typically

Overview

When adding comprehensive type annotations to python projects, you often have to choose where to import identical types.

For instance, you might want to use a TypedDict, so you would naturally import:

from typing import TypedDict

The issue is that now if you use your typed dictionary in a pydantic model, you will get an exception. This is because you actually "should have" imported:

from typing_extensions import TypedDict

There's also types, collections, collections.abc, annotated_types, etc.

This package is intended to create a single source of core types for python projects with very few required dependencies, and maybe at most a few optional ones (eg: pydantic).

Suggested pattern:

import typically as t

Installation

With pip:

pip install typically

Or with uv:

uv add typically

Notes

• Type annotations really shouldn't impose any overhead. Of course, that's possible to avoid, but I'm still aiming to keep the extra import time less than 100 ms. The goal is (maybe) to pack as much commonly used stuff into a single import without incurring overhead.
• All members of builtins are explicitly imported in the __init__ file, as sometimes one might like to override them, but also be able to access original builtin type.
• Various type variables are included. These are always named T_... where the ellipses is the type the type variable is bound to, if any.

TODO

• Implement paramspec objects for functools caching.
• Add pydantic as submodule, as it imposes overhead and should be optional.
• Add improved decimal type.