Class TriggerBuffer
Captures incoming data windows into labelled segments, controlled by a trigger input. Stores both classification and regression labels uniformly as a per-frame MathNet.Numerics.LinearAlgebra.Matrix<T> of target values.
JSON configuration
Use these block entries inside a pipeline configuration. Input names must match other blocks in that pipeline; device addresses and file paths must match your setup.
Block entry for a larger pipeline. Params is not used. Capture must be a Trigger-compatible block that provides targets and start/stop events. Features supplies the matching observations. Completed segments can be used for batch training.
{
"TriggerBuffer": {
"Type": "triggerbuffer",
"Inputs": ["Features", "Capture"]
}
}
How Type, Inputs and Params work · Block catalogue and parameter reference
Implements
Inherited Members
Namespace: MOSAIC.Models.FlowControl
Assembly: MOSAIC.dll
Syntax
public class TriggerBuffer : BaseBlock, INotifyPropertyChanged, INotifyPropertyChanging, ISubscriber, IPublisher, IBaseBlockProps, IJsonExportable, IDisposable, IAsyncDisposable
Remarks
Requires exactly two inputs in the pipeline JSON:
-
A trigger block (first input). Sending a MathNet.Numerics.LinearAlgebra.Vector<T> starts /
updates the active target; sending
nullstops the segment and finalises it. The sender is identified by class name containing"Trigger". -
A data provider (second input, e.g.
SlidingWindoworMockUltrasoundSource). While capture is active, each published MathNet.Numerics.LinearAlgebra.Matrix<T> or MathNet.Numerics.LinearAlgebra.Vector<T> is accumulated as one data row in the segment.
Target shape — uniform across classification and regression.
Each entry in dB has a MathNet.Numerics.LinearAlgebra.Matrix<T> Targets
with one row per data row. For the classic click-driven Trigger
block (which publishes the target Vector once on start), the buffer caches
it as _currentTarget and queues an identical row for every
data frame received during the segment — so classification consumers can
read Targets.Row(0) and ignore the rest. For a streaming trigger
(StreamTrigger, future DataGlove, etc.) that publishes a new
target Vector each tick, the buffer queues whichever target is most recent
at the moment each data row arrives — so regression consumers see a true
per-frame target stream.
Constructors
TriggerBuffer(string)
Declaration
public TriggerBuffer(string name)
Parameters
| Type | Name | Description |
|---|---|---|
| string | name |
Fields
_collecting
Active recording flag — toggled by trigger start / stop.
Declaration
private bool _collecting
Field Value
| Type | Description |
|---|---|
| bool |
_currentLabel
Human-readable name attached to the segment, when known.
Declaration
private string _currentLabel
Field Value
| Type | Description |
|---|---|
| string |
_currentTarget
Most recently received target vector. Held between trigger messages so that data frames arriving in between get labelled with the latest known target — this is what makes classic single-publish triggers and per-tick streaming triggers transparently interchangeable.
Declaration
private Vector<double>? _currentTarget
Field Value
| Type | Description |
|---|---|
| Vector<double> |
_matrices
The data frames themselves; stacked vertically on stop.
Declaration
private readonly List<Matrix<double>> _matrices
Field Value
| Type | Description |
|---|---|
| List<Matrix<double>> |
_targetRows
One row per captured data frame; finalised into a Matrix on stop.
Declaration
private readonly List<Vector<double>> _targetRows
Field Value
| Type | Description |
|---|---|
| List<Vector<double>> |
Properties
JsonTypeName
Block type identifier for JSON serialization. Override for a custom name.
Declaration
protected override string JsonTypeName { get; }
Property Value
| Type | Description |
|---|---|
| string |
Overrides
MaxInputs
Needs exactly two inputs (e.g. a data stream + a Trigger/TriggerBuffer).
Declaration
public override int MaxInputs { get; }
Property Value
| Type | Description |
|---|---|
| int |
Overrides
MinInputs
Needs exactly two inputs (e.g. a data stream + a Trigger/TriggerBuffer).
Declaration
public override int MinInputs { get; }
Property Value
| Type | Description |
|---|---|
| int |
Overrides
dB
All captured segments as (Label, Targets, Data) triples.
Targets.RowCount equals the number of data frames received; the
row at index k is the target that was active when data frame
k arrived. For classification, every row is identical and
consumers can use Targets.Row(0); for regression, rows vary
across time.
