At the time of creating this lib, there were (at least) three implementation of default_encoder()
copy-pasted from one place, evolving its own way and about to being copy-pasted further.
If you have read the story about AI and paperclips, you have an idea where this would lead us to.
To prevent this from happening, this library should unify all the implementation, and provide reusable implementation of the JSON encoding.
Add kiwi-json
into your requirements.in file
kiwi-json
If you use your own JSON encoder as a class, use default_encoder()
in there.
import simplejson
from kw.json import default_encoder, mask_dict
class OurJSONEncoder(simplejson.JSONEncoder):
def default(self, obj):
return default_encoder(obj, mask_dict)
kiwi-json
provides a simple implementation for masking dictionary values with kw.json.mask_dict
.
Or you can create a masking function for it by yourself. It supports customizing placeholder, blacklist and whitelist:
from kw.json import mask_dict_factory
mask_dict = mask_dict_factory(placeholder='0_0', blacklist={'secret'}, whitelist={'not-so-secret'})
If you want to use json.dumps
directly, you can do it the following way:
import simplejson
from kw.json import default_encoder
dumps = partial(simplejson.dumps, default=default_encoder)
If you have simplejson
installed, you can use Decimal as JSON number
type:
from decimal import Decimal
from kw.json import dumps, loads
assert dumps({"num": Decimal("1.234")}, use_decimal=True) == '{"num": 1.234}'
assert loads('{"num": 1.234}', use_decimal=True) == {"num": Decimal("1.234")}
Flask-based application could utilize the extension:
from kw.json.flask import JSONExtension
def create_app():
...
JSONExtension(app)
...
Extension will install an encoder to given app.
If you want to make sure that the encoder dumps classes, you can use the raw_encoder
:
from kw.json import raw_encoder, dumps
dumps(data, default=raw_encoder)
To dump dates and datetimes as unix time, use date_as_unix_time=True
:
import arrow
from datetime import datetime
from kw.json import dumps
dumps({1: datetime.now(), 2: arrow.now()}, date_as_unix_time=True)
If you want to combine the powers of date_as_unix_time
and raw_encoder
,
you can create your own encoder using partial:
from kw.json import dumps, raw_encoder
from functools import partial
my_encoder = partial(raw_encoder, date_as_unix_time=True)
dumps(obj, default=my_encoder)
To run the tests we use tox. Before you can run the tests please make sure you have postgres database running and the DATABASE_URI env variable set
export DATABASE_URI='postgres://[username]:[password]@[host]:[port]/[database]'
Once you have this set up just execute:
tox