Skip to main content

A single-file utility to allow better runtime handling of types by providing a predictable way of transforming data into the shape of a given type hint.

Project description

typewire

A single-file utility to allow better runtime handling of types by providing a predictable way of transforming data into the shape of a given type hint.

Why?

Python's standard library provides tools for describing types via type hints, but it doesn't provide a unified way of actually enforcing those type hints at runtime, even buried deep in the typing module.

Our goal is to allow for x: T to behave transparently as usual whilst also allowing the user to convert to that type, regardless of whether that is x: int or x: float | None or x: dict[str, list[int | dict[float, User]]]. Just like int(x) will (to the best of its ability) turn x into an int, typewire.as_type(x, int) will do the same, but with the added benefit of working on type hints that aren't callable (like list[float]).

>>> from typewire import as_type

>>> as_type("3.2", float)
3.2

>>> as_type(78.1, int)
78

>>> as_type("3.2", int, transparent_int=True)
3

>>> as_type(["13.2", "18", "-1.2"], list[int | float])
[13.2, 18, -1.2]

>>> as_type([("a", "1"), ("b", "2"), ("c", "3"), ("z", "26")], dict[str, float])
{'a': 1.0, 'b': 2.0, 'c': 3.0, 'z': 26.0}

>>> from pathlib import Path
>>> data = {"logs": ["/tmp/app.log", "123"]}
>>> hint = dict[str, list[Path | int]]
>>> as_type(data, hint)
{'logs': [Path('/tmp/app.log'), 123]}

Installation

typewire is supported on Python 3.10 and onward and can be easily installed with a package manager such as:

# using pip
$ pip install typewire

# using uv
$ uv add typewire

typewire does not have any additional dependencies.

Documentation

TypeHint

TypeHint is provided as a top-level alias for typing.Any.

is_union, is_mapping, is_iterable

These three functions check whether a given type hint is a union type (e.g., int | str | bytes), a mapping type (e.g., dict[str, Any]), or an iterable type (e.g., list[str]).

Note that is_iterable specifically excludes str and bytes: is_iterable(str) == False as, while str does support iteration, for the purposes of type casting, it's not really an iterable/container type.

as_type

The signature is

def as_type(value: Any, to: TypeHint, *, transparent_int: bool = False, semantic_bool: bool = False) -> Any:
  ...

In particular, it casts the given value to the given to type, regardless of whether to is:

# a plain type
>>> as_type(3.2, int)
3

# typing.Literal, returning the value as-is if it's a valid entry
>>> as_type("abc", Literal["abc", "def"])
'abc'

>>> as_type("80", Literal[80, 443])
ValueError(...)

# a union type, casting to the first valid type
>>> as_type("3", float | int)
3.0

>>> as_type("3", int | float)
3

# an optional type
>>> as_type(43, int | None)
43

>>> as_type(None, int | None)
None

# a mapping type
>>> as_type({"a": "1", "b": "2.0"}, dict[str, float])
{'a': 1.0, 'b': 2.0}

# a container/iterable type
>>> as_type([1.2, -3, 449], list[str])
['1.2', '-3', '449']

>>> as_type([1.2, -3, 449], tuple[str, ...])
('1.2', '-3', '449')

# even if it's just a blank generic continer: it'll act like T[Any]
>>> as_type(["a", 3.2, None, [1, "a"]], tuple)
('a', 3.2, None, [1, 'a'])

# typing.Annotated, treating it as the bare type
>>> as_type("3", Annotated[int, "some metadata"])
3

# a typing.NewType, treating it as the supertype
>>> UserId = NewType("UserId", int)
>>> as_type("3", UserId)
3
>>> type(as_type("3", UserId))  # unfortunately, the UserId type doesn't exist at runtime
<class 'int'>

# an abstract collections.abc.Iterable/Mapping, cast as concrete list/dict
>>> as_type([1.2, -3, 449], Iterable[str])
['1.2', '-3', '449']

>>> as_type({"a": "1", "b": "2.0"}, Mapping[str, float])
{'a': 1.0, 'b': 2.0}

# ...unless it's a string being cast as Iterable[str]
>>> as_type("hello world", Iterable[str])
'hello world'

On a failure, ValueError is raised.

transparent_int

This flag (default = False) allows for a nonstrict cast to int.

