serializers.IntegerField(write_only=True)
class Meta:
model = models.Like
fields = ('id', 'created', )
def save(self, **kwargs):
# Remove book from validated data, so the serializer does
# not try to save it
self.validated_data.pop('book', None)
# Call model serializer save method
return super().save(**kwargs)
Related
Roughly said, I have the following schema in ORM:
class Page(models.Model):
title = models.CharField(max_length=255, null=False, blank=False)
#property
def content(self):
return [Video.objects.all()[0], Text.objects.all()[0], Video.objects.all()[1]]
and I have the following set of classes to support serialization for detailed view:
class ContentSerializer(serializers.ListSerializer):
class Meta:
model = ???
fields = '???'
class PageDetailSerializer(serializers.ModelSerializer):
content = ContentSerializer(many=True)
class Meta:
model = Page
fields = ('title', 'content', )
So I'm looking for a way to serialize that Page.content property - which is:
a list;
will contain heterogeneous data (combination of, let's say Video, Audio, Text and other models.
So I need somehow patch one of builtin serializers to iterate thru the list and check type of each object. And then decide how to serialize each one. E.g. I could prepare kind of dynamically created ModelSerializer with:
obj_type = type(obj)
class ContentModelSerializer(serializers.ModelSerializer):
class Meta:
model = obj_type
fields = '__all__'
serialized_obj = ContentModelSerializer(obj)
How could I implement that?
You can simply achieve this by overriding the to_representation method of Page serializer. like this:
class PageDetailSerializer(serializers.ModelSerializer):
class Meta:
model = Page
fields = ('title', 'content', )
def to_representation(self, instance):
ctx = super(PageDetailSerializer, self).to_representation(instance)
content = instance.content # property field of page, will return list of items
serialized_content = []
for c in content:
if type(c) == Video:
serialized_content.append({... serialized data of video type ..})
elif type(c) == ...
# other conditions here..
I had googled a lot before found the solution. This article has a reference to SerializerMethodField, which let you add custom handler for a field. And the final solution, which worked for me is:
class PageDetailSerializer(serializers.ModelSerializer):
_cache_serializers = {}
content = serializers.SerializerMethodField()
class Meta:
model = Page
fields = ('title', 'content', )
def _get_content_item_serializer(self, content_item_type):
if content_item_type not in self._cache_serializers:
class ContentItemSerializer(serializers.ModelSerializer):
class Meta:
model = content_item_type
exclude = ('id', 'page', )
self._cache_serializers[content_item_type] = ContentItemSerializer
return self._cache_serializers[content_item_type]
def get_content(self, page):
return [
self._get_content_item_serializer(type(content_item))(content_item).data for content_item in page.content
]
I've a serializer. I want to restrict updating a field. How would I do that?
class ABCSerializer(serializers.ModelSerializer):
class Meta:
"""Meta."""
model = ModelA
fields = ('colA', 'colB', 'colC',)
colA is a required field while creating the object. However, it should not be allowed to update. How can I do that??
Sounds like you need different serializers for PUT and POST methods. In the serializer for the PUT method you can set the colA field to readonly
class ABCViewSet(ModelViewSet):
serializer_class = ABCSerializer
def get_serializer_class(self):
serializer_class = self.serializer_class
if self.request.method == 'PUT':
serializer_class = SerializerWithReadOnlyColA
return serializer_class
You can use Django REST Frameworks field-level validation by validating that field has not changed on update like so:
from rest_framework.exceptions import ValidationError
class ABCSerializer(serializers.ModelSerializer):
colA = serializers.CharField(max_length=100)
def validate_colA(self, value):
if self.instance and self.instance.colA != value:
raise ValidationError("You may not edit colA")
return value
class Meta:
"""Meta."""
model = ModelA
fields = ('colA', 'colB', 'colC',)
This will check whether or not this is an update (via checking if an instance is populated on the serializer) and if so it will then check to see if you have made a change to the field and if you have it will throw a ValidationError. The benefit of this approach is that you can keep your view code the same as before and continue to keep your validation behaviour in your serializer.
You can override the serializer's update method to only update fields that you want.
class ABCSerializer(serializers.ModelSerializer):
def update(self, instance, validated_data):
instance.colB = validated_data.get('colB', instance.colB)
instance.colC = validated_data.get('colC', instance.colC)
# do nothing to instance.colA
instance.save()
return instance
class Meta:
model = ModelA
fields = ('colA', 'colB', 'colC',)
Or if you have many fields, and just want to omit updating colA, you could write your update method like this:
def update(self, instance, validated_data):
validated_data.pop('colA') # validated_data no longer has colA
return super().update(instance, validated_data)
You can read more about overriding update here: https://www.django-rest-framework.org/api-guide/serializers/#saving-instances
I think it's too late to answer but this may be useful for others:)
you can solve your problem this way:
class ABCSerializer(serializers.ModelSerializer):
class Meta:
model = ModelA
fields = ('colA', 'colB', 'colC',)
def get_fields(self):
fields = super().get_fields()
if self.instance:
fields["colA"].read_only = True
return fields
When you want to create, the self.instance is None, it will pass the if clause, and in case of updating the if clause will make the field read only and non-editable.
