Structural Pattern Matching for Types

[jorenham (Joren Hammudoglu)]

FWIW, I suggested something similar in Make `@overload` less verbose - #10 by jorenham, with a syntax like

@jorenham

from typing import Literal as L

case type ResetIndexResult[InplaceT: L[True], DropT: bool, VT] = None
case type ResetIndexResult[InplaceT: L[False], DropT: L[True], VT] = Series[VT]
case type ResetIndexResult[InplaceT: L[False], DropT: L[False], VT] = DataFrame

def reset_index[InplaceT: bool = L[False], DropT: bool = L[True]](
    self,
    level: Sequence[Level] | Level | None = ...,
    *,
    drop: DropT = ...,
    name: Level = ...,
    inplace: InplaceT = ..., 
    allow_duplicates: bool = ...,
) -> ResetIndexResult[InplaceT, DropT, S1]: ...

Note sure if it’s better, but I thought I’d just throw it out there.

Unless I’m missing something, I think that this these “case type”'s or “type mappings” (whatever you want to call it), generalize overloads. Put differently, I think that all overloads could be rewritten in terms using these type-mappings. It’s even more DRY that way, because it allows you to recycle (parts of) you overload signatures, because it decouples the overloaded signature from the function definition. It also provides a natural way for composing these definitions, which allows you can re-use some of the overloads.