Grave Settings

A library for automatic serialization of python object hierarchies to storage. Most python objects are compatible without any extra code, and there are several tools for fine-grain control.

The framework uses either duck typing or strong typing to allow objects define custom behaviors and defines a Formatter class to allow for user defined input/output formats

Pre-defined formats:

• json (via built in json module)
• toml (read via tomlib, write enabled when tomli-w module is installed)
• bson (only tested experimentally)

Red the documenation for examples and important considerations.

Install

pip install grave-settings

Features

• Save / Import types to reconstruct object hierarchies
• Preserve "is" relationships automatically by scanning object ids
• Abstractions for object version management for easing continuous deployment
• OrderedHandler objects allow for custom serialization / deserialization logic for types without using inheritance or duck typing
• Detect circular references
• Event handlers for finalizing serialization / deserialization (intended to fix circular references)
• Semantic objects for communicating options to the formatter (ex. security options for loading of arbitrary types)

Default standard handlers

Defining new handlers or adding functionality to the default handlers is easy, but some types have already been done:

Name	Description
Type	built-in python type object
NoneType	None
Iterable	General catch all for Iterable defined in collections module
Mapping	General catch all for Mapping defined in collections module
FunctionType	Python user-defined function
date	from datetime module
datetime	from datetime module (experimental timezone support)
timedelta	from datetime module
Enum	from enum module
partial	builtin partial class
bytes	builtin bytes data
Complex / Rational	builtin numerical classes
Path	from pathlib

There is still a ways to go before most of the built in types have handlers. To see how the handlers work read: default_handlers.py

In progress

• Validation support
• Automated Qt GUI interface for editing settings objects

Code Example

Here is a quick example that serializes and then deserializes a custom object hierarchy with the default built in classes

from datetime import date

from grave_settings.formatters.toml import TomlFormatter


class Color:
    def __init__(self, r, g, b):
        self.r = r
        self.g = g
        self.b = b

    def __str__(self):
        return f'Color(r={self.r}, g={self.g}, b={self.b})'


class MyObject:
    def __init__(self):
        self.integer = 1
        self.string = 'b'
        self.function = print  # demonstrates handling a function
        self.type_object = int  # demonstrates handling a type
        # demonstrates handling a list of date objects
        self.dates = [date(year=2022, month=1, day=1), date(year=2023, month=2, day=2)]
        # demonstrates handling a nested object
        self.color = Color(255, 255, 255)

formatter = TomlFormatter()
obj_str = formatter.dumps(MyObject())
print(obj_str)

This will output:

__class__ = "__main__.MyObject"
integer = 1
string = "b"

[function]
__class__ = "types.FunctionType"
state = "builtins.print"

[type_object]
__class__ = "builtins.type"
state = "builtins.int"

[[dates]]
__class__ = "datetime.date"
state = [
    2022,
    1,
    1,
]

[[dates]]
__class__ = "datetime.date"
state = [
    2023,
    2,
    2,
]

[color]
__class__ = "__main__.Color"
r = 255
g = 255
b = 255

Now if we take this string and deserialize it with the same (or equivalent) formatter:

from grave_settings.formatters.toml import TomlFormatter
formatter = TomlFormatter()
remade_obj = formatter.loads(obj_str)
print(f'''
integer = {remade_obj.integer}
string = {remade_obj.string}
function = {remade_obj.function}
type_object = {remade_obj.type_object}
dates = {remade_obj.dates}
color = {remade_obj.color}
''')

We get:

integer = 1
string = b
function = <built-in function print>
type_object = <class 'int'>
dates = [datetime.date(2022, 1, 1), datetime.date(2023, 2, 2)]
color = Color(r=255, g=255, b=255)

It is recommended to use the Serializable type (in this module) or handlers instead of using the naked objects like above. Read the documentation for more on this.