🥮 mooncakes.io

moonsignalkit

Streaming telemetry statistics and change detection for MoonBit.

signal
timeseries
streaming
telemetry
cusum
anomaly-detection
moon add cn-wn/moonsignalkit@0.3.0
Download zip
Author
Version
0.3.0
License
Apache-2.0
Repository
Last updated
last month
Downloads
17
README

#MoonSignalKit

MoonSignalKit supplies portable time-series and streaming telemetry primitives for MoonBit: batch summaries, bounded rolling quantiles, filters, CSV interchange, online moments, timestamp monitoring, and two-sided CUSUM.

#Add to a project

moon add cn-wn/moonsignalkit

///|
test {
  let monitor = TimestampMonitor::new(10)
  let _first = monitor.observe(Sample::new(100, 1.0))
  let observation = monitor.observe(Sample::new(115, 1.0))
  assert_true(observation.is_gap)
}

The package has no filesystem, network, browser, or FFI dependency, so its API is shared by Native, JavaScript, Wasm, and Wasm-GC targets. For full examples, quality gates, scope, and related work, see the repository README.

#
ChangeDirection

source
pub(all) enum ChangeDirection {
  Rising
  Falling
} derive(Eq, 
Debug)

#
ChangePoint

source
pub(all) struct ChangePoint {
  sample : Sample
  direction : ChangeDirection
  magnitude : Double
  observation : Int
} derive(
Debug)

A sustained level shift reported by CusumDetector.

#
ChangePoint::to_json

source
fn ChangePoint::to_json(self : ChangePoint) -> String

#
CsvError

source
pub(all) enum CsvError {
  EmptyInput
  MissingColumns(Int)
  InvalidTime(Int)
  InvalidValue(Int)
} derive(Eq, 
Debug)

Errors reported by the intentionally small, dependency-free telemetry CSV reader.

The format is two columns: time,value. Quoted fields and locale-specific numeric formats are intentionally out of scope for a portable core library.

#
CusumDetector

source
pub(all) struct CusumDetector {
  target : Double
  drift : Double
  threshold : Double
  positive_sum : Double
  negative_sum : Double
  observations : Int
} derive(
Debug)

Stateful two-sided CUSUM detector for streaming telemetry.

target is the expected process level, drift suppresses ordinary noise, and threshold controls how much cumulative evidence triggers an event.

#
CusumDetector::new

source
fn CusumDetector::new(target : Double, drift : Double, threshold : Double) -> CusumDetector

#
CusumDetector::push

source
fn CusumDetector::push(self : CusumDetector, sample : Sample) -> ChangePoint?

Processes one sample in O(1) time and returns an event only on a level shift.

#
CusumDetector::reset_evidence

source
fn CusumDetector::reset_evidence(self : CusumDetector) -> Unit

Clears accumulated evidence while preserving configuration and count.

#
EwmaFilter

source
pub(all) struct EwmaFilter {
  alpha : Double
  has_value : Bool
  value : Double
} derive(
Debug)

O(1)-memory exponentially weighted moving-average filter.

#
EwmaFilter::new

source
fn EwmaFilter::new(alpha_per_mille : Int) -> EwmaFilter

#
EwmaFilter::push

source
fn EwmaFilter::push(self : EwmaFilter, sample : Sample) -> Sample

#
EwmaFilter::reset

source
fn EwmaFilter::reset(self : EwmaFilter) -> Unit

#
EwmaFilter::value

source
fn EwmaFilter::value(self : EwmaFilter) -> Double?

#
MedianFilter

source
pub(all) struct MedianFilter {
  window : RollingWindow
} derive(
Debug)

A bounded median filter for suppressing isolated telemetry spikes.

#
MedianFilter::new

source
fn MedianFilter::new(window : Int) -> MedianFilter

#
MedianFilter::push

source
fn MedianFilter::push(self : MedianFilter, sample : Sample) -> Sample

#
MedianFilter::reset

source
fn MedianFilter::reset(self : MedianFilter) -> Unit

#
OnlineMoments

source
pub(all) struct OnlineMoments {
  count : Int
  mean : Double
  m2 : Double
  min : Double
  max : Double
} derive(
Debug)

Numerically stable online statistics using Welford's algorithm.

Each sample is processed in O(1) time and O(1) memory, so the accumulator can be used for unbounded telemetry streams without retaining raw samples.

#
OnlineMoments::new

source

#
OnlineMoments::push

source
fn OnlineMoments::push(self : OnlineMoments, value : Double) -> Unit

Adds one observation using a stable one-pass update.

#
OnlineMoments::std_dev

source
fn OnlineMoments::std_dev(self : OnlineMoments) -> Double

#
OnlineMoments::summary

source
fn OnlineMoments::summary(self : OnlineMoments) -> Summary

Takes an immutable summary without resetting the accumulator.

#
OnlineMoments::variance

source
fn OnlineMoments::variance(self : OnlineMoments) -> Double

#
RollingWindow

source
pub(all) struct RollingWindow {
  capacity : Int
  values : Array[Sample]
  size : Int
  next : Int
} derive(
Debug)

A bounded chronological telemetry window.

