Float literal formatting outputs different things dependent on where it is used

[coastalwhite]

Checks

• I have checked that this issue has not already been reported.
• I have confirmed this bug exists on the latest version of Polars.

Reproducible example

pl.DataFrame().with_columns(a=pl.lit(0.0)).with_columns(b=pl.col(“a”).cast(pl.String), c=pl.lit(0.0).cast(pl.String))

Outputs:

shape: (1, 3)
┌─────┬─────┬─────┐
│ a   ┆ b   ┆ c   │
│ --- ┆ --- ┆ --- │
│ f64 ┆ str ┆ str │
╞═════╪═════╪═════╡
│ 0.0 ┆ 0.0 ┆ 0   │
└─────┴─────┴─────┘

Log output

No response

Issue description

Inconsistency

Expected behavior

Consistency

Installed versions

Replace this line with the output of pl.show_versions(). Leave the backticks in place.

[ritchie46]

Likely due to the new scalar. All the formatting of columns should ensure it goes into polars-core/fmt.

Ah, nevermind.

I am not entirely sure it should be the same. One is formatting behavior (which is even subjective to env vars), the other is casting behavior. They should be independent.

We don't want that the casting for floats changes because the formatting of the tables have changed.

EDIT:

I now see that the cast is different depending on it being a column source or literal source. That should be fixed indeed.