BUG: `to_datetime` can return `NaTType` about pandas-stubs HOT 3 CLOSED

We made a conscious decision to not have pd.to_datetime() return Union[pd.NaT, Timestamp] because then you'd have to use cast and write this in lots of code:

out = cast(pd.Timestamp, pd.to_datetime("2022-07-05"))

to force out to be pd.Timestamp, which is what you want most of the time.

There's not a way in static typing to test if the timestamp is valid.

from pandas-stubs.

@Dr-Irv why not @overload with errors:Literal["coerce"] with a return type [Timestamp, NaTType] and errors:Literal["ignore","raise"] with return type Timestamp? This type of overloading would let most users ignore the NaT and the cast while letting those who use errors="coerce" get correct typing.

Reading python/mypy#5090 both convinced me that str and Sequence[str] are the same, and that the role of a typechecker is to be as honest as possible with the underlying language, including undesirable warts.

from pandas-stubs.

@Dr-Irv why not @overload with errors:Literal["coerce"] with a return type [Timestamp, NaTType] and errors:Literal["ignore","raise"] with return type Timestamp? This type of overloading would let most users ignore the NaT and the cast while letting those who use errors="coerce" get correct typing.

Reasonable suggestion. Open to a PR. Have reopened the issue.

Reading python/mypy#5090 both convinced me that str and Sequence[str] are the same, and that the role of a typechecker is to be as honest as possible with the underlying language, including undesirable warts.

Not sure that is relevant here - I think you're referring to #82, so comment there on your perspective on this issue.

from pandas-stubs.