incomplete type declaration at `sqlite3.Connection.create_aggregate()`

Bug report

Bug description:

The create_aggregate documentation does not specify a result type of the aggregation and even allows for multiple aggregating values per row (n_arg parameter).

However, the type declaration at dbapi2.pyi does not reflect that (comments added by me):

def create_aggregate(self, name: str, n_arg: int, aggregate_class: Callable[[], _AggregateProtocol]) -> None: ...

class _AggregateProtocol(Protocol):
    def step(self, value: int, /) -> object: ...  # <-- just one aggregating value of type int
    def finalize(self) -> int: ...   # <-- result has to be int

I think that this is not correct. At least, I have working code that aggregates multiple strings into a combined one.

Also, it is in stark contrast to other, way more generic, type declarations at the same place (comments added by me):

class _AnyParamWindowAggregateClass(Protocol):
    def step(self, *args: Any) -> object: ... # <-- multiple aggregating values of any type allowed
    def inverse(self, *args: Any) -> object: ...
    def value(self) -> _SqliteData: ...
    def finalize(self) -> _SqliteData: ... # <-- result can be many more types

_SqliteData: TypeAlias = str | ReadableBuffer | int | float | None

CPython versions tested on:

3.12

Operating systems tested on:

Windows

Commencez par le stub sqlite3 dbapi2.pyi et comparez create_aggregate avec le comportement documenté de sqlite3.Connection.create_aggregate. Examinez _AnyParamWindowAggregateClass et _SqliteData dans le même fichier, puis déterminez la déclaration compatible pour plusieurs valeurs d’entrée et des résultats d’agrégation non entiers ; c’est terminé lorsque le stub accepte correctement les cas d’utilisation documentés.