jayqi/typenames

String representations of Python type annotations

typenames : String representations of type annotations

typenames is a configurable Python library for creating string representations of type annotations. By default, it produces compact representations by removing standard library module names. Configurable options include standardizing on | operator syntax for unions or standard collections classes for generics.

import typing
from typenames import typenames

typenames(int)
#> 'int'
typenames(dict[str, typing.Any])
#> 'dict[str, Any]'
typenames(str | int)
#> 'str | int'
typenames(typing.Optional[str])
#> 'Optional[str]'

Why use this library?

String representations of Python type objects, type aliases, and special typing forms are inconsistent and often verbose. Here are some comparisons using default settings against built-in string representations:

Input	With str(...)	With typenames(...)
int	<class 'int'>	int
list	<class 'list'>	list
typing.Optional[int]	typing.Optional[int]	Optional[int]
collections.abc.Iterator[typing.Any]	collections.abc.Iterator[typing.Any]	Iterator[Any]
typing.Literal[MyEnum.NAME]	typing.Literal[<MyEnum.NAME: 'value'>]	Literal[MyEnum.NAME]

typenames also has handy configurable functionality, such as:

• Forcing standardization on | operator union syntax (e.g., Union[int, str] to int | str) or vice versa
• Forcing standardization on | operator optional syntax (e.g., Optional[int] to int | None) or vice versa
• Forcing standardization on standard collection types for generics (e.g., List[int] to list[int]) or vice versa
• Controlling exactly which module names to remove using regex patterns.

No need for string manipulation to get what you want!

Installation

typenames is available on PyPI:

pip install typenames

Basic Usage

The main way to use the library is the typenames function. Calling it on a type annotation renders a string representation:

import typing
from typenames import typenames

typenames(int)
#> 'int'
typenames(typing.Optional[str])
#> 'Optional[str]'
typenames(collections.abc.Callable[[int], tuple[str, ...]])
#> 'Callable[[int], tuple[str, ...]]

Under the hood, typenames parses a type annotation as a tree structure. If you need to see the parsed tree, use the parse_type_tree function to return the root node. You can get the rendered string representation by calling str(...) on root node.

import typing
from typenames import parse_type_tree

tree = parse_type_tree(typing.Union[typing.Any, list[typing.Any]])
tree
#> <GenericNode typing.Union[<TypeNode typing.Any>, <GenericNode <class 'list'>[<TypeNode typing.Any>]>]>
str(tree)
#> 'Union[Any, list[Any]]'

Configurable options

All configuration options can be passed as keyword arguments to either the typenames or parse_type_tree functions.

Union Syntax (union_syntax)

This option controls how unions are rendered. It supports both the typing.Union special form and the | operator (bitwise or) syntax from PEP 604. Valid options are defined by the enum UnionSyntax and include:

• "as_given" (default): render the union as it is given without changing syntax.
• "or_operator": render all type unions using the | operator.
• "special_form": render all type unions using the typing.Union special form.

Note that runtime use of the | operator between types is new in Python 3.10. To use in earlier versions of Python, you will need to use postponed evaluation of annotations à la PEP 563 with from __future__ import__annotations__. Support for the | operator is only a limitation on providing type annotation inputs to typenames, and not a limitation on output rendering.

Limitations: Python automatically flattens unions when evaluating them at runtime. Since typenames uses runtime type objects, it will only see the flattened result and not know if your original input was nested. Furthermore, any mixing of | operator syntax and any typing module types will result in a typing.Union union, so as_given will always render such inputs with typing.Union.

Optional Syntax (optional_syntax)

This option controls how optional types are rendered. It supports both the typing.Optional special form and the | operator (bitwise or) syntax from PEP 604. Valid options are defined by the enum OptionalSyntax and include:

• "as_given" (default): render the optional type as it is given without changing syntax
• "or_operator": render all optional types using the | operator
• "union_special_form": render all optional types using the typing.Optional special form
• "optional_special_form": render all optional types using the typing.Optional special form

Note that runtime use of the | operator between types is new in Python 3.10. To use in earlier versions of Python, you will need to use postponed evaluation of annotations à la PEP 563 with from __future__ import__annotations__. Support for the | operator is only a limitation on providing type annotation inputs to typenames, and not a limitation on output rendering.

