drf_orjson_renderer
is JSON renderer and parser for Django Rest Framework
using the orjson library. Backed by
Rust, orjson is safe, correct and fast. ⚡️
In addition, unlike some performance optimized DRF renderers, It also renders pretty printed JSON when requests are made via RestFramework's BrowsableAPI.
You get:
- The safety of Rust
- The speed of orjson when requests are made with
Accept: appliation/json
HTTP header or when requests are made with an unspecifiedAccept
header. - The convenience of formatted output when requested with
Accept: text/html
. - The ability to pass your own
default
function definition.
pip install drf_orjson_renderer
You can then set the ORJSONRenderer
class as your default renderer in your settings.py
REST_FRAMEWORK = {
"DEFAULT_RENDERER_CLASSES": (
"drf_orjson_renderer.renderers.ORJSONRenderer",
"rest_framework.renderers.BrowsableAPIRenderer",
),
}
To modify how data is serialized, specify options in your settings.py
REST_FRAMEWORK = {
"ORJSON_RENDERER_OPTIONS": (
orjson.OPT_NON_STR_KEYS,
orjson.OPT_SERIALIZE_DATACLASS,
orjson.OPT_SERIALIZE_NUMPY,
),
}
Also you can set the ORJSONParser
class as your default parser in your settings.py
REST_FRAMEWORK = {
"DEFAULT_PARSER_CLASSES": (
"drf_orjson_renderer.parsers.ORJSONParser",
),
}
By default, the ORJSONRenderer
will pass a default
function as a helper for
serializing objects that orjson doesn't recognize. That should cover the most
common cases found in a Django web application. If you find you have an object
it doesn't recognize you can pass your own default function by overriding the
get_renderer_context()
method of your view:
from rest_framework.views import APIView
from rest_framework.response import Response
class MyView(APIView):
def default(self, obj):
if isinstance(obj, MyComplexData):
return dict(obj)
def get_renderer_context(self):
renderer_context = super().get_renderer_context()
renderer_context["default_function"] = self.default
return renderer_context
def get(self, request, *args, **kwargs):
my_complex_data = MyComplexData()
return Response(data=my_complex_data)
If you know your data is already in a format orjson natively
recognizes you can get a small
performance boost by passing None
to the renderer_context
:
def get_renderer_context(self):
renderer_context = super().get_renderer_context()
renderer_context["default_function"] = None
return renderer_context
As of ORJSON version 3, 2-space indenting is supported in serialization. In
order to take advantage of the RestFramework Browsable API, when the
requested media type is not application/json
, the ORJSON renderer will add
orjson.OPT_INDENT_2
to the options mask to pretty print your output.
When this package was originally written ORJSON did not natively support serializing numpy types. This package provided an encoder class that overrides the DjangoJSONEncoder with support for numpy types. This encoder is no longer necessary but included for backwards compatibility.
from drf_orjson_renderer.encoders import DjangoNumpyJSONEncoder
from rest_framework.views import APIView
class MyView(APIView):
def get_renderer_context(self):
renderer_context = super().get_renderer_context()
renderer_context["django_encoder_class"] = DjangoNumpyJSONEncoder
return renderer_context
See the orjson Benchmarks for more information