Skip to content

Latest commit

 

History

History
663 lines (523 loc) · 21.3 KB

File metadata and controls

663 lines (523 loc) · 21.3 KB

Usage

The DJA package implements a custom renderer, parser, exception handler, and pagination. To get started enable the pieces in settings.py that you want to use.

Many features of the JSON:API format standard have been implemented using Mixin classes in serializers.py. The easiest way to make use of those features is to import ModelSerializer variants from rest_framework_json_api instead of the usual rest_framework

Configuration

We suggest that you copy the settings block below and modify it if necessary.

REST_FRAMEWORK = {
    'PAGE_SIZE': 10,
    'EXCEPTION_HANDLER': 'rest_framework_json_api.exceptions.exception_handler',
    'DEFAULT_PAGINATION_CLASS':
        'rest_framework_json_api.pagination.JsonApiPageNumberPagination',
    'DEFAULT_PARSER_CLASSES': (
        'rest_framework_json_api.parsers.JSONParser',
        'rest_framework.parsers.FormParser',
        'rest_framework.parsers.MultiPartParser'
    ),
    'DEFAULT_RENDERER_CLASSES': (
        'rest_framework_json_api.renderers.JSONRenderer',
        # If you're performance testing, you will want to use the browseable API
        # without forms, as the forms can generate their own queries.
        # If performance testing, enable:
        # 'example.utils.BrowsableAPIRendererWithoutForms',
        # Otherwise, to play around with the browseable API, enable:
        'rest_framework.renderers.BrowsableAPIRenderer'
    ),
    'DEFAULT_METADATA_CLASS': 'rest_framework_json_api.metadata.JSONAPIMetadata',
}

Pagination

DJA pagination is based on DRF pagination.

When pagination is enabled, the renderer will return a meta object with record count and a links object with the next, previous, first, and last links.

Configuring the Pagination Style

Pagination style can be set on a particular viewset with the pagination_class attribute or by default for all viewsets by setting REST_FRAMEWORK['DEFAULT_PAGINATION_CLASS'] and by setting REST_FRAMEWORK['PAGE_SIZE'].

You can configure fixed values for the page size or limit -- or allow the client to choose the size or limit via query parameters.

Two pagination classes are available:

  • JsonApiPageNumberPagination breaks a response up into pages that start at a given page number with a given size (number of items per page). It can be configured with the following attributes:
    • page_query_param (default page[number])
    • page_size_query_param (default page[size]) Set this to None if you don't want to allow the client to specify the size.
    • max_page_size (default 100) enforces an upper bound on the page_size_query_param. Set it to None if you don't want to enforce an upper bound.
  • JsonApiLimitOffsetPagination breaks a response up into pages that start from an item's offset in the viewset for a given number of items (the limit). It can be configured with the following attributes:
    • offset_query_param (default page[offset]).
    • limit_query_param (default page[limit]).
    • max_limit (default 100) enforces an upper bound on the limit. Set it to None if you don't want to enforce an upper bound.

These examples show how to configure the parameters to use non-standard names and different limits:

from rest_framework_json_api.pagination import JsonApiPageNumberPagination, JsonApiLimitOffsetPagination

class MyPagePagination(JsonApiPageNumberPagination):
    page_query_param = 'page_number'
    page_size_query_param = 'page_size'
    max_page_size = 1000

class MyLimitPagination(JsonApiLimitOffsetPagination):
    offset_query_param = 'offset'
    limit_query_param = 'limit'
    max_limit = None

Performance Testing

If you are trying to see if your viewsets are configured properly to optimize performance, it is preferable to use example.utils.BrowsableAPIRendererWithoutForms instead of the default BrowsableAPIRenderer to remove queries introduced by the forms themselves.

Serializers

It is recommended to import the base serializer classes from this package rather than from vanilla DRF. For example,

from rest_framework_json_api import serializers

class MyModelSerializer(serializers.ModelSerializer):
    # ...

Setting the resource_name

You may manually set the resource_name property on views, serializers, or models to specify the type key in the json output. In the case of setting the resource_name property for models you must include the property inside a JSONAPIMeta class on the model. It is automatically set for you as the plural of the view or model name except on resources that do not subclass rest_framework.viewsets.ModelViewSet:

Example - resource_name on View:

class Me(generics.GenericAPIView):
    """
    Current user's identity endpoint.