Declaration
public List<(string Label, Matrix<double> Targets, Matrix<double> Data)> dB { get; }
Property Value
| Type | Description |
|---|---|
| List<(string Label, Matrix<double> Targets, Matrix<double> Data)> |
Methods
Clear()
Clears all captured segments and resets state.
Declaration
public void Clear()
ConfigureInput(IServiceProvider, JsonModel)
Declaration
public static TriggerBuffer ConfigureInput(IServiceProvider sp, JsonModel m)
Parameters
| Type | Name | Description |
|---|---|---|
| IServiceProvider | sp | |
| JsonModel | m |
Returns
| Type | Description |
|---|---|
| TriggerBuffer |
Dispose()
Performs application-defined tasks associated with freeing, releasing, or resetting unmanaged resources.
Declaration
public override void Dispose()
Overrides
FormatVector(Vector<double>)
Declaration
private static string FormatVector(Vector<double> v)
Parameters
| Type | Name | Description |
|---|---|---|
| Vector<double> | v |
Returns
| Type | Description |
|---|---|
| string |
GetJsonParams()
Block-specific parameters for JSON. Override to provide custom parameters.
Declaration
protected override IReadOnlyList<object>? GetJsonParams()
Returns
| Type | Description |
|---|---|
| IReadOnlyList<object> | The values for Params, in the order this block's |
Overrides
HandleTrigger(object?, object?, string)
Declaration
private void HandleTrigger(object? sender, object? data, string senderName)
Parameters
| Type | Name | Description |
|---|---|---|
| object | sender | |
| object | data | |
| string | senderName |
OnReceive(object, object?)
Processes incoming data from an upstream block. Must be implemented by derived classes.
Declaration
protected override void OnReceive(object sender, object? data)
Parameters
| Type | Name | Description |
|---|---|---|
| object | sender | The publishing block, or whatever else delivered the value. |
| object | data |
Overrides
Remarks
Called on the dispatching thread, not the UI thread.
RemoveSegment(int)
Removes a single segment from dB by index. Bounds-checked; silently no-ops if the index is out of range (e.g. the UI raced with a concurrent Clear()). Does NOT republish dB's count — downstream consumers only react to additions, and a deletion shouldn't force them to retrain mid-session.
Declaration
public bool RemoveSegment(int index)
Parameters
| Type | Name | Description |
|---|---|---|
| int | index | Zero-based index in dB. |
Returns
| Type | Description |
|---|---|
| bool | true if a segment was removed. |
ResetSegmentState()
Declaration
private void ResetSegmentState()
ResolveSegmentLabel(object?, string)
Pick a human-readable label for the segment from whichever trigger-like
block sent the start message. The classic Trigger block exposes
CurrentActionName (e.g. "THUMB"); a future StreamTrigger
exposes SessionName instead. Try both via reflection so the buffer
doesn't need a compile-time reference to either type; fall back to the
sender's class name only if neither property exists or is empty.
Declaration
private static string ResolveSegmentLabel(object? sender, string fallbackTypeName)
Parameters
| Type | Name | Description |
|---|---|---|
| object | sender | |
| string | fallbackTypeName |
Returns
| Type | Description |
|---|---|
| string |
SaveToCsv(string)
Save all segments to CSV. Each row of output is one data frame with
the corresponding target vector prepended:
label, t0, t1, …, tD-1, d0, d1, …, dC-1.
Declaration
public void SaveToCsv(string filePath)
Parameters
| Type | Name | Description |
|---|---|---|
| string | filePath |
StackVectorsAsRows(List<Vector<double>>)
Stack a list of equal-length vectors into a single matrix with one row per vector. Handles the case where vectors have differing lengths by padding shorter rows with zeros up to the widest vector seen.
Declaration
private static Matrix<double> StackVectorsAsRows(List<Vector<double>> rows)
Parameters
| Type | Name | Description |
|---|---|---|
| List<Vector<double>> | rows |
Returns
| Type | Description |
|---|---|
| Matrix<double> |
StartCapture(Vector<double>, string, object?)
Declaration
private void StartCapture(Vector<double> firstLabel, string senderName, object? sender)
Parameters
| Type | Name | Description |
|---|---|---|
| Vector<double> | firstLabel | |
| string | senderName | |
| object | sender |
StopCapture()
Declaration
private void StopCapture()