Relational fields are used to represent model relationships. They can be applied to `ForeignKey`, `ManyToManyField` and `OneToOneField` relationships, as well as to reverse relationships, and custom relationships such as `GenericForeignKey`.
**Note:** The relational fields are declared in `relations.py`, but by convention you should import them from the `serializers` module, using `from rest_framework import serializers` and refer to fields as `serializers.<FieldName>`.
**Note:** REST Framework does not attempt to automatically optimize querysets passed to serializers in terms of `select_related` and `prefetch_related` since it would be too much magic. A serializer with a field spanning an orm relation through its source attribute could require an additional database hit to fetch related objects from the database. It is the programmer's responsibility to optimize queries to avoid additional database hits which could occur while using such a serializer.
For example, the following serializer would lead to a database hit each time evaluating the tracks field if it is not prefetched:
class AlbumSerializer(serializers.ModelSerializer):
tracks = serializers.SlugRelatedField(
many=True,
read_only=True,
slug_field='title'
)
class Meta:
model = Album
fields = ['album_name', 'artist', 'tracks']
# For each album object, tracks should be fetched from database
qs = Album.objects.all()
print(AlbumSerializer(qs, many=True).data)
If `AlbumSerializer` is used to serialize a fairly large queryset with `many=True` then it could be a serious performance problem. Optimizing the queryset passed to `AlbumSerializer` with:
When using the `ModelSerializer` class, serializer fields and relationships will be automatically generated for you. Inspecting these automatically generated fields can be a useful tool for determining how to customize the relationship style.
To do so, open the Django shell, using `python manage.py shell`, then import the serializer class, instantiate it, and print the object representation…
>>> from myapp.serializers import AccountSerializer
In order to explain the various types of relational fields, we'll use a couple of simple models for our examples. Our models will be for music albums, and the tracks listed on each album.
*`queryset` - The queryset used for model instance lookups when validating the field input. Relationships must either set a queryset explicitly, or set `read_only=True`.
*`pk_field` - Set to a field to control serialization/deserialization of the primary key's value. For example, `pk_field=UUIDField(format='hex')` would serialize a UUID primary key into its compact hex representation.
**Note**: This field is designed for objects that map to a URL that accepts a single URL keyword argument, as set using the `lookup_field` and `lookup_url_kwarg` arguments.
This is suitable for URLs that contain a single primary key or slug argument as part of the URL.
If you require more complex hyperlinked representation you'll need to customize the field, as described in the [custom hyperlinked fields](#custom-hyperlinked-fields) section, below.
*`view_name` - The view name that should be used as the target of the relationship. If you're using [the standard router classes][routers] this will be a string with the format `<modelname>-detail`. **required**.
*`queryset` - The queryset used for model instance lookups when validating the field input. Relationships must either set a queryset explicitly, or set `read_only=True`.
*`lookup_field` - The field on the target that should be used for the lookup. Should correspond to a URL keyword argument on the referenced view. Default is `'pk'`.
*`lookup_url_kwarg` - The name of the keyword argument defined in the URL conf that corresponds to the lookup field. Defaults to using the same value as `lookup_field`.
*`format` - If using format suffixes, hyperlinked fields will use the same format suffix for the target unless overridden by using the `format` argument.
## SlugRelatedField
`SlugRelatedField` may be used to represent the target of the relationship using a field on the target.
By default this field is read-write, although you can change this behavior using the `read_only` flag.
When using `SlugRelatedField` as a read-write field, you will normally want to ensure that the slug field corresponds to a model field with `unique=True`.
*`slug_field` - The field on the target that should be used to represent it. This should be a field that uniquely identifies any given instance. For example, `username`. **required**
*`queryset` - The queryset used for model instance lookups when validating the field input. Relationships must either set a queryset explicitly, or set `read_only=True`.
This field can be applied as an identity relationship, such as the `'url'` field on a HyperlinkedModelSerializer. It can also be used for an attribute on the object. For example, the following serializer:
class AlbumSerializer(serializers.HyperlinkedModelSerializer):
*`view_name` - The view name that should be used as the target of the relationship. If you're using [the standard router classes][routers] this will be a string with the format `<model_name>-detail`. **required**.
*`lookup_field` - The field on the target that should be used for the lookup. Should correspond to a URL keyword argument on the referenced view. Default is `'pk'`.
*`lookup_url_kwarg` - The name of the keyword argument defined in the URL conf that corresponds to the lookup field. Defaults to using the same value as `lookup_field`.
