Handling of Polars decimals in scales

[jnolis]

Hello!

As an R user more and more using Python I am a huge fan of this package!

I ran into a problem recently due to a Decimal column in a Polars DataFrame. It seems like in some contexts plotnine treats them as discrete and not continuous values. I have a work around of changing the columns to Floats, but ideally that wouldn't be necssary.

Thank you!

This is an example of something I would expect to work but does not:

df = pl.DataFrame({"x": ["a","b","c"], "y": [1,2,3]}, schema= {"x":pl.String(),"y": pl.Decimal()})
ggplot(df, aes(x = "x", y = "y")) + geom_col() + scale_y_continuous(limits = [0,5])

Returns TypeError: Discrete value supplied to continuous scale

[has2k1]

The issue here is that internally plotnine uses pandas which does not have a decimal dtype.
For pandas, a column of decimal is of type object, which is the catch all for discrete variables.

We could do a cheap hack and convert polars decimals to floats instead of the user having to do it. I think that would be sufficient for most expectations.