You can do this with the read_only_fieldsoption
class ABCSerializer(serializers.ModelSerializer):
class Meta:
"""Meta."""
model = ModelA
fields = ('colB', 'colC',)
read_only_fields = ('colA',)
I only want the AnchorSerializer() to serialize on a GET request so it return the serialized object as a response. On a POST request when creating an AnchorToUser object an integer is expected.
class AnchorToUserSerializer(serializers.ModelSerializer):
# Add the username from the user object relation.
user = serializers.ReadOnlyField(source='user.username')
# Serialize the nested anchor.
anchor = AnchorSerializer() # Should only be used with GET.
class Meta:
model = AnchorToUser
fields = (
'anchor',
'user',
'created_at'
)
Maybe you can use different serializers for GET and POST.
class AnchorToUserGetSerializer(serializers.ModelSerializer):
user = serializers.ReadOnlyField(source='user.username')
anchor = AnchorSerializer(read_only=True) # only used for serialization
class Meta:
model = AnchorToUser
fields = ('anchor', 'user', 'created_at')
class AnchorToUserPostSerializer(serializers.ModelSerializer):
user = serializers.ReadOnlyField(source='user.username')
anchor = serializers.IntegerField(write_only=True) # accept integer values
class Meta:
model = AnchorToUser
fields = ('anchor', 'user', 'created_at')
To decide what serializer to use, you can then override the get_serializer_class() method.
In your views or viewsets, you can do something like:
def get_serializer_class(self):
if request.method == 'POST':
return AnchorToUserPostSerializer
return AnchorToUserGetSerializer
Not sure if this has changed since 2016, but I had to have my viewset as such for this to work;
def get_serializer_class(self):
if self.action == 'create':
return AnchorToUserPostSerializer
return AnchorToUserGetSerializer
I want to hide specific fields of a model on the list display at persons/ and show all the fields on the detail display persons/jane
I am relatively new to the rest framework and the documentation feels like so hard to grasp.
Here's what I am trying to accomplish.
I have a simple Person model,
# model
class Person(models.Model):
first_name = models.CharField(max_length=30, blank=True)
last_name = models.CharField(max_length=30, blank=True)
nickname = models.CharField(max_length=20)
slug = models.SlugField()
address = models.TextField(max_length=300, blank=True)
and the serializer class
# serializers
class PersonListSerializer(serializers.ModelSerializer):
class Meta:
model = Person
fields = ('nickname', 'slug')
class PersonSerializer(serializers.ModelSerializer):
class Meta:
model = Person
fields = ('first_name', 'last_name', 'nickname', 'slug', 'address')
and the viewsets.
# view sets (api.py)
class PersonListViewSet(viewsets.ModelViewSet):
queryset = Person.objects.all()
serializer_class = PersonListSerializer
class PersonViewSet(viewsets.ModelViewSet):
queryset = Person.objects.all()
serializer_class = PersonSerializer
at the url persons I want to dispaly list of persons, just with fields nickname and slug and at the url persons/[slug] I want to display all the fields of the model.
my router configurations,
router = routers.DefaultRouter()
router.register(r'persons', api.PersonListViewSet)
router.register(r'persons/{slug}', api.PersonViewSet)
I guess the second configuration is wrong, How can I achieve what I am trying to do?
update:
the output to persons/slug is {"detail":"Not found."} but it works for person/pk
Thank you
For anyone else stumbling across this, I found overriding get_serializer_class on the viewset and defining a serializer per action was the DRY-est option (keeping a single viewset but allowing for dynamic serializer choice):
class MyViewset(viewsets.ModelViewSet):
serializer_class = serializers.ListSerializer
permission_classes = [permissions.IsAdminUser]
renderer_classes = (renderers.AdminRenderer,)
queryset = models.MyModel.objects.all().order_by('-updated')
def __init__(self, *args, **kwargs):
super(MyViewset, self).__init__(*args, **kwargs)
self.serializer_action_classes = {
'list':serializers.AdminListSerializer,
'create':serializers.AdminCreateSerializer,
'retrieve':serializers.AdminRetrieveSerializer,
'update':serializers.AdminUpdateSerializer,
'partial_update':serializers.AdminUpdateSerializer,
'destroy':serializers.AdminRetrieveSerializer,
}
def get_serializer_class(self, *args, **kwargs):
"""Instantiate the list of serializers per action from class attribute (must be defined)."""
kwargs['partial'] = True
try:
return self.serializer_action_classes[self.action]
except (KeyError, AttributeError):
return super(MyViewset, self).get_serializer_class()
Hope this helps someone else.
