210

I'm using json.dumps to convert into json like

countries.append({"id":row.id,"name":row.name,"timezone":row.timezone})
print json.dumps(countries)

The result i have is:

[
   {"timezone": 4, "id": 1, "name": "Mauritius"}, 
   {"timezone": 2, "id": 2, "name": "France"}, 
   {"timezone": 1, "id": 3, "name": "England"}, 
   {"timezone": -4, "id": 4, "name": "USA"}
]

I want to have the keys in the following order: id, name, timezone - but instead I have timezone, id, name.

How should I fix this?

8 Answers 8

317

Both Python dict (before Python 3.7) and JSON object are unordered collections. You could pass sort_keys parameter, to sort the keys:

>>> import json
>>> json.dumps({'a': 1, 'b': 2})
'{"b": 2, "a": 1}'
>>> json.dumps({'a': 1, 'b': 2}, sort_keys=True)
'{"a": 1, "b": 2}'

If you need a particular order; you could use collections.OrderedDict:

>>> from collections import OrderedDict
>>> json.dumps(OrderedDict([("a", 1), ("b", 2)]))
'{"a": 1, "b": 2}'
>>> json.dumps(OrderedDict([("b", 2), ("a", 1)]))
'{"b": 2, "a": 1}'

Since Python 3.6, the keyword argument order is preserved and the above can be rewritten using a nicer syntax:

>>> json.dumps(OrderedDict(a=1, b=2))
'{"a": 1, "b": 2}'
>>> json.dumps(OrderedDict(b=2, a=1))
'{"b": 2, "a": 1}'

See PEP 468 – Preserving Keyword Argument Order.

If your input is given as JSON then to preserve the order (to get OrderedDict), you could pass object_pair_hook, as suggested by @Fred Yankowski:

>>> json.loads('{"a": 1, "b": 2}', object_pairs_hook=OrderedDict)
OrderedDict([('a', 1), ('b', 2)])
>>> json.loads('{"b": 2, "a": 1}', object_pairs_hook=OrderedDict)
OrderedDict([('b', 2), ('a', 1)])
Sign up to request clarification or add additional context in comments.

5 Comments

OrderedDict's init really ugly
@jean: the initial value has nothing to do with OrderedDict(), you can pass a dict to OrderedDict(), you can pass a list of ordered pairs to dict() too -- though the order is lost in both of these cases.
I mean init it when preserve the order, needs typing many '(' and ')'
Also, if you load JSON using d = json.load(f, object_pairs_hook=OrderedDict), a later json.dump(d) will retain the order of the original elements.
24

As others have mentioned the underlying dict is unordered. However there are OrderedDict objects in python. ( They're built in in recent pythons, or you can use this: http://code.activestate.com/recipes/576693/ ).

I believe that newer pythons json implementations correctly handle the built in OrderedDicts, but I'm not sure (and I don't have easy access to test).

Old pythons simplejson implementations dont handle the OrderedDict objects nicely .. and convert them to regular dicts before outputting them.. but you can overcome this by doing the following:

class OrderedJsonEncoder( simplejson.JSONEncoder ):
   def encode(self,o):
      if isinstance(o,OrderedDict.OrderedDict):
         return "{" + ",".join( [ self.encode(k)+":"+self.encode(v) for (k,v) in o.iteritems() ] ) + "}"
      else:
         return simplejson.JSONEncoder.encode(self, o)

now using this we get:

>>> import OrderedDict
>>> unordered={"id":123,"name":"a_name","timezone":"tz"}
>>> ordered = OrderedDict.OrderedDict( [("id",123), ("name","a_name"), ("timezone","tz")] )
>>> e = OrderedJsonEncoder()
>>> print e.encode( unordered )
{"timezone": "tz", "id": 123, "name": "a_name"}
>>> print e.encode( ordered )
{"id":123,"name":"a_name","timezone":"tz"}

Which is pretty much as desired.

Another alternative would be to specialise the encoder to directly use your row class, and then you'd not need any intermediate dict or UnorderedDict.

3 Comments

Note that JSON objects are unordered still; a JSON client can read the object definition and completely ignore the order of the keys and be fully RFC compliant.
Martijn is correct, this does not effect RFC compliance, but it can certainly still be valuable if you want to have a consistent format for your JSON (For example if the file is under version control, or to make it easier for a human reader to comprehend, of to make entry order match your documentation.)
In which case you just set sort_keys to True when calling json.dumps(); for order stability (for testing, stable caching or VCS commits), sorting keys is enough.
16

hey i know it is so late for this answer but add sort_keys and assign false to it as follows :

json.dumps({'****': ***},sort_keys=False)

this worked for me

2 Comments

The default value is already False.
If sort_keys is true (default: False), then the output of dictionaries will be sorted by key. Thanks for the hint!
9

The order of a dictionary doesn't have any relationship to the order it was defined in. This is true of all dictionaries, not just those turned into JSON.

>>> {"b": 1, "a": 2}
{'a': 2, 'b': 1}

Indeed, the dictionary was turned "upside down" before it even reached json.dumps:

>>> {"id":1,"name":"David","timezone":3}
{'timezone': 3, 'id': 1, 'name': 'David'}

Comments

8

json.dump() will preserve the ordder of your dictionary. Open the file in a text editor and you will see. It will preserve the order regardless of whether you send it an OrderedDict.

But json.load() will lose the order of the saved object unless you tell it to load into an OrderedDict(), which is done with the object_pairs_hook parameter as J.F.Sebastian instructed above.

It would otherwise lose the order because under usual operation, it loads the saved dictionary object into a regular dict and a regular dict does not preserve the oder of the items it is given.

1 Comment

This is actually a better fix as maintaining the order on load takes care of dump time ordering. Thanks for this answer.
2

Provided you're using Python 3.7+ , it does preserve the order.

Prior to Python 3.7, dict was not guaranteed to be ordered, so inputs and outputs were typically scrambled unless collections.OrderedDict was specifically requested. Starting with Python 3.7, the regular dict became order preserving, so it is no longer necessary to specify collections.OrderedDict for JSON generation and parsing.

https://docs.python.org/3/library/json.html#json.dump

Comments

2

Based on Michael Anderson's answer but also works when you pass in an array

class OrderedJsonEncoder(simplejson.JSONEncoder):
    def encode(self, o, first=True):
        if type(o) == list and first:
            return '[' + ",".join([self.encode(val, first=False) for val in o]) + ']'
        if type(o) == OrderedDict:
            return "{" + ",".join(
                [self.encode(k, first=False) + ":" + self.encode(v) for (k, v) in o.iteritems()]
            ) + "}"
        else:
            return simplejson.JSONEncoder.encode(self, o)

Comments

-3

Python 3.6.1:

Python 3.6.1 (default, Oct 10 2020, 20:16:48)
[GCC 7.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.

>>> import json
>>> json.dumps({'b': 1, 'a': 2})
'{"b": 1, "a": 2}'

Python 2.7.5:

Python 2.7.5 (default, Nov 20 2015, 02:00:19) 
    [GCC 4.8.5 20150623 (Red Hat 4.8.5-4)] on linux2
    Type "help", "copyright", "credits" or "license" for more information.

>>> import json
>>> json.dumps({'b': 1, 'a': 2})
'{"a": 2, "b": 1}'

Comments

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.