ddn.columnar.window

The window-function layer of ddn.df (ADR-2026-10-24-17): over (partitioned aggregation scattering back to input rows) and rolling windows (fixed-width receding frames). Both reuse the existing eager kernels — no new computation paths.

Copyright

License

BSD-3-Clause

enum RollingAgg

struct NumericDoubles

fn fillDoubles numericDoublesOf rolling rollingT rollingTime

Types 2

private enumRollingAgg

sumOp

meanOp

countOp

minOp

maxOp

stdOp

varOp

private structNumericDoubles

Fields

double[] data

byte[] valid

Functions 6

Series rolling(Series s,  size_t windowSize,  size_t minPeriods,
       string agg)

@trustedComputes a rolling aggregation over fixed-width receding frames (ADR-2026-10-24-17): row `i` sees `[max(0, i - windowSize + 1) .. i]` inclusive; `minPeriods` (default 1) nulls rows with too few val...

Series rolling(Series s,  size_t windowSize,  size_t minPeriods,
       string agg,  size_t ddof)

@trustedDitto, with the dispersion degrees of freedom: `ddof` applies to the `"std"` and `"variance"` aggregations only (default 1; a window whose valid count is <= ddof reads null; ADR-2026-10-28-01).

private fn

Series rollingT(T)(Series s,  size_t windowSize,
       size_t minPeriods,  string agg,  size_t ddof)

@trusted

Series rollingTime(Series values,  Series times,  long  window,
       size_t minPeriods =  1,  string agg =  "sum",  size_t ddof =  1)

@trustedThe time-based rolling aggregation (ADR-2026-10-30-03): over an explicit (values, times) pair - no hidden index column. The times are DATETIME and must be sorted ascending (validated, throws; never...

private fnNumericDoubles numericDoublesOf(Series s) @trusted

private fnvoid fillDoubles(T)(Series s, double[] data, byte[] valid)