    GET /me
    """
    resource_name = 'users'
    serializer_class = identity_serializers.IdentitySerializer
    allowed_methods = ['GET']
    permission_classes = (permissions.IsAuthenticated, )

If you set the resource_name property on the object to False the data will be returned without modification.

Example - resource_name on Model:

class Me(models.Model):
    """
    A simple model
    """
    name = models.CharField(max_length=100)

    class JSONAPIMeta:
        resource_name = "users"

If you set the resource_name on a combination of model, serializer, or view in the same hierarchy, the name will be resolved as following: view > serializer > model. (Ex: A view resource_name will always override a resource_name specified on a serializer or model). Setting the resource_name on the view should be used sparingly as serializers and models are shared between multiple endpoints. Setting the resource_name on views may result in a different type being set depending on which endpoint the resource is fetched from.

Inflecting object and relation keys

This package includes the ability (off by default) to automatically convert json api field names of requests and responses from the python/rest_framework's preferred underscore to a format of your choice. To hook this up include the following setting in your project settings:

JSON_API_FORMAT_FIELD_NAMES = 'dasherize'

Possible values:

  • dasherize
  • camelize (first letter is lowercase)
  • capitalize (camelize but with first letter uppercase)
  • underscore

Note: due to the way the inflector works address_1 can camelize to address1 on output but it cannot convert address1 back to address_1 on POST or PUT. Keep this in mind when naming fields with numbers in them.

Example - Without format conversion:

{
    "data": [{
        "type": "identities",
        "id": 3,
        "attributes": {
            "username": "john",
            "first_name": "John",
            "last_name": "Coltrane",
            "full_name": "John Coltrane"
        },
    }],
    "meta": {
        "pagination": {
          "count": 20
        }
    }
}

Example - With format conversion set to dasherize:

{
    "data": [{
        "type": "identities",
        "id": 3,
        "attributes": {
            "username": "john",
            "first-name": "John",
            "last-name": "Coltrane",
            "full-name": "John Coltrane"
        },
    }],
    "meta": {
        "pagination": {
          "count": 20
        }
    }
}

Types

A similar option to JSON_API_FORMAT_KEYS can be set for the types:

JSON_API_FORMAT_TYPES = 'dasherize'

Example without format conversion:

{
	"data": [{
        "type": "blog_identity",
        "id": 3,
        "attributes": {
                ...
        },
        "relationships": {
            "home_town": {
                "data": [{
                    "type": "home_town",
                    "id": 3
                }]
            }
        }
    }]
}

When set to dasherize:

{
	"data": [{
        "type": "blog-identity",
        "id": 3,
        "attributes": {
                ...
        },
        "relationships": {
            "home_town": {
                "data": [{
                    "type": "home-town",
                    "id": 3
                }]
            }
        }
    }]
}

It is also possible to pluralize the types like so:

JSON_API_PLURALIZE_TYPES = True

Example without pluralization:

{
	"data": [{
        "type": "identity",
        "id": 3,
        "attributes": {
                ...
        },
        "relationships": {
            "home_towns": {
                "data": [{
                    "type": "home_town",
                    "id": 3
                }]
            }
        }
    }]
}

When set to pluralize:

{
	"data": [{
        "type": "identities",
        "id": 3,
        "attributes": {
                ...
        },
        "relationships": {
            "home_towns": {
                "data": [{
                    "type": "home_towns",
                    "id": 3
                }]
            }
        }
    }]
}

Related fields

Because of the additional structure needed to represent relationships in JSON API, this package provides the ResourceRelatedField for serializers, which works similarly to PrimaryKeyRelatedField. By default, rest_framework_json_api.serializers.ModelSerializer will use this for related fields automatically. It can be instantiated explicitly as in the following example:

from rest_framework_json_api import serializers
from rest_framework_json_api.relations import ResourceRelatedField

from myapp.models import Order, LineItem, Customer


class OrderSerializer(serializers.ModelSerializer):
    class Meta:
        model = Order

    line_items = ResourceRelatedField(
        queryset=LineItem.objects,
        many=True  # necessary for M2M fields & reverse FK fields
    )

    customer = ResourceRelatedField(
        queryset=Customer.objects  # queryset argument is required
    )                              # except when read_only=True