You can override the 'get_fields' method your serializer class and to add something like that:
def get_fields(self, *args, **kwargs):
fields = super().get_fields(*args, **kwargs)
request = self.context.get('request')
if request is not None and not request.parser_context.get('kwargs'):
fields.pop('your_field', None)
return fields
In this case when you get detail-view there is 'kwargs': {'pk': 404} and when you get list-view there is 'kwargs': {}
I wrote an extension called drf-action-serializer (pypi) that adds a serializer called ModelActionSerializer that allows you to define fields/exclude/extra_kwargs on a per-action basis (while still having the normal fields/exclude/extra_kwargs to fall back on).
The implementation is nice because you don't have to override your ViewSet get_serializer method because you're only using a single serializer. The relevant change is that in the get_fields and get_extra_kwargs methods of the serializer, it inspects the view action and if that action is present in the Meta.action_fields dictionary, then it uses that configuration rather than the Meta.fields property.
In your example, you would do this:
from action_serializer import ModelActionSerializer
class PersonSerializer(ModelActionSerializer):
class Meta:
model = Person
fields = ('first_name', 'last_name', 'nickname', 'slug', 'address')
action_fields = {
'list': {'fields': ('nickname', 'slug')}
}
Your ViewSet would look something like:
class PersonViewSet(viewsets.ModelViewSet):
queryset = Person.objects.all()
serializer_class = PersonSerializer
And your router would look normal, too:
router = routers.DefaultRouter()
router.register(r'persons', api.PersonViewSet)
Implementation
If you're curious how I implemented this:
I added a helper method called get_action_config which gets the current view action and returns that entry in the action_fields dict:
def get_action_config(self):
"""
Return the configuration in the `Meta.action_fields` dictionary for this
view's action.
"""
view = getattr(self, 'context', {}).get('view', None)
action = getattr(view, 'action', None)
action_fields = getattr(self.Meta, 'action_fields', {})
I changed get_field_names of ModelSerializer:
From:
fields = getattr(self.Meta, 'fields', None)
exclude = getattr(self.Meta, 'exclude', None)
To:
action_config = self.get_action_config()
if action_config:
fields = action_config.get('fields', None)
exclude = action_config.get('exclude', None)
else:
fields = getattr(self.Meta, 'fields', None)
exclude = getattr(self.Meta, 'exclude', None)
Finally, I changed the get_extra_kwargs method:
From:
extra_kwargs = copy.deepcopy(getattr(self.Meta, 'extra_kwargs', {}))
To:
action_config = self.get_action_config()
if action_config:
extra_kwargs = copy.deepcopy(action_config.get('extra_kwargs', {}))
else:
extra_kwargs = copy.deepcopy(getattr(self.Meta, 'extra_kwargs', {}))
If you want to change what fields are displayed in the List vs Detail view, the only thing you can do is change the Serializer used. There's no field that I know of that lets you specify which fields of the Serializer gets used.
The field selection on you serializers should be working, but I don't know what might be happening exactly. I have two solutions you can try:
1 Try to change the way you declare you serializer object
#If you aren't using Response:
from rest_framework.response import Response
class PersonListViewSet(viewsets.ModelViewSet):
def get(self, request):
queryset = Person.objects.all()
serializer_class = PersonListSerializer(queryset, many=True) #It may change the things
return Response(serializer_class.data)
class PersonViewSet(viewsets.ModelViewSet):
def get(self, request, pk): #specify the method is cool
queryset = Person.objects.all()
serializer_class = PersonSerializer(queryset, many=True) #Here as well
#return Response(serializer_class.data)
2 The second way around would change your serializers
This is not the most normal way, since the field selector should be working but you can try:
class PersonListSerializer(serializers.ModelSerializer):
nickname = serializers.SerializerMethodField() #Will get the attribute my the var name
slug = serializers.SerializerMethodField()
class Meta:
model = Person
def get_nickname(self, person):
#This kind of method should be like get_<fieldYouWantToGet>()
return person.nickname
def get_slug(self, person):
#This kind of method should be like get_<fieldYouWantToGet>()
return person.slug
I hope it helps. Try to see the APIview class for building your view too.
