add sma ema macd

go.mod

@@ -7,8 +7,10 @@ require (
 	github.com/goccy/go-json v0.10.5
 	github.com/gocql/gocql v1.5.0
 	github.com/lxzan/gws v1.8.9
+	github.com/markcheno/go-talib v0.0.0-20250114000313-ec55a20c902f
 	github.com/panjf2000/ants/v2 v2.11.3
 	github.com/scylladb/gocqlx/v3 v3.0.1
+	github.com/shopspring/decimal v1.4.0
 	github.com/stretchr/testify v1.10.0
 	github.com/testcontainers/testcontainers-go v0.38.0
 	github.com/zeromicro/go-zero v1.8.5
@@ -55,7 +57,6 @@ require (
 	github.com/power-devops/perfstat v0.0.0-20210106213030-5aafc221ea8c // indirect
 	github.com/scylladb/go-reflectx v1.0.1 // indirect
 	github.com/shirou/gopsutil/v4 v4.25.5 // indirect
-	github.com/shopspring/decimal v1.4.0 // indirect
 	github.com/sirupsen/logrus v1.9.3 // indirect
 	github.com/tklauser/go-sysconf v0.3.12 // indirect
 	github.com/tklauser/numcpus v0.6.1 // indirect

go.sum

@@ -156,6 +156,8 @@ github.com/magiconair/properties v1.8.10 h1:s31yESBquKXCV9a/ScB3ESkOjUYYv+X0rg8S
 github.com/magiconair/properties v1.8.10/go.mod h1:Dhd985XPs7jluiymwWYZ0G4Z61jb3vdS329zhj2hYo0=
 github.com/mailru/easyjson v0.7.7 h1:UGYAvKxe3sBsEDzO8ZeWOSlIQfWFlxbzLZe7hwFURr0=
 github.com/mailru/easyjson v0.7.7/go.mod h1:xzfreul335JAWq5oZzymOObrkdz5UnU4kGfJJLY9Nlc=
+github.com/markcheno/go-talib v0.0.0-20250114000313-ec55a20c902f h1:iKq//xEUUaeRoXNcAshpK4W8eSm7HtgI0aNznWtX7lk=
+github.com/markcheno/go-talib v0.0.0-20250114000313-ec55a20c902f/go.mod h1:3YUtoVrKWu2ql+iAeRyepSz3fy6a+19hJzGS88+u4u0=
 github.com/mattn/go-colorable v0.1.14 h1:9A9LHSqF/7dyVVX6g0U9cwm9pG3kP9gSzcuIPHPsaIE=
 github.com/mattn/go-colorable v0.1.14/go.mod h1:6LmQG8QLFO4G5z1gPvYEzlUgJ2wF+stgPZH1UqBm1s8=
 github.com/mattn/go-isatty v0.0.20 h1:xfD0iDuEKnDkl03q4limB+vH+GxLEtL/jb4xVJSWWEY=