In the JSON API spec, relationship objects contain links to related objects. To make this work on a serializer we need to tell the ResourceRelatedField about the corresponding view. Use the HyperlinkedModelSerializer and instantiate the ResourceRelatedField with the relevant keyword arguments:

from rest_framework_json_api import serializers
from rest_framework_json_api.relations import ResourceRelatedField

from myapp.models import Order, LineItem, Customer


class OrderSerializer(serializers.HyperlinkedModelSerializer):
    class Meta:
        model = Order

    line_items = ResourceRelatedField(
        queryset=LineItem.objects,
        many=True,
        related_link_view_name='order-lineitems-list',
        related_link_url_kwarg='order_pk',
        self_link_view_name='order_relationships'
    )

    customer = ResourceRelatedField(
        queryset=Customer.objects,
        related_link_view_name='order-customer-detail',
        related_link_url_kwarg='order_pk',
        self_link_view_name='order-relationships'
    )
  • related_link_view_name is the name of the route for the related view.

  • related_link_url_kwarg is the keyword argument that will be passed to the view that identifies the 'parent' object, so that the results can be filtered to show only those objects related to the 'parent'.

  • self_link_view_name is the name of the route for the RelationshipView (see below).

In this example, reverse('order-lineitems-list', kwargs={'order_pk': 3} should resolve to something like /orders/3/lineitems, and that route should instantiate a view or viewset for LineItem objects that accepts a keyword argument order_pk. The drf-nested-routers package is useful for defining such nested routes in your urlconf.

The corresponding viewset for the line-items-list route in the above example might look like the following. Note that in the typical use case this would be the same viewset used for the /lineitems endpoints; when accessed through the nested route /orders/<order_pk>/lineitems the queryset is filtered using the order_pk keyword argument to include only the lineitems related to the specified order.

from rest_framework import viewsets

from myapp.models import LineItem
from myapp.serializers import LineItemSerializer


class LineItemViewSet(viewsets.ModelViewSet):
    queryset = LineItem.objects
    serializer_class = LineItemSerializer

    def get_queryset(self):
        queryset = super(LineItemViewSet, self).get_queryset()

        # if this viewset is accessed via the 'order-lineitems-list' route,
        # it wll have been passed the `order_pk` kwarg and the queryset
        # needs to be filtered accordingly; if it was accessed via the
        # unnested '/lineitems' route, the queryset should include all LineItems
        if 'order_pk' in self.kwargs:
            order_pk = self.kwargs['order_pk']
            queryset = queryset.filter(order__pk=order_pk)

        return queryset

RelationshipView

rest_framework_json_api.views.RelationshipView is used to build relationship views (see the JSON API spec). The self link on a relationship object should point to the corresponding relationship view.

The relationship view is fairly simple because it only serializes Resource Identifier Objects rather than full resource objects. In most cases the following is sufficient:

from rest_framework_json_api.views import RelationshipView

from myapp.models import Order


class OrderRelationshipView(RelationshipView):
    queryset = Order.objects

The urlconf would need to contain a route like the following:

url(
    regex=r'^orders/(?P<pk>[^/.]+)/relationships/(?P<related_field>[^/.]+)$',
    view=OrderRelationshipView.as_view(),
    name='order-relationships'
)

The related_field kwarg specifies which relationship to use, so if we are interested in the relationship represented by the related model field Order.line_items on the Order with pk 3, the url would be /orders/3/relationships/line_items. On HyperlinkedModelSerializer, the ResourceRelatedField will construct the url based on the provided self_link_view_name keyword argument, which should match the name= provided in the urlconf, and will use the name of the field for the related_field kwarg. Also we can override related_field in the url. Let's say we want the url to be: /order/3/relationships/order_items - all we need to do is just add field_name_mapping dict to the class:

field_name_mapping = {
        'line_items': 'order_items'
    }

Working with polymorphic resources

Polymorphic resources allow you to use specialized subclasses without requiring special endpoints to expose the specialized versions. For example, if you had a Project that could be either an ArtProject or a ResearchProject, you can have both kinds at the same URL.

DJA tests its polymorphic support against django-polymorphic. The polymorphic feature should also work with other popular libraries like django-polymodels or django-typed-models.

Writing polymorphic resources

A polymorphic endpoint can be set up if associated with a polymorphic serializer. A polymorphic serializer takes care of (de)serializing the correct instances types and can be defined like this:

class ProjectSerializer(serializers.PolymorphicModelSerializer):
    polymorphic_serializers = [ArtProjectSerializer, ResearchProjectSerializer]

    class Meta:
        model = models.Project

It must inherit from serializers.PolymorphicModelSerializer and define the polymorphic_serializers list. This attribute defines the accepted resource types.