Limitations:

• Python automatically converts typing.Union[..., None] to typing.Optional[...] when evaluating at runtime. Since typenames uses runtime type objects, it will only see the result using typing.Optional and not know the form of your original input.
• Python automatically flattens unions when evaluating them at runtime. Since typenames uses runtime type objects, it will only see the flattened result and not know if your original input was nested. Furthermore, any mixing of | operator syntax and any typing module types will result in a typing.Union union, so as_given will always render such inputs with typing module special forms.
• The typing.Optional special form only accepts exactly one parameter. By default, typenames will render cases with multiple parameters with Optional[Union[...]]. You can use the union_syntax option to control the inner union's syntax.

Standard Collection Syntax (standard_collection_syntax)

This option controls how parameterized standard collection generic types are rendered. It supports both the typing module's generic aliases (e.g., typing.List[...]) and the standard class (e.g., list[...]) syntax from PEP 585. Valid options are defined by the enum StandardCollectionSyntax and include:

• "as_given" (default): render the parameterized generic type as it is given without changing syntax
• "standard_class": render all parameterized standard collection generic types using their class
• "typing_module": render all parameterized standard collection generic types using the typing module's generic alias

Note that runtime use of standard collection classes as parameterized generic types is new in Python 3.9. To use in earlier versions of Python, you will need to use postponed evaluation of annotations à la PEP 563 with from __future__ import__annotations__. Support for standard collection classes for parameterized generic types is only a limitation on providing type annotation inputs to typenames, and not a limitation on output rendering.

Removing Module Names (remove_modules)

This option controls how module names are removed from the rendered output. It takes a list of inputs, which can either be a string of the module name or a re.Pattern regex pattern directly (the result of re.compile). String inputs are templated into the following regex pattern:

module: str  # Given module name
re.compile(r"^{}\.".format(module.replace(".", r"\.")))

Note that module names are removed in the given order, so having entries that are submodules of other entries can potentially lead to the wrong behavior. You can either order them from higher-depth to lower-depth, or directly provide a compiled pattern with optional groups. For example, the pattern re.compile(r"^collections\.(abc\.)?") will match both "collections." and "collections.abc.".

The default list of module names include the standard library modules relevant to PEP 585 plus types and typing. It can be accessed at DEFAULT_REMOVE_MODULES.

DEFAULT_REMOVE_MODULES: List[Union[str, re.Pattern]] = [
    "__main__",
    "builtins",
    re.compile(r"^collections\.(abc\.)?"),
    "contextlib",
    "re",
    "types",
    "typing",
]

If you are trying to add additional modules to this option (rather than overriding the defaults), the easiest way to do so is to concatenate with the default list:

from typing import Optional
from typenames import typenames, DEFAULT_REMOVE_MODULES, BaseNode

typenames(Optional[BaseNode])
#> 'Optional[typenames.BaseNode]'
typenames(Optional[BaseNode], remove_modules=["typenames"])
#> 'typing.Optional[BaseNode]'
typenames(
    Optional[BaseNode],
    remove_modules=DEFAULT_REMOVE_MODULES + ["typenames"],
)
#> 'Optional[BaseNode]'

To remove all module names, you can use REMOVE_ALL_MODULES, which contains the pattern re.compile(r"^(<?\w+>?\.)+").

Annotated (include_extras)

This option controls whether to render typing.Annotated (or typing_extensions.Annotated if using Python 3.8) and the extra metadata. Annotated is a typing special form introduced in Python 3.9 and originally specified by PEP 593. Many libraries like Pydantic, FastAPI, and Typer use it to attach metadata to type annotations that are used at runtime.

By default, typenames will not render Annotated and extra metadata. Set include_extras=True to render them.

from typing import Annotated
from typenames import typenames

typenames(Annotated[int, "some metadata"])
#> 'int'
typenames(Annotated[int, "some metadata"], include_extras=True)
#> "Annotated[int, 'some metadata']"

---

^{Reproducible examples created by reprexlite.}