Define your JSON schema as Python dataclasses
It's kinda like Pydantic but better.
Now with JSON-schema generation!
pip install nvelope
With nvelope
you can define dataclasses which know how to convert themselves from/to JSON.
All with custom checks and custom defined conversions from/to JSON for any type you want to put into your dataclass.
This library was designed with extensibility in mind, so it relies on interfaces (for the most part) rather than some weird inheritance stuff and other magic.
You can (and probably should) take a look at the code! The code base is microscopic compared to Pydantic.
Say you have a JSON representing a user in your app looking something like this
{
"id": 530716139,
"username": "johndoe",
"language_code": "en"
}
You define an envelope for it
from dataclasses import dataclass
from nvelope import (Obj, int_conv, string_conv)
@dataclass # note the @dataclass decorator, it is important
class User(Obj):
_conversion = {
"id": int_conv,
"language_code": string_conv,
"username": string_conv,
}
id: int
language_code: str
username: str
Now you have a model that knows how to read data from the JSON
(not the raw string, actually, but to the types that are allowed by the
standard json.dumps
function e.g. dict
, list
, str
, int
, float
, bool
, None
) ...
user = User.from_json(
{
"id": 530716139,
"username": "johndoe",
"language_code": "en"
}
)
... and knows how to convert itself into JSON
User(
id=530716139,
username="johndoe",
language_code="en",
).as_json()
# returns a dictionary {
# "id": 530716139,
# "username": "johndoe",
# "language_code": "en"
# }
You can also define compound envelopes.
Say we want to define a message and include info about the sender.
Having defined the User
envelope, we can do it like this:
from nvelope import CompoundConv
@dataclass
class Message(Obj):
_conversion = {
"message_id": int_conv,
"from_": CompoundConv(User),
"text": string_conv,
}
from_: User
text: str
message_id: int
and use it the same way:
# reading an obj from parsed json like this
Message.from_json(
{
"message_id": 44,
"text": "hello there",
"from_": {
"id": 530716139,
"username": "johndoe",
"language_code": "en"
}
}
)
# and dumping an object to json like this
import json
json.dumps(
Message(
message_id=44,
text="whatever",
from_=User(
id=530716139,
username="johndoe",
language_code="en",
)
).as_json()
)
This is how you define arrays:
from nvelope import Arr, CompoundConv
class Users(Arr):
conversion = CompoundConv(User)
# Same API inherited from nvelope.Compound interface
Users.from_json(
[
{
"id": 530716139,
"username": "johndoe",
"language_code": "en",
},
{
"id": 452341341,
"username": "ivandrago",
"language_code": "ru",
}
]
)
Users(
[
User(
id=530716139,
username="johndoe",
language_code="en",
),
User(
id=452341341,
username="ivandrago",
language_code="ru",
),
]
).as_json()
At some point you may need to define an envelope for an API containing certain field names which cannot be
used in python since they are reserved keywords (such as def
, from
, etc.).
There's a solution for this:
from dataclasses import dataclass
from nvelope import Obj, string_conv, CompoundConv, AliasTable
@dataclass
class Comment(Obj):
_conversion = {
"text": string_conv,
"from_": CompoundConv(User),
}
_alias_table = AliasTable({"from_": "from"})
text: str
from_: User
In this case from
key gets replaced by from_
in the python model.
The from_
field gets translated back to from
when calling .as_json()
There's a difference between fields that can be set to None
and fields which may be missing in the JSON at all.
This is how you specify that a some field may be missing from the JSON and that's OK:
from dataclasses import dataclass
from typing import Optional
from nvelope import MaybeMissing, Obj, OptionalConv, AliasTable
@dataclass
class Comment(Obj):
_alias_table = AliasTable(
{"from_": "from"}
)
text: str
img: Optional[str] # this field can be set to None (null), but is must always be present in the JSON
from_: MaybeMissing[User] # this field can be missing from JSON body
_conversion = {
"text": string_conv,
"img": OptionalConv(string_conv), # note the wrapping with OptionalConv
"from_": CompoundConv(User),
}
This is how you check if the MaybeMissing
field is actually missing
comment.from_.has() # returns False if the field is missing
and this is how you get the value:
comment.value() # raises an error if there's no value,
# so it is recommended to check the output of .has()
# before calling .value()
The Comment
model from we have defined generates schema like this:
Comment.schema()
with the returned schema looking like this:
{
"type": "object",
"properties": {
"from": {
"properties": {
"id": {"type": "integer"},
"language_code": {"type": "string"},
"username": {"type": "string"},
},
"required": ["id", "language_code", "username"],
"type": "object",
},
"img": {"type": ["string", "null"]},
"text": {"type": "string"},
},
"required": ["text", "img"],
}
NOTE: nvelope
does not perform json schema checks.
You may define a custom conversions inheriting from nvelope.nvelope.Conversion
abstract base class
or using nvelope.nvelope.ConversionOf
class.
For example, this is how datetime_iso_format_conv
is defined:
from nvelope import WithTypeCheckOnDump, ConversionOf
datetime_iso_format_conv = WithTypeCheckOnDump(
datetime.datetime,
ConversionOf(
to_json=lambda v: v.isoformat(),
from_json=lambda s: datetime.datetime.fromisoformat(s),
),
)
Say we want to jsonify a datetime
field as POSIX timestamp, instead of storing it in ISO string format.
datetime_timestamp_conv = ConversionOf(
to_json=lambda v: v.timestamp(),
from_json=lambda s: datetime.datetime.fromtimestamp(s),
schema={"type": "number"},
)
We could also add WithTypeCheckOnDump
wrapper in order to add explicit check that
the value passed into .from_json()
is indeed float
.
from nvelope import ConversionOf
datetime_timestamp_conv = WithTypeCheckOnDump(
float,
ConversionOf(
to_json=lambda v: v.timestamp(),
from_json=lambda s: datetime.datetime.fromtimestamp(s),
schema={"type": "number"},
)
)
You may also go further and implement custom conversion.
Inherit from nvelope.Conversion
interface, implement its abstract methods, and you are good to go.
You can also define custom alternatives to nvelope.Obj
and nvelope.Arr
.
It will work fine as long as they inherit nvelope.Compound
interface.
It currently required 3 methods:
from_json
as_json
schema