Wrong polars dtype for empty Series

[Northern-Arc]

Wrong polars dtype for empty Series

[nameexhaustion]

see: Consistently support Python types as aliases for polars types (#13117)

Polars sometimes supports Python types (e.g. str, float, int) as aliases for polars types, and sometimes not.

[JulianCologne]

hi @Northern-Arc

native python dtypes are not suppored afaik, you need to use the polars dtypes

• int -> null BUT
• pl.Int32 -> Int32 (you example
• str -> null BUT
• pl.String -> str

pl.Series('str', dtype=pl.String)

shape: (0,)
Series: 'str' [str]
[
]

[alexander-beedie]

native python dtypes are not suppored afaik, you need to use the polars dtypes

We do actually have a dedicated codepath for interpreting the standard python types, but we definitely prefer the more specific polars-native equivalents (as python int can't tell you if it's i8, i32, u64, for example).

[Northern-Arc]

hi @Northern-Arc

native python dtypes are not suppored afaik, you need to use the polars dtypes

* `int` -> `null` BUT

* `pl.Int32` -> `Int32` (you example

* `str` -> `null` BUT

* `pl.String` -> `str`

pl.Series('str', dtype=pl.String)

shape: (0,)
Series: 'str' [str]
[
]

Thank you!