Python JSON serialize a Decimal object

2018-12-31 17:48发布

I have a Decimal('3.9') as part of an object, and wish to encode this to a JSON string which should look like {'x': 3.9}. I don't care about precision on the client side, so a float is fine.

Is there a good way to serialize this? JSONDecoder doesn't accept Decimal objects, and converting to a float beforehand yields {'x': 3.8999999999999999} which is wrong, and will be a big waste of bandwidth.

14条回答
牵手、夕阳
2楼-- · 2018-12-31 18:14

You can create a custom JSON encoder as per your requirement.

import json
from datetime import datetime, date
from time import time, struct_time, mktime
import decimal

class CustomJSONEncoder(json.JSONEncoder):
    def default(self, o):
        if isinstance(o, datetime):
            return str(o)
        if isinstance(o, date):
            return str(o)
        if isinstance(o, decimal.Decimal):
            return float(o)
        if isinstance(o, struct_time):
            return datetime.fromtimestamp(mktime(o))
        # Any other serializer if needed
        return super(CustomJSONEncoder, self).default(o)

The Decoder can be called like this,

import json
from decimal import Decimal
json.dumps({'x': Decimal('3.9')}, cls=CustomJSONEncoder)

and the output will be:

>>'{"x": 3.9}'
查看更多
春风洒进眼中
3楼-- · 2018-12-31 18:18

How about subclassing json.JSONEncoder?

class DecimalEncoder(json.JSONEncoder):
    def _iterencode(self, o, markers=None):
        if isinstance(o, decimal.Decimal):
            # wanted a simple yield str(o) in the next line,
            # but that would mean a yield on the line with super(...),
            # which wouldn't work (see my comment below), so...
            return (str(o) for o in [o])
        return super(DecimalEncoder, self)._iterencode(o, markers)

Then use it like so:

json.dumps({'x': decimal.Decimal('5.5')}, cls=DecimalEncoder)
查看更多
初与友歌
4楼-- · 2018-12-31 18:20

From the JSON Standard Document, as linked in json.org:

JSON is agnostic about the semantics of numbers. In any programming language, there can be a variety of number types of various capacities and complements, fixed or floating, binary or decimal. That can make interchange between different programming languages difficult. JSON instead offers only the representation of numbers that humans use: a sequence of digits. All programming languages know how to make sense of digit sequences even if they disagree on internal representations. That is enough to allow interchange.

So it's actually accurate to represent Decimals as numbers (rather than strings) in JSON. Bellow lies a possible solution to the problem.

Define a custom JSON encoder:

import json


class CustomJsonEncoder(json.JSONEncoder):

    def default(self, obj):
        if isinstance(obj, Decimal):
            return float(obj)
        return super(CustomJsonEncoder, self).default(obj)

Then use it when serializing your data:

json.dumps(data, cls=CustomJsonEncoder)

As noted from comments on the other answers, older versions of python might mess up the representation when converting to float, but that's not the case anymore.

To get that exact decimal back in Python:

Decimal(str(value))

This solution is hinted in Python 3.0 documentation on decimals:

To create a Decimal from a float, first convert it to a string.

查看更多
泪湿衣
5楼-- · 2018-12-31 18:23

This is what I have, extracted from our class

class CommonJSONEncoder(json.JSONEncoder):

    """
    Common JSON Encoder
    json.dumps(myString, cls=CommonJSONEncoder)
    """

    def default(self, obj):

        if isinstance(obj, decimal.Decimal):
            return {'type{decimal}': str(obj)}

class CommonJSONDecoder(json.JSONDecoder):

    """
    Common JSON Encoder
    json.loads(myString, cls=CommonJSONEncoder)
    """

    @classmethod
    def object_hook(cls, obj):
        for key in obj:
            if isinstance(key, six.string_types):
                if 'type{decimal}' == key:
                    try:
                        return decimal.Decimal(obj[key])
                    except:
                        pass

    def __init__(self, **kwargs):
        kwargs['object_hook'] = self.object_hook
        super(CommonJSONDecoder, self).__init__(**kwargs)

Which passes unittest:

def test_encode_and_decode_decimal(self):
    obj = Decimal('1.11')
    result = json.dumps(obj, cls=CommonJSONEncoder)
    self.assertTrue('type{decimal}' in result)
    new_obj = json.loads(result, cls=CommonJSONDecoder)
    self.assertEqual(new_obj, obj)

    obj = {'test': Decimal('1.11')}
    result = json.dumps(obj, cls=CommonJSONEncoder)
    self.assertTrue('type{decimal}' in result)
    new_obj = json.loads(result, cls=CommonJSONDecoder)
    self.assertEqual(new_obj, obj)

    obj = {'test': {'abc': Decimal('1.11')}}
    result = json.dumps(obj, cls=CommonJSONEncoder)
    self.assertTrue('type{decimal}' in result)
    new_obj = json.loads(result, cls=CommonJSONDecoder)
    self.assertEqual(new_obj, obj)
查看更多
零度萤火
6楼-- · 2018-12-31 18:24

My $.02!

I extend a bunch of the JSON encoder since I am serializing tons of data for my web server. Here's some nice code. Note that it's easily extendable to pretty much any data format you feel like and will reproduce 3.9 as "thing": 3.9

JSONEncoder_olddefault = json.JSONEncoder.default
def JSONEncoder_newdefault(self, o):
    if isinstance(o, UUID): return str(o)
    if isinstance(o, datetime): return str(o)
    if isinstance(o, time.struct_time): return datetime.fromtimestamp(time.mktime(o))
    if isinstance(o, decimal.Decimal): return str(o)
    return JSONEncoder_olddefault(self, o)
json.JSONEncoder.default = JSONEncoder_newdefault

Makes my life so much easier...

查看更多
若你有天会懂
7楼-- · 2018-12-31 18:27

I tried switching from simplejson to builtin json for GAE 2.7, and had issues with the decimal. If default returned str(o) there were quotes (because _iterencode calls _iterencode on the results of default), and float(o) would remove trailing 0.

If default returns an object of a class that inherits from float (or anything that calls repr without additional formatting) and has a custom __repr__ method, it seems to work like I want it to.

import json
from decimal import Decimal

class fakefloat(float):
    def __init__(self, value):
        self._value = value
    def __repr__(self):
        return str(self._value)

def defaultencode(o):
    if isinstance(o, Decimal):
        # Subclass float with custom repr?
        return fakefloat(o)
    raise TypeError(repr(o) + " is not JSON serializable")

json.dumps([10.20, "10.20", Decimal('10.20')], default=defaultencode)
'[10.2, "10.20", 10.20]'
查看更多
登录 后发表回答