Is there a way to add duplicate keys to json with python?
From my understanding, you can't have duplicate keys in python dictionaries. Usually, how I go about creating json is to create the dictionary and then json.dumps. However, I need duplicated keys within the JSON for testing purposes. But I can't do so because I can't add duplicate keys in a python dictionary. I am trying to doing this in python 3
You could always construct such a string value by hand.
On the other hand, one can make the CPython json module to encode duplicate keys. This is very tricky in Python 2 because json module does not respect duck-typing at all.
The straightforward solution would be to inherit from collections.Mapping - well you can't, since "MyMapping is not a JSON serializable."
Next one tries to subclass a dict - well, but if json.dumps notices that the type is dict, it skips from calling __len__, and sees the underlying dict directly - if it is empty, {} is output directly, so clearly if we fake the methods, the underlying dictionary must not be empty.
The next source of joy is that actually __iter__ is called, which iterates keys; and for each key, the __getitem__ is called, so we need to remember what is the corresponding value to return for the given key... thus we arrive to a very ugly solution for Python 2:
class FakeDict(dict):
def __init__(self, items):
# need to have something in the dictionary
self['something'] = 'something'
self._items = items
def __getitem__(self, key):
return self.last_val
def __iter__(self):
def generator():
for key, value in self._items:
self.last_val = value
yield key
return generator()
In CPython 3.3+ it is slightly easier... no, collections.abc.Mapping does not work, yes, you need to subclass a dict, yes, you need to fake that your dictionary has content... but the internal JSON encoder calls items instead of __iter__ and __getitem__!
Thus on Python 3:
import json
class FakeDict(dict):
def __init__(self, items):
self['something'] = 'something'
self._items = items
def items(self):
return self._items
print(json.dumps(FakeDict([('a', 1), ('a', 2)])))
prints out
{"a": 1, "a": 2}
'{}'
obj = json.loads(data, object_pairs_hook=FakeDict) and then json.dumps(obj).Thanks a lot Antti Haapala, I figured out you can even use this to convert an array of tuples into a FakeDict:
def function():
array_of_tuples = []
array_of_tuples.append(("key","value1"))
array_of_tuples.append(("key","value2"))
return FakeDict(array_of_tuples)
print(json.dumps(function()))
Output:
{"key": "value1", "key": "value2"}
And if you change the FakeDict class to this Empty dictionaries will be correctly parsed:
class FakeDict(dict):
def __init__(self, items):
if items != []:
self['something'] = 'something'
self._items = items
def items(self):
return self._items
def test():
array_of_tuples = []
return FakeDict(array_of_tuples)
print(json.dumps(test()))
Output:
"{}"
Actually, it's very easy:
$> python -c "import json; print json.dumps({1: 'a', '1': 'b'})"
{"1": "b", "1": "a"}
