expanding.apply in polars

Question

In pandas I could call data.expanding(min_periods=1).apply(lambda_func) to call a func on expanding or a cumsum-like view.

How to do the same thing with polars? I could only find rolling_apply or apply.

Answer 1

From polars>=0.13.36 you can run cumulative_eval to run polars expressions on a cumulative increasing window.

Note that you should not use this for an aggregation like sum as this will have O(n^2) complexity.

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

df.select([
    pl.col("values").cumulative_eval(pl.element().first() - pl.element().last() ** 2)
])

shape: (5, 1)
┌────────┐
│ values │
│ ---    │
│ f64    │
╞════════╡
│ null   │
├╌╌╌╌╌╌╌╌┤
│ -3.0   │
├╌╌╌╌╌╌╌╌┤
│ -8.0   │
├╌╌╌╌╌╌╌╌┤
│ -15.0  │
├╌╌╌╌╌╌╌╌┤
│ -24.0  │
└────────┘