Files
QuanTAlib/lib/averages/trima/Trima.cs
T
Miha Kralj 3ed35322a5 Add TRIMA implementation and benchmarks; optimize WMA with SIMD
- Introduced `TrimaVector` class for multi-period Triangular Moving Average (TRIMA) calculations, optimized for SIMD.
- Implemented last-value substitution for invalid inputs in TRIMA.
- Added methods for calculating TRIMA for entire series and individual updates.
- Enhanced `Wma` class with periodic resync to prevent floating-point drift and introduced SIMD optimizations for performance.
- Updated benchmark suite to include TRIMA calculations alongside existing SMA, EMA, and WMA benchmarks.
2025-12-04 13:05:56 -08:00

294 lines
8.5 KiB
C#

using System.Buffers;
using System.Runtime.CompilerServices;
using System.Runtime.InteropServices;
namespace QuanTAlib;
/// <summary>
/// TRIMA: Triangular Moving Average
/// </summary>
/// <remarks>
/// TRIMA is a weighted moving average where the weights increase linearly to the middle
/// of the period and then decrease linearly. It places the most weight on the middle
/// portion of the data series.
///
/// Calculation:
/// TRIMA(period) = SMA(SMA(period1), period2)
/// where:
/// period1 = period / 2 + 1
/// period2 = (period + 1) / 2
///
/// This implementation uses a flattened structure with two internal SMA buffers
/// to ensure correct handling of warmup periods and bar corrections without
/// the overhead of composed objects.
///
/// Key characteristics:
/// - Smoother than SMA
/// - Double smoothing (lag is higher than SMA)
/// - Weights form a triangle
/// - O(1) time complexity
/// - O(period) space complexity
///
/// Sources:
/// - https://www.investopedia.com/terms/t/triangularaverage.asp
/// </remarks>
[SkipLocalsInit]
public sealed class Trima
{
private readonly int _period;
private readonly int _p1;
private readonly int _p2;
private readonly RingBuffer _buffer1;
private readonly RingBuffer _buffer2;
// SMA1 State
private double _sum1;
private double _p_sum1;
private double _p_lastInput1;
private double _lastValidValue1;
private double _p_lastValidValue1;
private int _tickCount1;
// SMA2 State
private double _sum2;
private double _p_sum2;
private double _p_lastInput2;
private int _tickCount2;
private int _sampleCount;
private const int ResyncInterval = 1000;
/// <summary>
/// Display name for the indicator.
/// </summary>
public string Name { get; }
/// <summary>
/// Creates TRIMA with specified period.
/// </summary>
/// <param name="period">Number of values to average (must be > 0)</param>
public Trima(int period)
{
if (period <= 0)
throw new ArgumentException("Period must be greater than 0", nameof(period));
_period = period;
_p1 = period / 2 + 1;
_p2 = (period + 1) / 2;
_buffer1 = new RingBuffer(_p1);
_buffer2 = new RingBuffer(_p2);
Name = $"Trima({period})";
}
/// <summary>
/// Current TRIMA value.
/// </summary>
public TValue Value { get; private set; }
/// <summary>
/// True if the TRIMA has enough data to produce valid results.
/// </summary>
public bool IsHot => _sampleCount >= _period;
/// <summary>
/// Gets a valid input value, using last-value substitution for non-finite inputs.
/// </summary>
[MethodImpl(MethodImplOptions.AggressiveInlining)]
private double GetValidValue(double input)
{
if (double.IsFinite(input))
{
_lastValidValue1 = input;
return input;
}
return _lastValidValue1;
}
/// <summary>
/// Updates TRIMA with the given value.
