dependent-types

Dependent types for Python.

Installation

python3 -m pip install dependent-types

Abstract

• Make illegal states unrepresentable.
• Parse, don't validate
• Abuse __instancecheck__ and type-guards.
• Values are checked at runtime but no extra instances/subclasses are instantiated.

Usage

Builtin types

dept.datetime

• TZAware
• TZNaive

dept.numeric

• Natural
• NegativeInt
• Portion

dept.re

• Match

dept.sized

• NonEmpty
• Empty

Creating dependent types

To create new dependent types, subclass dept.base.Dependent and define an __instancecheck__ method:

from typing import Any
from typing import TYPE_CHECKING

from dept.base import Dependent


class Greeting(str, Dependent):
    @classmethod
    def __instancecheck__(cls, instance: Any) -> bool:
        return (
            isinstance(instance, str)
            and instance.startswith(("Hello", "Hi"))
        )


hello = "Hello there"
# We can narrow types using mypy's type guards
assert isinstance(hello, Greeting)
# or explicitly when we need to
hi = Greeting.from_instance("Hi there")

# The runtime types are unchanged and will still be str for our greetings
assert type(hello) is str
assert type(hi) is str

# But their static types will be Greeting, retaining the information that our
# strings are not just any strs
if TYPE_CHECKING:
    reveal_type(hello)
    reveal_type(hi)

# As this string doesn't fulfill our __instancecheck__, it will not be an
# instance of Greeting.
assert not isinstance("Goodbye", Greeting)

Checkout out the dacite example for how to create dataclasses with rich dependently typed fields without duplicating type definitions or losing parsed information.