fix(rust): Fix cum_min and cum_max does not preserve inf or -inf values at series start

[Athsus]

What’s Changed

Improved cum_min / cum_max for floating-point numbers, and NaN handling.

This update refines the

1. Value init when doing cum_max_numeric and cum_min_numeric.
   It now checks the type: if it's a float, the starting value for the calculation is NaN. Otherwise, it uses the standard min/max possible value for that type.

2. Correct NaN handling in det_min and det_max
   Now it uses MinMax::min_ignore_nan and MinMax::max_ignore_nan to compare the values. For example, if a series starts with NaN, the result will also start with NaN, as expected.

3. New cum_scan_numeric function.
   A new helper function to handle the common scanning logic.

Added Test

def test_cum_agg_with_infs() -> None:
    # confirm that inf values are handled correctly
    s = pl.Series([float("inf"), 0.0, 1.0])
    assert_series_equal(s.cum_min(), pl.Series([float("inf"), 0.0, 0.0]))

    s = pl.Series([float("-inf"), 0.0, 1.0])
    assert_series_equal(s.cum_max(), pl.Series([float("-inf"), 0.0, 1.0]))

Would close the issue #22855

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 80.31%. Comparing base (f230f12) to head (cc01fb0).
Report is 9 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #22896      +/-   ##
==========================================
- Coverage   80.61%   80.31%   -0.31%     
==========================================
  Files        1677     1682       +5     
  Lines      222297   223181     +884     
  Branches     2801     2804       +3     
==========================================
+ Hits       179205   179248      +43     
- Misses      42425    43264     +839     
- Partials      667      669       +2     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Athsus]

Hi @orlp , please review and suggest if needed anything.

[orlp]

You can save a lot of repetition by defining

fn cum_scan_numeric<T, F>(ca: &ChunkedArray<T>, reverse: bool, init: T::Native, update: F) -> ChunkedArray<T>
where
    T: PolarsNumericType,
    ChunkedArray<T>: FromIterator<Option<T::Native>>,
    F: Fn(&mut T::Native, Option<T::Native>) -> Option<Option<T::Native>>
{
    let out: ChunkedArray<T> = match reverse {
        false => ca.iter().scan(init, update).collect_trusted(),
        true => ca.iter().rev().scan(init, update).collect_reversed(),
    };
    out.with_name(ca.name().clone())
}

and then using

let init = if <$T>::is_float() { <$T>::nan_value() } else { <$T>::min_value() };
cum_scan_numeric(ca, reverse, init, det_max)

and similarly for the minimum in the dispatch.

Note that your current init value for floats is incorrect, it needs to be NaN (as a test case try adding a series with NaN as the first element).

[orlp]

See above.

[Athsus]

You can save a lot of repetition by defining

fn cum_scan_numeric<T, F>(ca: &ChunkedArray<T>, reverse: bool, init: T::Native, update: F) -> ChunkedArray<T>
where
    T: PolarsNumericType,
    ChunkedArray<T>: FromIterator<Option<T::Native>>,
    F: Fn(&mut T::Native, Option<T::Native>) -> Option<Option<T::Native>>
{
    let out: ChunkedArray<T> = match reverse {
        false => ca.iter().scan(init, update).collect_trusted(),
        true => ca.iter().rev().scan(init, update).collect_reversed(),
    };
    out.with_name(ca.name().clone())
}

and then using

let init = if <$T>::is_float() { <$T>::nan_value() } else { <$T>::min_value() };
cum_scan_numeric(ca, reverse, init, det_max)

and similarly for the minimum in the dispatch.

Note that your current init value for floats is incorrect, it needs to be NaN (as a test case try adding a series with NaN as the first element).

Thanks! That's very helpful. I've made some new changes on it. @orlp

[Athsus]

Wait, I missed implementing cum_scan_numeric to cum_sum_numeric and cum_prod_numeric

[Athsus]

Hi @orlp , I've done new commits on it.
Though the CI fails, I think it's not because of my modifications. As I don't have the permission to re-run it, could you do the re-run if it's available and necessary?

[orlp]

Thank you for helping out :)

[Athsus]

Thank you for helping out :)

All good! Appreciations.