Polymorphic relations can also be handled with relations.PolymorphicResourceRelatedField like this:

class CompanySerializer(serializers.ModelSerializer):
    current_project = relations.PolymorphicResourceRelatedField(
        ProjectSerializer, queryset=models.Project.objects.all())
    future_projects = relations.PolymorphicResourceRelatedField(
        ProjectSerializer, queryset=models.Project.objects.all(), many=True)

    class Meta:
        model = models.Company

They must be explicitly declared with the polymorphic_serializer (first positional argument) correctly defined. It must be a subclass of serializers.PolymorphicModelSerializer.

Note: Polymorphic resources are not compatible with resource_name defined on the view.

Meta

You may add metadata to the rendered json in two different ways: meta_fields and get_root_meta.

On any rest_framework_json_api.serializers.ModelSerializer you may add a meta_fields property to the Meta class. This behaves in the same manner as the default fields property and will cause SerializerMethodFields or model values to be added to the meta object within the same data as the serializer.

To add metadata to the top level meta object add:

def get_root_meta(self, resource, many):
    if many:
      # Dealing with a list request
      return {
          'size': len(resource)
      }
    else:
      # Dealing with a detail request
      return {
        'foo': 'bar'
      }

to the serializer. It must return a dict and will be merged with the existing top level meta.

To access metadata in incoming requests, the JSONParser will add the metadata under a top level _meta key in the parsed data dictionary. For instance, to access meta data from a serializer object, you may use serializer.initial_data.get("_meta"). To customize the _meta key, see here.

Links

Adding url to fields on a serializer will add a self link to the links key.

Related links will be created automatically when using the Relationship View.

Included

JSON API can include additional resources in a single network request. The specification refers to this feature as Compound Documents. Compound Documents can reduce the number of network requests which can lead to a better performing web application. To accomplish this, the specification permits a top level included key. The list of content within this key are the extra resources that are related to the primary resource.

To make a Compound Document, you need to modify your ModelSerializer. The two required additions are included_resources and included_serializers.

For example, suppose you are making an app to go on quests, and you would like to fetch your chosen knight along with the quest. You could accomplish that with:

class KnightSerializer(serializers.ModelSerializer):
    class Meta:
        model = Knight
        fields = ('id', 'name', 'strength', 'dexterity', 'charisma')


class QuestSerializer(serializers.ModelSerializer):
    included_serializers = {
        'knight': KnightSerializer,
    }

    class Meta:
        model = Quest
        fields = ('id', 'title', 'reward', 'knight')

    class JSONAPIMeta:
        included_resources = ['knight']

included_resources informs DJA of what you would like to include. included_serializers tells DJA how you want to include it.

Performance improvements

Be aware that using included resources without any form of prefetching WILL HURT PERFORMANCE as it will introduce m*(n+1) queries.

A viewset helper was designed to allow for greater flexibility and it is automatically available when subclassing rest_framework_json_api.views.ModelViewSet:

from rest_framework_json_api import views

# When MyViewSet is called with ?include=author it will dynamically prefetch author and author.bio
class MyViewSet(views.ModelViewSet):
    queryset = Book.objects.all()
    prefetch_for_includes = {
        '__all__': [],
        'author': ['author', 'author__bio'],
        'category.section': ['category']
    }

An additional convenience DJA class exists for read-only views, just as it does in DRF.

from rest_framework_json_api import views

class MyReadOnlyViewSet(views.ReadOnlyModelViewSet):
    # ...

The special keyword __all__ can be used to specify a prefetch which should be done regardless of the include, similar to making the prefetch yourself on the QuerySet.

Using the helper to prefetch, rather than attempting to minimise queries via select_related might give you better performance depending on the characteristics of your data and database.

For example:

If you have a single model, e.g. Book, which has four relations e.g. Author, Publisher, CopyrightHolder, Category.

To display 25 books and related models, you would need to either do:

a) 1 query via selected_related, e.g. SELECT * FROM books LEFT JOIN author LEFT JOIN publisher LEFT JOIN CopyrightHolder LEFT JOIN Category

b) 4 small queries via prefetch_related.

If you have 1M books, 50k authors, 10k categories, 10k copyrightholders in the select_related scenario, you've just created a in-memory table with 1e18 rows which will likely exhaust any available memory and slow your database to crawl.

The prefetch_related case will issue 4 queries, but they will be small and fast queries.