/// </summary>
/// <param name="input">Input value</param>
/// <param name="isNew">True for new bar, false for update to current bar (default: true)</param>
/// <returns>Current TRIMA value</returns>
[MethodImpl(MethodImplOptions.AggressiveInlining)]
public TValue Update(TValue input, bool isNew = true)
{
if (isNew)
{
_sampleCount++;
// SMA 1 Update
double val1 = GetValidValue(input.Value);
double removed1 = _buffer1.Count == _buffer1.Capacity ? _buffer1.Oldest : 0.0;
_sum1 = _sum1 - removed1 + val1;
_buffer1.Add(val1);
// Resync SMA1
_tickCount1++;
if (_buffer1.IsFull && _tickCount1 >= ResyncInterval)
{
_tickCount1 = 0;
_sum1 = _buffer1.Sum();
}
// Save SMA1 state
_p_sum1 = _sum1;
_p_lastInput1 = val1;
_p_lastValidValue1 = _lastValidValue1;
// SMA 1 Result
double sma1Result = _sum1 / _buffer1.Count;
// SMA 2 Update (Input is sma1Result)
// Note: sma1Result is always finite if input stream has at least one finite value
double removed2 = _buffer2.Count == _buffer2.Capacity ? _buffer2.Oldest : 0.0;
_sum2 = _sum2 - removed2 + sma1Result;
_buffer2.Add(sma1Result);
// Resync SMA2
_tickCount2++;
if (_buffer2.IsFull && _tickCount2 >= ResyncInterval)
{
_tickCount2 = 0;
_sum2 = _buffer2.Sum();
}
// Save SMA2 state
_p_sum2 = _sum2;
_p_lastInput2 = sma1Result;
// Final Result
double trimaResult = _sum2 / _buffer2.Count;
Value = new TValue(input.Time, trimaResult);
}
else
{
// SMA 1 Correction
_lastValidValue1 = _p_lastValidValue1;
double val1 = GetValidValue(input.Value);
_sum1 = _p_sum1 - _p_lastInput1 + val1;
_buffer1.UpdateNewest(val1);
double sma1Result = _sum1 / _buffer1.Count;
// SMA 2 Correction
_sum2 = _p_sum2 - _p_lastInput2 + sma1Result;
_buffer2.UpdateNewest(sma1Result);
double trimaResult = _sum2 / _buffer2.Count;
Value = new TValue(input.Time, trimaResult);
}
return Value;
}
/// <summary>
/// Updates TRIMA with the entire series.
/// </summary>
/// <param name="source">Input series</param>
/// <returns>TRIMA series</returns>
public TSeries Update(TSeries source)
{
if (source.Count == 0) return new TSeries(new List<long>(), new List<double>());
// Use the static Calculate method for performance
int len = source.Count;
var t = new List<long>(len);
var v = new List<double>(len);
CollectionsMarshal.SetCount(t, len);
CollectionsMarshal.SetCount(v, len);
var tSpan = CollectionsMarshal.AsSpan(t);
var vSpan = CollectionsMarshal.AsSpan(v);
var sourceValues = source.Values;
var sourceTimes = source.Times;
Calculate(sourceValues, vSpan, _period);
sourceTimes.CopyTo(tSpan);
// Restore state by replaying the last part
// We need to replay enough to fill both SMAs
int lookback = _p1 + _p2;
int startIndex = Math.Max(0, len - lookback);
// Reset internal state
Reset();
// Replay
for (int i = startIndex; i < len; i++)
{
Update(new TValue(sourceTimes[i], sourceValues[i]), isNew: true);
}
Value = new TValue(tSpan[len - 1], vSpan[len - 1]);
return new TSeries(t, v);
}
/// <summary>
/// Calculates TRIMA for the entire series using a new instance.
/// </summary>
public static TSeries Calculate(TSeries source, int period)
{
var trima = new Trima(period);
return trima.Update(source);
}
/// <summary>
/// Calculates TRIMA in-place.
/// Uses ArrayPool to allocate temporary buffer and chains optimized SMA calculations.
/// </summary>
public static void Calculate(ReadOnlySpan<double> source, Span<double> output, int period)
{
if (source.Length != output.Length)
throw new ArgumentException("Source and output must have the same length");
if (period <= 0)
throw new ArgumentException("Period must be greater than 0", nameof(period));
int p1 = period / 2 + 1;
int p2 = (period + 1) / 2;
// Rent a temporary buffer for the intermediate SMA
double[] tempArray = ArrayPool<double>.Shared.Rent(source.Length);
Span<double> tempSpan = tempArray.AsSpan(0, source.Length);
try
{
// SMA 1
Sma.Calculate(source, tempSpan, p1);
// SMA 2 (TRIMA)
Sma.Calculate(tempSpan, output, p2);
}
finally
{
ArrayPool<double>.Shared.Return(tempArray);
}
}
/// <summary>
/// Resets the TRIMA state.
/// </summary>
public void Reset()
{
_buffer1.Clear();
_buffer2.Clear();
_sum1 = 0;
_p_sum1 = 0;
_p_lastInput1 = 0;
_lastValidValue1 = 0;
_p_lastValidValue1 = 0;
_tickCount1 = 0;
_sum2 = 0;
_p_sum2 = 0;
_p_lastInput2 = 0;
_tickCount2 = 0;
_sampleCount = 0;
Value = default;
}
}