How to right split n times in python polars dataframe (mimic pandas rsplit)

Question

I have a column of strings where the end portion has some information I need to parse into its own columns. Pandas has the rsplit function to split a string from the right, which does exactly what I need:

import polars as pl
df = pl.DataFrame(
    {
        "name": [
            "some_name_set_1",
            "some_name_set_1b",
            "some_other_name_set_2",
            "yet_another_name_set_2",
        ]
    }
)

df.to_pandas()["name"].str.rsplit("_", n=2, expand=True)

                  0    1   2
0         some_name  set   1
1         some_name  set  1b
2   some_other_name  set   2
3  yet_another_name  set   2

How can I mimic this behavior in Polars, which doesn't have an rsplit expression right now?

Answer 1

A single char separator can also be expressed in terms of regex:

df.select(
   pl.col("name").str.extract_groups(r"(.+)_([^_]+)_([^_]+)$")
     .struct.unnest()
)

shape: (4, 3)
┌──────────────────┬─────┬─────┐
│ 1                ┆ 2   ┆ 3   │
│ ---              ┆ --- ┆ --- │
│ str              ┆ str ┆ str │
╞══════════════════╪═════╪═════╡
│ some_name        ┆ set ┆ 1   │
│ some_name        ┆ set ┆ 1b  │
│ some_other_name  ┆ set ┆ 2   │
│ yet_another_name ┆ set ┆ 2   │
└──────────────────┴─────┴─────┘

But it is not a general solution (limited to single chars) and the regexes can be awkward to re-read.