>>> int("3.2")
ValueError # invalid literal for int() with base 10: '3.2'

>>> as_type("3.2", int)
ValueError # invalid literal for int() with base 10: '3.2'

>>> as_type("3.2", int, transparent_int = True)
3

In practice, this flag results in a call of int(float(value)) instead of just int(value).

semantic_bool

This flag (default = False) allows for a nonstrict cast to bool.

>>> bool("false")  # non-empty string
True

>>> as_type("false", bool)
True

>>> as_type("false", bool, semantic_bool = True)
False

In practice, if value is a string and is one of ["false", "no", "0", "off"] (case-insensitive), then it will be cast as False with this flag enabled.

unwrap

unwrap recursively removes Annotated, NewType, and Union layers, returning a list of component types. Note that it does not unwrap other containers, such as list. This is because unwrap is working to identify what the type "is", rather to find all of the structural components. From that perspective, list[T] is itself a leaf: the type represents a list.

# bare types are just themselves
>>> unwrap(int)
[int]

>>> unwrap(list[int])
[list[int]]

# note that None is interpreted to NoneType, i.e., type(None)
>>> unwrap(None)
[<class 'NoneType'>]

# union types return their components in order
>>> unwrap(int | str)
[int, str]
>>> unwrap(str | int)
[str, int]

# resulting types are deduplicated
>>> unwrap(int | str | int)
[int, str]

# also works with old-style Optional
>>> unwrap(Optional[int])
[int, <class 'NoneType'>]

# annotated layer gets removed
>>> unwrap(Annotated[int, "some metadata"])
[int]

# NewType gets unwrapped, returning the supertype
>>> UserId = NewType("UserId", int)
>>> unwrap(UserId)
[int]

# unwrap can handle matroyska dolls of nesting
# resulting order is depth-first
>>> T1, T2, T3 = TypeVar("T1"), TypeVar("T2"), TypeVar("T3")
>>> D1 = NewType("D1", T1)
>>> matroyska_doll = Optional[Annotated[T1 | Annotated[T2 | D1 | None | Annotated[T3 | T1, "level 2"], "level 1"], "level 2"]]
>>> unwrap(matroyska_doll)
[T1, T2, <class 'NoneType'>, T3]  # D1 -> T1, so it doesn't appear in the result, nor does the final T1

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

typewire-1.1.0.tar.gz (6.5 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

typewire-1.1.0-py3-none-any.whl (7.5 kB view details)

Uploaded Python 3

File details

Details for the file typewire-1.1.0.tar.gz.

File metadata

  • Download URL: typewire-1.1.0.tar.gz
  • Upload date:
  • Size: 6.5 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.9.22 {"installer":{"name":"uv","version":"0.9.22","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Manjaro Linux","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

File hashes

Hashes for typewire-1.1.0.tar.gz
Algorithm Hash digest
SHA256 3831c9129ad939b06ed7eca346b852abcfc7c8bb6e55742c93273e1f826597e8
MD5 0937329c67aa1caca37940423992800f
BLAKE2b-256 1cbe923c521391f3cdb76d6434e3792c35126156c0743e75f5ace78161e19099

See more details on using hashes here.

File details

Details for the file typewire-1.1.0-py3-none-any.whl.

File metadata

  • Download URL: typewire-1.1.0-py3-none-any.whl
  • Upload date:
  • Size: 7.5 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: uv/0.9.22 {"installer":{"name":"uv","version":"0.9.22","subcommand":["publish"]},"python":null,"implementation":{"name":null,"version":null},"distro":{"name":"Manjaro Linux","version":null,"id":null,"libc":null},"system":{"name":null,"release":null},"cpu":null,"openssl_version":null,"setuptools_version":null,"rustc_version":null,"ci":null}

File hashes

Hashes for typewire-1.1.0-py3-none-any.whl
Algorithm Hash digest
SHA256 ec7e71b1e44b729490d919a267c7251357f6ddd18c28360e60e3e26d184291f7
MD5 9a64a2ed5efd8b0852dbcfcdee38ae41
BLAKE2b-256 cbbfa14280eaf71c3de129edad2a2599026c423ca6b45c3fc296c00ab005a11c

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page