Polars equivalent of pandas factorize

Question

Does polars have the function to encode string column into integers (1, 2, 3) like pandas.factorize?

Didn't find it in the polars documentation

Answer 1

Perhaps you're looking for a dense rank or the categorical type.

df = pl.DataFrame({"column": ["foo", "bar", "baz", "foo", "foo"]})

df.with_columns(rank = pl.col("column").rank("dense"))

shape: (5, 2)
┌────────┬──────┐
│ column | rank │
│ ---    | ---  │
│ str    | u32  │
╞════════╪══════╡
│ foo    | 3    │
│ bar    | 1    │
│ baz    | 2    │
│ foo    | 3    │
│ foo    | 3    │
└────────┴──────┘

AFAIK - "first seen" order is a little more involved.

(df.with_row_index("index")
   .with_columns(rank = pl.col("index").first().over("column").rank("dense"))
)

shape: (5, 3)
┌───────┬────────┬──────┐
│ index ┆ column ┆ rank │
│ ---   ┆ ---    ┆ ---  │
│ u32   ┆ str    ┆ u32  │
╞═══════╪════════╪══════╡
│ 0     ┆ foo    ┆ 1    │
│ 1     ┆ bar    ┆ 2    │
│ 2     ┆ baz    ┆ 3    │
│ 3     ┆ foo    ┆ 1    │
│ 4     ┆ foo    ┆ 1    │
└───────┴────────┴──────┘