How to check if element value in Rust-Polars column is 0

I am trying to account for a zero in the denominator when calculating a normalized percentage deviation from the term in the denominator.

Here is the code snippet -

let out: DataFrame = input_df
    .lazy()
    .with_columns(vec![
        when(col("write") == lit(0.0) && result.wr == 0.0)
            .then(lit(0.0))
            .otherwise((lit(result.wr) - col("write")) / col("write") * lit(100))
            .alias("write%"),
    ])
    .collect()
    .unwrap();

The problem is col("write") == lit(0.0) resolves to false even when the respective element in col("write") is 0.0. I'm unclear why or how to solve it.

col("write") is of type f64, so is result.wr.

I tried changing to just lit(0) which yielded the same result. Not using lit() throws a 'Expected Expr' error.

There are a few issues here. The first being a logical one. You are stating you want to account for a zero in the denominator, however the following line will only ever evaluate to true if zero is in the denominator and the numerator:

when(col("write") == lit(0.0) && result.wr == 0.0)

The second issue is with the ==, when comparing column values to literals, you should be using the built-in expressions for the columns. In this case, you can use col("write").eq(lit(0.0)).

Here's a working example, note that I removed the numerator check as it's not needed:

let mut input_df = df!("write" => &[0.0, 3.0, 999.0]).unwrap();
let wr = 66.0;
let out: DataFrame = input_df
	.lazy()
	.with_columns(vec![
			when(col("write").eq(lit(0.0)))
					.then(lit(0.0))
					.otherwise((lit(wr) - col("write")) / col("write") * lit(100))
					.alias("write%"),
	])
	.collect()
	.unwrap();
println!("{:?}", out);

And the output:

shape: (3, 2)

┌───────┬────────────┐
│ write ┆ write%     │
│ ---   ┆ ---        │
│ f64   ┆ f64        │
╞═══════╪════════════╡
│ 0.0   ┆ 0.0        │
│ 3.0   ┆ 2100.0     │
│ 999.0 ┆ -93.393393 │
└───────┴────────────┘