JSON Encoding with Django adding extra \\ characters

I'm trying to create a function that will convert a dictionary containing a message and a Django model instance into JSON, that I can pass back to the client. For example, I have the model Test defined in models.py.

from django.db import models

class Test(models.Model):
    test_field = models.CharField(max_length=40)

I've defined this extension of the simplejson JSONEncoder based on the this stackoverflow question:

from django.core.serializers import serialize
from django.utils.simplejson import dumps, loads, JSONEncoder
from django.db.models.query import QuerySet
from django.db import models
from django.utils.functional import curry

class DjangoJSONEncoder(JSONEncoder):
    def default(self, obj):
        if isinstance(obj, QuerySet):
            # `default` must return a python serializable
            # structure, the easiest way is to load the JSON
            # string produced by `serialize` and return it
            return loads(serialize('json', obj))
        if isinstance(obj, models.Model):
            #do the same as above by making it a queryset first
            set_obj = [obj]
            set_str = serialize('json', set_obj)
            #eliminate brackets in the beginning and the end 
            str_obj = set_str[1:len(set_str)-2]
            return str_obj
        return JSONEncoder.default(self,obj)

# partial function, we can now use dumps(my_dict) instead
# of dumps(my_dict, cls=DjangoJSONEncoder)
dumps = curry(dumps, cls=DjangoJSONEncoder)

Then I go about creating an instance of this along with a status message:

t = Test(test_field="hello")
d = {"entry": t, "message": "Congratulations"}
json = dumps(d)

The contents of json are:

{"entry": "{\\"pk\\": null, \\"model\\": \\"hours.test\\", \\"fields\\": {\\"test_field\\": \\"hello\\"}", "message": "Congratulations"}

Which is basically what I want except for all the extra \\ character开发者_如何转开发s. Why are these being inserted into the json? How can I modify my DjangoJSONEncoder so it doesn't insert the \ characters?

NOTE

If I just encode the model instance manually I don't get all the extra \\ characters.

s = serialize('json', [t])
s[1:len(s)-2]

This outputs:

{"pk": null, "model": "hours.test", "fields": {"test_field": "hello"}

EDIT

Based on the advice of Daniel Roseman and Leopd I modified the DjangoJSONEncoder class to the following:

class DjangoJSONEncoder(JSONEncoder):
    def default(self, obj):
        if isinstance(obj, QuerySet):
            # `default` must return a python serializable
            # structure, the easiest way is to load the JSON
            # string produced by `serialize` and return it
            return loads(serialize('python', obj))
        if isinstance(obj, models.Model):
            #do the same as above by making it a list first
            return serialize('python', [obj])[0]
        return JSONEncoder.default(self,obj)

Your logic is wrong, unfortunately. Your "easiest way", as you state, returns a string - but you don't want a string at that point, you want a dictionary. You end up serializing a string within a string, hence the extra quotes which need to be escaped.

Luckily, one of the format options for the serialize function is python - which "serializes" the queryset to a Python dictionary. So you just need:

return serialize('python', obj))

You're serializing your Test model object into a JSON string, and then the entry field in d is the string, being serialized, not the data structure.