Somehow close:
If you just want to skip fields in the serilaizer
class UserSerializer(serializers.ModelSerializer):
user_messages = serializers.SerializerMethodField()
def get_user_messages(self, obj):
if self.context.get('request').user != obj:
# do somthing here check any value from the request:
# skip others msg
return
# continue with your code
return SystemMessageController.objects.filter(user=obj, read=False)
I rewrite ModelViewSet list function to modify serializer_class.Meta.fields attribute, code like this:
class ArticleBaseViewSet(BaseViewSet):
def list(self, request, *args, **kwargs):
exclude = ["content"]
self.serializer_class.Meta.fields = [f.name for f in self.serializer_class.Meta.model._meta.fields if f.name not in exclude]
queryset = self.filter_queryset(self.get_queryset()).filter(is_show=True, is_check=True)
page = self.paginate_queryset(queryset)
if page is not None:
serializer = self.get_serializer(page, many=True)
return self.get_paginated_response(serializer.data)
serializer = self.get_serializer(queryset, many=True)
return Response(serializer.data)
class BannerArticleViewSet(ArticleBaseViewSet):
queryset = BannerArticle.objects.filter(is_show=True, is_check=True).all()
serializer_class = BannerArticleSerializer
permission_classes = (permissions.AllowAny,)
But it looks not stable, so i will not use it, just share to figure out the best way
My solution.
class BaseSerializerMixin(_ModelSerializer):
class Meta:
exclude: tuple[str, ...] = ()
exclude_in_list: tuple[str, ...] = ()
model: Type[_models.Model]
def get_action(self) -> Optional[str]:
if 'request' not in self.context:
return None
return self.context['request'].parser_context['view'].action
def get_fields(self):
fields = super().get_fields()
if self.get_action() == 'list':
[fields.pop(i) for i in list(fields) if i in self.Meta.exclude_in_list]
return fields
I think it should be like this:
router.register(r'persons/?P<slug>/', api.PersonViewSet)
and you should include a line like this:
lookup_field='slug'
in your serializer class. Like this:
class PersonSerializer(serializers.ModelSerializer):
lookup_field='slug'
class Meta:
model = Person
fields = ('first_name', 'last_name', 'nickname', 'slug', 'address')
I have a Django Model w/ a m2m relationship that uses a through model:
models.py
class ModelA(models.Model):
name = models.CharField(max_length=64)
class ModelB(models.Model):
name = models.CharField(max_length=64)
other_models = models.ManyToManyField("ModelA", through="ModelC")
class ModelC(models.Model):
model_a = models.ForeignKey("ModelA", related_name="link_to_model_a")
model_b = models.ForeignKey("ModelB", related_name="link_to_model_b")
some_other_info = models.TextField()
class Meta:
unique_together = ("model_a", "model_b", )
I want to serialize this using django-rest-framework:
serializers.py
class ModelCSerializer(ModelSerializer):
class Meta:
model = ModelC
fields = ('id', 'model_a', 'model_b', 'some_other_info', )
class QModelBSerializer(ModelSerializer):
class Meta:
model = ModelB
fields = ('id', 'other_models', )
other_models = ModelCSerializer(many=True, required=False, source="link_to_model_b")
Now, for existing models the GET displays properly:
{
"id": 2,
"name": "i am an instance of model_b",
"other_models": [
{"id": 1, "model_a": 1,"model_b": 2, "some_other_info":"here is some other info"}
],
}
But, if I try to PUT some data it fails w/ a unique_together error. I thought that sending this as a PUT would cause an update (which shouldn't raise a unique_together error) not a create? Here is the code for PUT:
views.py
class ModelBDetail(APIView):
def put(self, request, pk, format=None):
model = ModelB.objects.get(id=pk)
serializer = ModelBSerializer(model, data=request.data, context={"request": request})
if serializer.is_valid(): # THIS IS RETURNING FALSE
serializer.save()
return Response(serializer.data)
return Response(serializer.errors, status=status.HTTP_400_BAD_REQUEST)
Any thoughts?
Django rest framework documentation states that it is up to developer to implement creates and updates for nested representations.
Although #Ivan was correct about writing my own create & update fn, the specific issue I was seeing was that the nested serialization did not have an instance attribute associated with it.
The new code looks like this:
serializers.py
class ModelBSerializer(ModelSerializer):
....
def update(self, model_instance, validated_data):
model_c_serializer = self.fields["other_models"]
model_c_data = validated_data.pop(model_c_serializer.source, [])
for key, value in validated_data.iteritems():
setattr(model_instance, key, value)
model_instance.save()
model_c_serializer.update(model_instance.link_to_model_b.all(),
model_c_data)
return model_instance
class ModelCSerializer(ModelSerializer):
...
def to_internal_value(self, data):
# this is as good a place as any to set the instance
try:
model_class = self.Meta.model
self.instance = model_class.objects.get(pk=data.get("id"))
except ObjectDoesNotExist:
pass
return super(ModelCSerializer, self).to_internal_value(data)
Basically, I call update for the nested serializers explicitly and I also force each nested serializer to check the data that is passed to them for an instance.