Kylie provides mappings between JSON data structures and Python objects. It provides a reasonable amount of power with only a tiny bit of magic, and it has 100% code coverage.
- Free software: BSD license
- Documentation: https://kylie.readthedocs.org/
- Allows name mapping between Models and python dictionary keys.
- Allows type conversion when serializing and de-serializing objects.
- Automatic serialization/deserialization of nested models.
- Not bound to JSON in any way, and should also be useful for MessagePack
- Supports Python 2.6+ & 3.3+
class SpanishInquisitionModel(Model):
inquisition_id = Attribute('id')
expected = Attribute(python_type=bool, serialized_type=int)
Then:
>>> surprise = SpanishInquisitionModel(inquisition_id=1234, expected=False)
>>> surprise.inquisition_id
1234
>>> surprise.serialize()
{'id': 1234, expected=0}
Note that the attribute inquisition_id
becomes the dict key "id"
, and
expected is mapped to 0
instead of False
.
We can now take this dict, dumps
it to JSON, and somewhere else call
the following on the json_data (which is a dict returned from loads
):
>>> my_surprise = SpanishInquisitionModel.deserialize(json_data)
>>> my_surprise.inquisition_id
1234
>>> my_surprise.expected
False
Kylie supports nested models, so you can embed other Model instances inside the data, and Kylie will manage serialization and deserialization of them for you.
So what doesn't Kylie do yet? Well, there are a few things, because it's very new:
- Doesn't have any mechanism for validation. I plan to add this once I decide the best way to do it. Ideas welcome!
- No post-deserialize option, unless you do it yourself.
This would allow wiring up of objects that are referred to by
id
and provided elsewhere in the serialized data-structure, for example.
So, still some important things to do, but I think Kylie is already useful.
Back in the late 80's (I'm old!) Kylie and Jason were today's Kim and Kanye. This Kylie works well with JSON. Geddit?