Appending a sample is O(1). Summaries and quantiles deliberately inspect the bounded window, so their cost is O(window_size) and does not grow with the lifetime of a stream.

#
RollingWindow::capacity

source
fn RollingWindow::capacity(self : RollingWindow) -> Int

#
RollingWindow::is_ready

source
fn RollingWindow::is_ready(self : RollingWindow) -> Bool

#
RollingWindow::length

source
fn RollingWindow::length(self : RollingWindow) -> Int

#
RollingWindow::median

source
fn RollingWindow::median(self : RollingWindow) -> Double

#
RollingWindow::new

source
fn RollingWindow::new(capacity : Int) -> RollingWindow

#
RollingWindow::push

source
fn RollingWindow::push(self : RollingWindow, sample : Sample) -> Unit

Adds one sample, evicting the oldest sample after the capacity is reached.

#
RollingWindow::quantile

source
fn RollingWindow::quantile(self : RollingWindow, permille : Int) -> Double

Returns an exact nearest-rank quantile over the retained values.

permille is clamped to the inclusive range 0..1000, making 500 the median.

#
RollingWindow::reset

source
fn RollingWindow::reset(self : RollingWindow) -> Unit

#
RollingWindow::samples

source
fn RollingWindow::samples(self : RollingWindow) -> Array[Sample]

Returns retained samples from oldest to newest.

#
RollingWindow::summary

source
fn RollingWindow::summary(self : RollingWindow) -> Summary

Computes a numerically stable summary over the current bounded window.

#
Sample

source
pub(all) struct Sample {
  time : Int
  value : Double
} derive(
Debug)

#
Sample::new

source
fn Sample::new(time : Int, value : Double) -> Sample

#
Series

source
pub(all) struct Series {
  name : String
  samples : Array[Sample]
} derive(
Debug)

#
Series::difference

source
fn Series::difference(self : Series, name? : String) -> Series

#
Series::exponential_smoothing

source
fn Series::exponential_smoothing(self : Series, alpha_per_mille : Int, name? : String) -> Series

#
Series::from_csv

source
fn Series::from_csv(name : String, csv : String, has_header? : Bool) -> Result[Series, CsvError]

Imports a two-column time,value CSV document.

Empty lines are ignored. The first non-empty row is treated as a header when has_header is true, which is the default and matches Series::to_csv.

#
Series::is_empty

source
fn Series::is_empty(self : Series) -> Bool

#
Series::length

source
fn Series::length(self : Series) -> Int

#
Series::max_value

source
fn Series::max_value(self : Series) -> Double

#
Series::mean

source
fn Series::mean(self : Series) -> Double

#
Series::min_value

source
fn Series::min_value(self : Series) -> Double

#
Series::moving_average

source
fn Series::moving_average(self : Series, window : Int, name? : String) -> Series

#
Series::new

source
fn Series::new(name : String, samples : Array[Sample]) -> Series

#
Series::normalize_minmax

source
fn Series::normalize_minmax(self : Series, name? : String) -> Series

#
Series::outliers

source
fn Series::outliers(self : Series, z_threshold : Double) -> Array[Sample]

#
Series::peak_count

source
fn Series::peak_count(self : Series, threshold : Double) -> Int

#
Series::peaks

source
fn Series::peaks(self : Series, threshold : Double) -> Array[Sample]

#
Series::range

source
fn Series::range(self : Series) -> Double

#
Series::rate_of_change

source
fn Series::rate_of_change(self : Series, name? : String) -> Series

#
Series::sum

source
fn Series::sum(self : Series) -> Double

#
Series::summary

source
fn Series::summary(self : Series) -> Summary

#
Series::summary_json

source
fn Series::summary_json(self : Series) -> String

#
Series::to_csv

source
fn Series::to_csv(self : Series) -> String

Exports a series as portable UTF-8 CSV with a time,value header.

#
Series::variance

source
fn Series::variance(self : Series) -> Double

#
Summary

source
pub(all) struct Summary {
  count : Int
  min : Double
  max : Double
  mean : Double
  range : Double
  variance : Double
  std_dev : Double
} derive(
Debug)

#
Summary::to_json

source
fn Summary::to_json(self : Summary) -> String

#
TimestampMonitor

source
pub(all) struct TimestampMonitor {
  max_gap : Int
  last_time : Int?
} derive(
Debug)

Detects late samples and excessive timestamp gaps without retaining a stream. The latest in-order timestamp is never moved backwards, making observations deterministic for out-of-order telemetry.

#
TimestampMonitor::new

source
fn TimestampMonitor::new(max_gap : Int) -> TimestampMonitor

#
TimestampMonitor::observe

source

#
TimestampMonitor::reset

source
fn TimestampMonitor::reset(self : TimestampMonitor) -> Unit

#
TimingObservation

source
pub(all) struct TimingObservation {
  previous_time : Int?
  time : Int
  delta : Int
  is_late : Bool
  is_gap : Bool
} derive(
Debug)

Timing quality observed while accepting one telemetry sample.

Source Files

Powered by MoonBit

Site sourceReport issuePackagesBuild queueSkillsStatistics

© 2026 mooncakes.io