Implement $cat$get_categories()

[etiennebacher]

library(polars)

df = pl$DataFrame(
  cats = factor(c("z", "z", "k", "a", "b")),
  vals = factor(c(3, 1, 2, 2, 3))
)
df
#> shape: (5, 2)
#> ┌──────┬──────┐
#> │ cats ┆ vals │
#> │ ---  ┆ ---  │
#> │ cat  ┆ cat  │
#> ╞══════╪══════╡
#> │ z    ┆ 3    │
#> │ z    ┆ 1    │
#> │ k    ┆ 2    │
#> │ a    ┆ 2    │
#> │ b    ┆ 3    │
#> └──────┴──────┘

df$select(
  pl$col("cats")$cat$get_categories()
)
#> shape: (4, 1)
#> ┌──────┐
#> │ cats │
#> │ ---  │
#> │ str  │
#> ╞══════╡
#> │ z    │
#> │ k    │
#> │ a    │
#> │ b    │
#> └──────┘
df$select(
  pl$col("vals")$cat$get_categories()
)
#> shape: (3, 1)
#> ┌──────┐
#> │ vals │
#> │ ---  │
#> │ str  │
#> ╞══════╡
#> │ 3    │
#> │ 1    │
#> │ 2    │
#> └──────┘

[sorhawell]

looks good to me :) thanks

[etiennebacher]

Thanks for the small fix. Quick question, how do you know that get_categories() is not fallible?

[sorhawell]

Thanks for the small fix. Quick question, how do you know that get_categories() is not fallible?

The rust-polars method return type Expr aka. pl::Expr
https://docs.rs/polars/latest/polars/prelude/struct.CategoricalNameSpace.html#method.get_categories

It would return PolarsResult if it was fallible. It could in theory also panic but those we just let crash and burn.

In generel any r-polars method which take no input and call a rust function/method which returns no result will be infallible. robj_to!() is fallible because user could give an erroneous input.

If binding together the output of two functions returning Expr but one is fallible the other one not. We can use Ok() to wrap it

let res_expr: Result<Expr> = if condition {
 fallible_f()
} else {
 Ok(infallible_f())
};

If only one infallible function, we don't need to wrap in an Ok, to just unwrap it again. It is not buggy, and everything would work as expected if we did.

[etiennebacher]

I see, thanks for the explanation, slowly getting more comfortable with the Rust side of polars :)