Split string column into multiple columns (feature request / use case) about explorer

[billylanchantin]

Hi @ryancurtin,

I would be ok supporting a version of the Polars approach:

(
    df["s"]
    .str.splitn(" ", 2)
    .struct.rename_fields(["first_part", "second_part"])
    .alias("fields")
    .to_frame()
    .unnest("fields")
)

A PR for splitn would be welcome :) Heads up: I think our lack of struct support may make that difficult.

As for other approaches, nothing comes to mind that would be vastly superior. Though that may just be my lack of imagination.

[josevalim]

@billylanchantin we do support structs :D maybe the API we need here is something like: split_into(string, " ", [:first_name, :last_name]). And that returns a struct we can unnest later?

[billylanchantin]

we do support structs :D

Sorry, I should've said "limited" support! I don't think we have rename_fields.

maybe the API we need here is something like: split_into(string, " ", [:first_name, :last_name]). And that returns a struct we can unnest later?

That's a good thought. Take their splitn and add the fields as an argument. Frankly, it seems odd that splitn forces you to name the fields afterward.

[ryancurtin]

@josevalim @billylanchantin Thanks a lot for your guidance. I will take a crack at implementing split_into this weekend and see what I come up with!