jsonbourne

jsonbourne

Install: pip install jsonbourne

• Python json lib/pkg that makes json feel like the JSON module in javascript/typescript:
  • from jsonbourne import JSON; JSON.parse(JSON.stringify({"key": "value"}))
  • Automatically uses best json-lib-backend available (orjson/python-rapidjson) ~ can be configured
• Hybrid dict/class object (jsonbourne.JsonObj):
  • Dot-notation getting/setting (featuring protected attributes!)
  • All your favorite python dictionary methods (items, keys, update, values) and more!
  • Works with pydantic and attrs
• FastAPI:
  • JSONBOURNEResponse ~ auto use the best
• No hard dependencies ~ works with python-stdlib-json as well as orjson and python-rapidjson
• jsonbourne.JsonObj uses list/dict comprehensions (some are recursive) everywhere because 'why not?' and it is a bit faster

Usage:

JSON ~ from jsonbourne import JSON

Importing:

# Importing JSON:
from jsonbourne import JSON

# or
import JSON

# Importing jsonbourne:
import jsonbourne
import david_webb  # jsonbourne's `True` identity

JSON basics:

import JSON  # Module included with jsonbourne

string_stringify = JSON.stringify(
    {"a": 1, "b": 2, "c": 3}
)  # '{"a": 1, "b": 2, "c": 3}'
string_dumps = JSON.dumps({"a": 1, "b": 2, "c": 3})  # '{"a": 1, "b": 2, "c": 3}'
string_dumps

'{"a":1,"b":2,"c":3}'

JSON option kwargs ~ pretty & sort_keys

pretty:

string_dumps = JSON.stringify(
    {"b": 2, "a": 1, "c": 3}, pretty=True
)  # '{"a": 1, "b": 2, "c": 3}'
print(string_dumps)

{
  "b": 2,
  "a": 1,
  "c": 3
}

sort_keys:

string_dumps = JSON.stringify(
    {"b": 2, "a": 1, "c": 3}, pretty=True, sort_keys=True
)  # '{"a": 1, "b": 2, "c": 3}'
print(string_dumps)

{
  "a": 1,
  "b": 2,
  "c": 3
}

JsonObj & JSON

• Python dictionary/object with dot access
• Protections against setting class/obj attributes
• Is as javascript-y as possible (keys have to be strings -- ints/floats will be converted to strings)
• Create a jsonbourne.JsonObj with jsonbourne.JSON
• Recursive jsonification
• Allows for kwarging (**json_obj)
• Works with pydantic and attrs

Make an empty JsonObj

The following 3 examples all produce the same thing

from jsonbourne import JSON
j = JSON()  # j => JsonObj(**{})
# OR
import JSON
j = JSON()  # j => JsonObj(**{})
# OR
from jsonbourne import JsonObj
j = JsonObj()  # j => JsonObj(**{})

From a dictionary o data

import datetime

data = {
    "key": "value",
    "list": [1, 2, 3, 4, 5],
    "dt": datetime.datetime(1970, 1, 1, 0, 0, 0, 1),
    "sub": {
        "b": 3,
        "key": "val",
        "a": 1,
    },
    "timedelta": datetime.timedelta(days=2),
}

JSON(data)

JsonObj(**{
    'dt': datetime.datetime(1970, 1, 1, 0, 0, 0, 1),
    'key': 'value',
    'list': [1, 2, 3, 4, 5],
    'sub': {'a': 1, 'b': 3, 'key': 'val'},
    'timedelta': datetime.timedelta(days=2)
})

Dot access

JSON(data).sub.b

3

stringified_data = JSON(data).stringify(pretty=True)
print(stringified_data)

{
  "key": "value",
  "list": [
    1,
    2,
    3,
    4,
    5
  ],
  "dt": "1970-01-01T00:00:00.000001",
  "sub": {
    "b": 3,
    "key": "val",
    "a": 1
  },
  "timedelta": 172800.0
}

parsed_data = JSON(stringified_data)
parsed_data

JsonObj(**{
    'dt': '1970-01-01T00:00:00.000001',
    'key': 'value',
    'list': [1, 2, 3, 4, 5],
    'sub': {'a': 1, 'b': 3, 'key': 'val'},
    'timedelta': 172800.0
})

list(parsed_data.keys())

['key', 'list', 'dt', 'sub', 'timedelta']

list(parsed_data.items())

[('key', 'value'),
 ('list', [1, 2, 3, 4, 5]),
 ('dt', '1970-01-01T00:00:00.000001'),
 ('sub', JsonObj(**{'b': 3, 'key': 'val', 'a': 1})),
 ('timedelta', 172800.0)]

list(parsed_data.dot_keys())

