Skip to content
Chelis docs

Rolling and EWM

Coral.Window is the first Coral module with a runtime-executed pandas parity lane in the checked-in test harness.

Validated operations in the current slice:

  • rolling_sum
  • rolling_mean
  • rolling_std using sample standard deviation (ddof=1)
  • rolling_min
  • rolling_max
  • ewm(alpha, adjust=False)
module Coral.BookWindow
import Coral.Window (rolling_mean, rolling_std, ewm)
export (main)
def main() -> f32 = {
  values = to_tensor([cast(1.0, f32), cast(2.0, f32), cast(3.0, f32), cast(4.0, f32), cast(5.0, f32)])
  means = rolling_mean(copy(values), cast(3, int64))
  stds = rolling_std(copy(values), cast(3, int64))
  smooth = ewm(values, cast(0.5, f32))
  mean_tail = index(to_list(copy(means)), cast(4, int64))
  std_tail = index(to_list(copy(stds)), cast(4, int64))
  smooth_tail = index(to_list(copy(smooth)), cast(4, int64))
  _ = drop(means)
  _ = drop(stds)
  _ = drop(smooth)
  add(mean_tail, add(std_tail, smooth_tail))
}