internal/lib/strategy/bool.go

@@ -0,0 +1,136 @@
+package strategy
+
+import (
+	"github.com/shopspring/decimal"
+	"math"
+)
+
+/************** O(1) 滾動標準差：Welford + ring buffer **************/
+
+// 滾動統計（固定窗寬）：維持 mean 與 M2（平方離差和）
+// 支援「移除最舊樣本」+「新增最新樣本」的 O(1) 更新
+type rollingWelford struct {
+	n     int               // 窗寬
+	buf   []decimal.Decimal // 環狀緩衝（固定長度 n）
+	head  int               // 下一個覆寫位置（最舊元素位置）
+	size  int               // 當前已填入數量（<= n）
+	mean  decimal.Decimal   // 目前視窗內的平均值
+	m2    decimal.Decimal   // 目前視窗內的平方離差和（∑(x-mean)^2）
+	ready bool              // 是否已填滿 n
+}
+
+func newRollingWelford(n int) *rollingWelford {
+	if n <= 0 {
+		panic("rollingWelford window must be > 0")
+	}
+	return &rollingWelford{
+		n:   n,
+		buf: make([]decimal.Decimal, n),
+	}
+}
+
+// add 一個樣本（標準 Welford 加法）
+func (rw *rollingWelford) add(x decimal.Decimal) {
+	// n_old = size, n_new = size+1
+	nOld := decimal.NewFromInt(int64(rw.size))
+	nNew := nOld.Add(decimal.NewFromInt(1))
+
+	delta := x.Sub(rw.mean)
+	meanNew := rw.mean.Add(delta.Div(nNew))
+	delta2 := x.Sub(meanNew)
+	rw.m2 = rw.m2.Add(delta.Mul(delta2))
+	rw.mean = meanNew
+
+	if rw.size < rw.n {
+		rw.size++
+		if rw.size == rw.n {
+			rw.ready = true
+		}
+	}
+}
+
+// remove 一個樣本（Welford 反向移除公式）
+func (rw *rollingWelford) remove(x decimal.Decimal) {
+	if rw.size <= 0 {
+		return
+	}
+	// n_old = size, n_new = size-1
+	nOld := decimal.NewFromInt(int64(rw.size))
+	nNew := nOld.Sub(decimal.NewFromInt(1))
+	if nNew.LessThanOrEqual(decimal.Zero) {
+		// 視窗將清空
+		rw.size = 0
+		rw.mean = decimal.Zero
+		rw.m2 = decimal.Zero
+		rw.ready = false
+		return
+	}
+	delta := x.Sub(rw.mean)                 // (x - mean_old)
+	meanNew := rw.mean.Sub(delta.Div(nNew)) // mean_new = mean_old - delta / n_new
+	// M2_new = M2_old - delta*(x - mean_new)
+	rw.m2 = rw.m2.Sub(delta.Mul(x.Sub(meanNew)))
+	rw.mean = meanNew
+	rw.size--
+	rw.ready = rw.size == rw.n // 正常來說 size==n-1，移除後會在 add 後再回到 n
+}
+
+// push：O(1) 更新（移除最舊 + 新增最新）
+func (rw *rollingWelford) push(x decimal.Decimal) {
+	if rw.size == rw.n {
+		old := rw.buf[rw.head]
+		rw.remove(old)
+	}
+	rw.buf[rw.head] = x
+	rw.head = (rw.head + 1) % rw.n
+	rw.add(x)
+}
+
+// mid() = 平均；std() = sqrt(variance)
+// 這裡採「母體方差」分母 = n（與很多圖表預設一致；如需樣本用 n-1）
+func (rw *rollingWelford) mid() decimal.Decimal { return rw.mean }
+
+func (rw *rollingWelford) std() decimal.Decimal {
+	if !rw.ready || rw.size == 0 {
+		return decimal.Zero
+	}
+	nDec := decimal.NewFromInt(int64(rw.n))
+	variance := rw.m2.Div(nDec)
+	f, _ := variance.Float64()
+	return decimal.NewFromFloat(math.Sqrt(f))
+}
+
+/************** Boll（Welford 版本） **************/
+
+type BollW struct {
+	rw *rollingWelford
+}
+
+type BollWOut struct {
+	Mid, Upper, Lower, Std decimal.Decimal
+	Ready                  bool
+}
+
+func NewBollW(n int) *BollW {
+	return &BollW{rw: newRollingWelford(n)}
+}
+
+// Push 輸入收盤價 close 與 k 倍標準差（典型 2.0）
+// 回傳：上下軌、均線、標準差、是否就緒（前 n 根為 false）
+func (b *BollW) Push(close, k decimal.Decimal) BollWOut {
+	b.rw.push(close)
+	if !b.rw.ready {
+		return BollWOut{Ready: false}
+	}
+	mid := b.rw.mid()
+	std := b.rw.std()
+
+	upper := mid.Add(k.Mul(std))
+	lower := mid.Sub(k.Mul(std))
+	return BollWOut{
+		Mid:   mid,
+		Upper: upper,
+		Lower: lower,
+		Std:   std,
+		Ready: true,
+	}
+}

