Allow polars.Expr.round to get negative decimals

[herrmann1981]

Description

Sometimes I want to round values not to a decimal position after the comma but before the comma. This is currently not possible with the round function.

df = pl.DataFrame({'number': [12, 15, 13.6, 13.69, 13.98456]}, schema={'number': pl.Float32})
df.with_columns(rounded=pl.col('number').round(-1))

This currently results in the following error:
OverflowError: can't convert negative int to unsigned

What I would expect is something like:

df.with_columns(rounded=pl.col('number').round(-1))

| number | rounded |
| --- |--- │
| f32 | f32 |
| | |
| 12.0 | 10.0 │
| 15.0 | 16.0 │
| 13.6 | 10.0 │
| 13.69 | 10.0 │
| 13.98456 | 10.0 │

[Julian-J-S]

I agree that this is useful and python has native support for that.

There are multiple "hacks" to achieve this also with the current methods e.g.:

• 1_234: round with -2
• divide by 10**2: 1234 / 100 = 12.34
• round(0) or cast to int -> 12
• multiply by 10**2 -> 1200 😉

[herrmann1981]

@Julian-J-S Thanks for the fast response. Yes I know about this workaround. I just thought it would be useful to have the shortcut in the framework.