Polars – can't compute entropy after conversion to pl.List

I’m trying to compute the entropy of a list, but I need to do a conversion first :

import polars as pl
df = pl.DataFrame({"Result": "1, 2, 3"})

df.select(pl.col("Result").str.split(",").cast(pl.List(pl.Float64)).entropy()).collect()

but this gives :

ComputeError: cannot cast List type (inner: 'Float64', to: 'Float64')

What’s wrong here?

>Solution :

For this problem you’ll need to do a couple of things:

1. properly parse the numbers (including the space after the comma)
2. use .list.eval(…entropy()) to calculate the entropy per list
3. the result returns a list of length 1, so we grab the calculated entropy

import polars as pl
print(pl.__version__) # 0.20.2

df = pl.DataFrame({"Result": ["1, 2, 3", "4, 5, 6"]})

print(
    df.select(
        pl.col("Result").str.split(", ")   # ①
        .cast(pl.List(pl.Float64))
        .list.eval(pl.element().entropy()) # ②
        .list.get(0)                       # ③
    )
    # shape: (2, 1)
    # ┌──────────┐
    # │ Result   │
    # │ ---      │
    # │ f64      │
    # ╞══════════╡
    # │ 1.011404 │
    # │ 1.085189 │
    # └──────────┘
)