Modeling Types

newtype

newtype makes a UserId distinct from a bare int for type checkers, while it behaves like one at runtime.

from pyrula import newtype, newtype_strict

UserId = newtype("UserId", int)
UserId(1)                       # 1, typed as UserId

Email = newtype_strict("Email", str, validate=lambda s: "@" in s)
Email("[email protected]")          # ok
Email("nope")                   # raises ValueError

newtype_strict also takes err_msg for the failure message.

Sealed unions

@sealed marks a closed set of cases and @case declares each as an immutable dataclass. match runs the first matching on(Type, handler) and raises if none match.

from pyrula import sealed, case, match, on

@sealed
class Shape:
    pass

@case
class Circle(Shape):
    radius: float

@case
class Rect(Shape):
    w: float
    h: float

def area(shape: Shape) -> float:
    return match(shape)(
        on(Circle, lambda c: 3.14159 * c.radius ** 2),
        on(Rect, lambda r: r.w * r.h),
    )

area(Circle(2.0))   # 12.566...

A sealed type accepts subclasses only from its defining module; any other raises TypeError at class creation, with no opt-out for tests.

Cases are frozen dataclasses, so dataclasses.replace and asdict work:

import dataclasses

c = Circle(2.0)
dataclasses.replace(c, radius=3.0)   # Circle(radius=3.0)
dataclasses.asdict(c)                # {"radius": 2.0}

The JSON form is {"CaseName": {fields}}; see JSON form.