*`format` - If using format suffixes, hyperlinked fields will use the same format suffix for the target unless overridden by using the `format` argument.
By default nested serializers are read-only. If you want to support write-operations to a nested serializer field you'll need to create `create()` and/or `update()` methods in order to explicitly specify how the child relationships should be saved:
To implement a custom relational field, you should override `RelatedField`, and implement the `.to_representation(self, value)` method. This method takes the target of the field as the `value` argument, and should return the representation that should be used to serialize the target. The `value` argument will typically be a model instance.
To provide a dynamic queryset based on the `context`, you can also override `.get_queryset(self)` instead of specifying `.queryset` on the class or when initializing the field.
If you want to support a writable hyperlinked field then you'll also want to override `get_object`, in order to map incoming URLs back to the object they represent. For read-only hyperlinked fields there is no need to override this method.
The return value of this method should the object that corresponds to the matched URL conf arguments.
May raise an `ObjectDoesNotExist` exception.
## Example
Say we have a URL for a customer object that takes two keyword arguments, like so:
/api/<organization_slug>/customers/<customer_pk>/
This cannot be represented with the default implementation, which accepts only a single lookup field.
In this case we'd need to override `HyperlinkedRelatedField` to get the behavior we want:
from rest_framework import serializers
from rest_framework.reverse import reverse
class CustomerHyperlink(serializers.HyperlinkedRelatedField):
# We define these as class attributes, so we don't need to pass them as arguments.
Note that if you wanted to use this style together with the generic views then you'd also need to override `.get_object` on the view in order to get the correct lookup behavior.
Generally we recommend a flat style for API representations where possible, but the nested URL style can also be reasonable when used in moderation.
The `queryset` argument is only ever required for *writable* relationship field, in which case it is used for performing the model instance lookup, that maps from the primitive user input, into a model instance.
In version 2.x a serializer class could *sometimes* automatically determine the `queryset` argument *if* a `ModelSerializer` class was being used.
This behavior is now replaced with *always* using an explicit `queryset` argument for writable relational fields.
Doing so reduces the amount of hidden 'magic' that `ModelSerializer` provides, makes the behavior of the field more clear, and ensures that it is trivial to move between using the `ModelSerializer` shortcut, or using fully explicit `Serializer` classes.
The built-in `__str__` method of the model will be used to generate string representations of the objects used to populate the `choices` property. These choices are used to populate select HTML inputs in the browsable API.
To provide customized representations for such inputs, override `display_value()` of a `RelatedField` subclass. This method will receive a model object, and should return a string suitable for representing it. For example:
class TrackPrimaryKeyRelatedField(serializers.PrimaryKeyRelatedField):
When rendered in the browsable API relational fields will default to only displaying a maximum of 1000 selectable items. If more items are present then a disabled option with "More than 1000 items…" will be displayed.
This behavior is intended to prevent a template from being unable to render in an acceptable timespan due to a very large number of relationships being displayed.
There are two keyword arguments you can use to control this behavior:
-`html_cutoff` - If set this will be the maximum number of choices that will be displayed by a HTML select drop down. Set to `None` to disable any limiting. Defaults to `1000`.
-`html_cutoff_text` - If set this will display a textual indicator if the maximum number of items have been cutoff in an HTML select drop down. Defaults to `"More than {count} items…"`
In cases where the cutoff is being enforced you may want to instead use a plain input field in the HTML form. You can do so using the `style` keyword argument. For example:
Note that reverse relationships are not automatically included by the `ModelSerializer` and `HyperlinkedModelSerializer` classes. To include a reverse relationship, you must explicitly add it to the fields list. For example:
You'll normally want to ensure that you've set an appropriate `related_name` argument on the relationship, that you can use as the field name. For example:
If you have not set a related name for the reverse relationship, you'll need to use the automatically generated related name in the `fields` argument. For example:
If you want to serialize a generic foreign key, you need to define a custom field, to determine explicitly how you want to serialize the targets of the relationship.
We could define a custom field that could be used to serialize tagged instances, using the type of each instance to determine how it should be serialized:
If you need the target of the relationship to have a nested representation, you can use the required serializers inside the `.to_representation()` method:
raise Exception('Unexpected type of tagged object')
return serializer.data
Note that reverse generic keys, expressed using the `GenericRelation` field, can be serialized using the regular relational field types, since the type of the target in the relationship is always known.
For more information see [the Django documentation on generic relations][generic-relations].
If you wish to represent [extra fields on a through model][django-intermediary-manytomany] then you may serialize the through model as [a nested object][dealing-with-nested-objects].
