tressed 0.0.4

Deserialize and serialize types, in any order

tressed - Deserialize and serialize types, in any order.

Tressed is a easy to use pure python library to deserialize and serialize to and from Python types.

---

Install tressed using pip:

$ pip install tressed

Now we can get going! For simple usecases, you can use the default loader and dumper directly:

>>> from tressed import load, dump
>>> from pprint import pprint
>>>
>>>
>>> value = load([1, [2, 2], [3, 4]], tuple[int, set[float], complex])
>>> pprint(value)
(1, {2.0}, (3+4j))
>>>
>>> pprint(dump(value))
[1, [2.0], [3.0, 4.0]]

For more advanced use cases instantiate a Loader and/or a Dumper:

>>> from dataclasses import dataclass, field
>>> from pprint import pprint
>>> 
>>> from tressed import Loader
>>> from tressed.alias import to_camel
>>> 
>>> @dataclass
... class SomeDataclass:
...     some_field: str
...     some_default_field: str = "bar"
...     some_default_factory_field: list[int] = field(default_factory=lambda: [1, 2, 3])
...     other_field: tuple[int, str] = field(metadata={"alias": "OTHER"}, kw_only=True)
...     
>>> raw_value = {
...     "someField": "foo",
...     "OTHER": (2, "humbug"),
... }
... 
>>> loader = Loader(alias_fn=to_camel)
>>> value = loader.load(raw_value, SomeDataclass)
>>> pprint(value)
SomeDataclass(some_field='foo',
              some_default_field='bar',
              some_default_factory_field=[1, 2, 3],
              other_field=(2, 'humbug'))
>>>
>>> from tressed import Dumper
>>> from tressed.alias import to_pascal
>>>
>>> dumper = Dumper(alias_fn=to_pascal)
>>> pprint(dumper.dump(value))
{'OTHER': [2, 'humbug'],
 'SomeDefaultFactoryField': [1, 2, 3],
 'SomeDefaultField': 'bar',
 'SomeField': 'foo'}

Features

Supported type forms

Tressed uses the proposed PEP 747 TypeForm combined with mypy 1.19 to provide accurate types for the Loader.load function.

The following type forms are supported out of the box:

• bool
• int
• float
• str
• complex
• tuple[T1, ..., Tn], typing.Tuple[T1, ..., Tn]
• tuple[T, ...], typing.Tuple[T, ...] (a.k.a homogeneous tuple)
• typing.NamedTuple
• list[T], typing.List[T]
• set[T], typing.Set[T]
• frozenset[T], typing.FrozenSet[T]
• dict[K, V], typing.Dict[K, V]
• typing.Literal[L1, .., Ln]
• typing.TypeAliasType
• T | None, typing.Optional[T]
• T1 | .. | Tn, typing.Union[T1, .., Tn] (Untagged Union)
• Annotated[T1 | ... | Tn, Discriminator(...)] (Discriminated Union)
• enum.Enum
• typing.TypedDict, typing_extensions.TypedDict (including support for PEP 728 closed and extra_items)
• dataclasses.dataclass
• typing.NewType[T]
• ipaddress.{IPv4Address, IPv6Address, IPv4Interface, IPv6Interface, IPv4Network, IPv6Network}
• uuid.UUID
• pathlib.Path
• datetime.{date,datetime,time} (ISO 8601 format, using datetime.{date,datetime,time}.fromisoformat)

It is easy to add support for custom types as needed when creating a loader.

Supported source types

Data can be loaded from the following basic types, matching types used by common serialization formats.

Additionally data can be loaded from argparse namespaces to support easily loading arguments into data types for simple CLI applications.

• int
• float
• bool
• str
• list
• dict
• argparse.Namespace

Installation

Using pip:

Or using uv:

uv add tressed

Then just import tressed and get coding.

import tressed

Goals

• Provide easy serialization and deserialization to and from built-in and standard library types.
• Provide easy way to add support for custom types.
• Provide easy yet powerful support for serializaiton and deserialization aliases.
• Pure python, no additional runtime dependencies beyond the standard library.
• Do as little work as possible to be suitable for CLIs, for example using lazy imports as much as possible.

Stretch goals

• Add first class support for concurrent loading and dumping, with support for asyncio and threading, for example deserializing a list of M objects on N threads.
• Use code generation internally similarly to dataclasses. Instead of doing metaprogramming each time, generate an actual function for the type that does (de)serialization. Interpret the first time, compile the second, using source or ast module, and then reuse the compiled function.