Module multi 

Multi-column and row-index reductions over a DataSet.

Aggregate semantics (nulls, all-null groups, casting) are documented in docs/REDUCE_AGG_SEMANTICS.md at the repository root.

Structs

FeatureMeanStd

Per-column mean and standard deviation (square root of variance under std_kind).

Functions

arg_max_row

Returns None if column is not in the schema. Otherwise [Some(None)] if there is no non-null comparable value, or [Some(Some((row_index, value)))] for the first row attaining the maximum (stable tie-break).

arg_min_row

Same as arg_max_row for the minimum.

feature_wise_mean_std

One pass over all rows: compute mean and std dev for each listed numeric column (Int64 / Float64). Nulls are ignored. If a column has no non-null values, both fields are Value::Null. Sample std dev is undefined for fewer than two values → Value::Null.

top_k_by_frequency

Non-null value frequencies; returns the top k pairs by count (desc), breaking ties by [value_sort_key] ascending. k == 0 yields an empty vector.