[('key',),
 ('list',),
 ('dt',),
 ('sub', 'b'),
 ('sub', 'key'),
 ('sub', 'a'),
 ('timedelta',)]

list(parsed_data.dot_items())

[(('key',), 'value'),
 (('list',), [1, 2, 3, 4, 5]),
 (('dt',), '1970-01-01T00:00:00.000001'),
 (('sub', 'b'), 3),
 (('sub', 'key'), 'val'),
 (('sub', 'a'), 1),
 (('timedelta',), 172800.0)]

parsed_data[("sub", "key")]

'val'

parsed_data.dot_lookup("sub.key")

'val'

{**parsed_data}

{'key': 'value',
 'list': [1, 2, 3, 4, 5],
 'dt': '1970-01-01T00:00:00.000001',
 'sub': JsonObj(**{'b': 3, 'key': 'val', 'a': 1}),
 'timedelta': 172800.0}

# fully eject
parsed_data.eject()

{'key': 'value',
 'list': [1, 2, 3, 4, 5],
 'dt': '1970-01-01T00:00:00.000001',
 'sub': {'b': 3, 'key': 'val', 'a': 1},
 'timedelta': 172800.0}

Protected keys

jsonbourne.JsonObj protects against setting attributes like 'items' through dot-notation.

from jsonbourne import JSON

j = JSON()
j.key = "value"
try:  # CANNOT set 'items' using dot-access
    j.items = [1, 2, 3, 4]
except ValueError:
    pass
# CAN set 'items' through key/item access
j["items"] = [1, 2, 3, 4]
print(j.__dict__)
print(j)
j_items = j.items
print("items", j_items)
# Getting 'items' through dot-access returns the `items()` method
assert j.items != [1, 2, 3, 4]
# Getting 'items' with key-access returns the stored value
assert j["items"] == [1, 2, 3, 4]

{'_data': {'key': 'value', 'items': [1, 2, 3, 4]}}
JsonObj(**{
    'items': [1, 2, 3, 4], 'key': 'value'
})
items <bound method JsonObj.items of JsonObj(**{'key': 'value', 'items': [1, 2, 3, 4]})>

pydantic & jsonbourne

• from jsonbourne.pydantic import JsonBaseModel
• Allows for aliases when getting/setting attribute(s)
• Supports __post_init__ (like dataclasses)

Basic usage:

from jsonbourne import JsonObj
from jsonbourne.pydantic import JsonBaseModel


class JsonSubObj(JsonBaseModel):
    herm: int

    def to_dict(self):
        return self.dict()

    def to_json(self, *args, **kwargs):
        return self.json()

    @classmethod
    def from_json(cls, json_string: str):
        return JsonSubObj(json.loads(json_string))


class JsonObjModel(JsonBaseModel):
    a: int
    b: int
    c: str
    d: JsonObj
    e: JsonSubObj

    #
    @property
    def a_property(self) -> str:
        return "prop_value"

    def to_json(self, *args, **kwargs):
        return self.json()

    @classmethod
    def from_json(cls, json_string: str):
        return cls(**json.loads(json_string))


obj = JsonObjModel(
    **{"a": 1, "b": 2, "c": "herm", "d": {"nested": "nestedval"}, "e": {"herm": 2}}
)
obj

JsonObjModel(**{
     'a': 1,
     'b': 2,
     'c': 'herm',
     'd': JsonObj(**{'nested': 'nestedval'}),
     'e': {'herm': 2}
})

# respects properties (which I don't think pydantic does(currently))
obj.a_property

'prop_value'

---

JSON backend/lib

jsonbourne finds the best json-lib (python-rapidjson/orjson) it can! On import jsonbourne gets to work spying on your python env. jsonbourne, the most highly qualified json-CIA-agent, will import the best python-json library it can find; if jsonbourne's cover is blown (meaning: the only json library found is the python stdlib json), jsonbourne will fallback to the python stdlib json.

jsonbourne will look for the following json-packages in the following order:

1. rapidjson
2. orjson

Custom lib preferences

from jsonbourne import import_json

json = import_json(("rapidjson", "orjson"))  # prefer rapidjson over orjson
string = json.dumps({"a": 1, "b": 2, "c": 3})
print(json)
print(string)

<class 'jsonbourne.jsonlib.RAPIDJSON'>
{"a":1,"b":2,"c":3}

Installing better JSON lib:

orjson

• pip install orjson [pip]

rapidjson/python-rapidjson

• pip install python-rapidjson [pip]
• conda install -c anaconda python-rapidjson [conda anaconda/defaults]
• conda install -c conda-forge python-rapidjson [conda-forge]