Dictionary deserializer is a package that aides in the deserializing of JSON (or other structures) that are converted to dicts, into composite classes.

# Dictionary deserializer

Dictionary deserializer is a project built to convert dictionaries into composite classes in an intuitive way. Special attention was also paid to being friendly to static type-checkers and IDE autocompletes.

It is expected that this library is used together with a JSON-to-dict deserializer like json.loads.

## Design

This project was originally meant as a proof of concept, to be used to find other projects that would be able to replace this, with the required feature set. That project was not found, and therefore, this project was expanded.

### Requirements

• Use type hints for type validation
• Allow polymorphism
• Through typing.Unions
• Through subclassing
• Support a large part of the typing module's types
• Allow validations on values
• Be able to validate and deserialize any compliant JSON structure
• Be compatible with static type checkers and IDE hinting
• Have a small impact on existing code starting to use this library

## Examples

None of this code is actually useful if you don't understand how to use it. It is very simple. Here are some examples:

### Specifying a structure

from typing import Optional

from dict_deserializer.deserializer import Deserializable

class User(Deserializable):
email: str                  # Type must be a string
username: str               # Type must be a string
password: Optional[str]     # Type must either be a string or a None


### Deserialization

from dict_deserializer.deserializer import deserialize, Rule

# Successful
deserialize(Rule(User), {
'email': 'pypi@rolfvankleef.nl',
})

# Fails because optional type is wrong
deserialize(Rule(User), {
'email': 'pypi@rolfvankleef.nl',
})


### Polymorphic structures

from typing import Optional, Any, List

from dict_deserializer.deserializer import Deserializable
from dict_deserializer.annotations import abstract

@abstract
class DirectoryObject(Deserializable):
name: str
meta: Any

class User(DirectoryObject):
full_name: str
first_name: Optional[str]

class Group(DirectoryObject):
members: List[DirectoryObject]


If you deserialize into Rule(DirectoryObject), the matching class will automatically be selected. If none of the subclasses match, an error is thrown since the DirectoryObject is declared abstract.

If you want to discriminate not by field names or types, but by their values, one can choose to define a @discriminator annotation.

### Value validations

The syntax for validating the value of a key is currently a bit weird. It is incompatible with existing syntax for defaults, but the type syntax is the same.

Example:

from typing import Optional

from dict_deserializer.deserializer import Deserializable
from dict_deserializer.annotations import validated

class Test(Deserializable):
name: Optional[str]

@validated(default='Unknown')
def name(self, value):
if len(value) > 20:
raise TypeError('Name may not be longer than 20 characters.')


## Limitations

This library uses the typing module extensively. It does, however, only support some of its types. This is a list of verified composite types:

• Union (Including Optional)
• List
• Tuple
• Any
• dict_deserializer.deserializer.Deserializable
• dict
• list

It supports these types as terminal types:

• int
• float
• str
• NoneType
• bool

## Planned features

• NamedTuples
• The anonymous namedtuple and the class-namedtuples with (optionally) type annotations.
• Dataclasses
• A way to allow deserializing into a class not extending Deserializable
• Enums
• Sets
• From lists

## Project details

Uploaded source
Uploaded py3