Module agg_kernels 

Specialized aggregate kernels for grouped operations.

These functions operate on gathered data slices with segment boundaries, or on arbitrary row-index groups. All f64 reductions use Kahan summation for deterministic, numerically stable results.

Functions

agg_count

Count per segment.

agg_first_f64

First f64 per segment.

agg_last_f64

Last f64 per segment.

agg_max_f64

Maximum f64 per segment. Returns NAN for empty segments.

agg_max_i64

Maximum i64 per segment.

agg_mean_f64

Kahan-stable mean over contiguous segments.

agg_median_f64

Median via sort per segment. Returns NAN for empty segments.

agg_min_f64

Minimum f64 per segment. Returns NAN for empty segments.

agg_min_i64

Minimum i64 per segment.

agg_n_distinct_i64

Count distinct i64 values per segment using BTreeSet (deterministic).

agg_n_distinct_str

Count distinct strings per segment using BTreeSet (deterministic).

agg_quantile_f64

Quantile via sort + linear interpolation. Returns NAN for empty segments.

agg_sd_f64

Standard deviation via Welford’s algorithm.

agg_sum_f64

Kahan-stable sum over contiguous segments.

agg_sum_i64

Sum for i64 (wrapping on overflow).

agg_var_f64

Variance via Welford’s online algorithm (numerically stable). Returns population variance (divide by N, not N-1).

gather_agg_mean_f64

Kahan-stable mean over gathered f64 rows.

gather_agg_n_distinct_str

Count distinct strings over gathered rows.

gather_agg_sum_f64

Kahan-stable sum over gathered f64 rows.

gather_agg_var_f64

Welford variance over gathered f64 rows.