internal/lib/strategy/bool_test.go

@@ -0,0 +1,65 @@
+package strategy
+
+import (
+	"github.com/shopspring/decimal"
+	"testing"
+)
+
+func TestBoll_WarmupAndSliding(t *testing.T) {
+	n := 5
+	k := d(2.0)
+	b := NewBollW(n)
+
+	// 前 n-1 根：not ready
+	for i, px := range []float64{10, 11, 12, 13} {
+		out := b.Push(d(int64(px)), k)
+		if out.Ready {
+			t.Fatalf("i=%d: should not be ready yet", i)
+		}
+	}
+
+	// 第 n 根開始：ready
+	out := b.Push(d(14), k)
+	if !out.Ready {
+		t.Fatalf("should be ready at %d-th push", n)
+	}
+	// 中線應為 (10+11+12+13+14)/5 = 12
+	if out.Mid.StringFixed(6) != d(12).StringFixed(6) {
+		t.Fatalf("mid expect 12, got %s", out.Mid)
+	}
+	// 標準差 > 0；上軌 > 中線 > 下軌
+	if !out.Std.GreaterThan(decimal.Zero) ||
+		!out.Upper.GreaterThan(out.Mid) ||
+		!out.Mid.GreaterThan(out.Lower) {
+		t.Fatalf("band ordering violated: U=%s M=%s L=%s Std=%s", out.Upper, out.Mid, out.Lower, out.Std)
+	}
+
+	// 再推兩根 -> 視窗滑到 [12,13,14,15,16]，中線=14
+	for _, px := range []float64{15, 16} {
+		out = b.Push(d(int64(px)), k)
+	}
+	if out.Mid.StringFixed(6) != d(14).StringFixed(6) {
+		t.Fatalf("sliding mid expect 14, got %s", out.Mid)
+	}
+	if !out.Std.GreaterThan(decimal.Zero) {
+		t.Fatalf("std should remain > 0 after slide, got %s", out.Std)
+	}
+}
+
+func TestBoll_FlatSeries(t *testing.T) {
+	n := 4
+	k := d(2.0)
+	b := NewBollW(n)
+
+	// 全部相同價格 -> 標準差=0、三條線重合
+	for i := 0; i < n; i++ {
+		_ = b.Push(d(10), k)
+	}
+	out := b.Push(d(10), k) // 視窗仍為相同值
+	if !out.Std.Equal(decimal.Zero) {
+		t.Fatalf("std should be 0 on flat series, got %s", out.Std)
+	}
+	if !(out.Upper.Equal(out.Mid) && out.Mid.Equal(out.Lower) && out.Mid.Equal(d(10))) {
+		t.Fatalf("bands should coincide at 10: U=%s M=%s L=%s", out.Upper, out.Mid, out.Lower)
+	}
+}

internal/lib/strategy/ema.go

@@ -32,6 +32,7 @@ func newEMACore(period uint) *emaCore {
 	if period == 0 {
 		panic("EMA period must be > 0")
 	}
+
 	return &emaCore{
 		period: period,
 		alpha:  alphaFromPeriod(period),

internal/lib/strategy/ema_sma_talib_test.go

@@ -0,0 +1,88 @@
+package strategy
+
+import (
+	"github.com/markcheno/go-talib"
+	"github.com/shopspring/decimal"
+	"math"
+	"testing"
+)
+
+func dv(v float64) decimal.Decimal { return decimal.NewFromFloat(v) }
+
+func genPrices(n int) []float64 {
+	out := make([]float64, n)
+	base := 10.0
+	for i := 0; i < n; i++ {
+		out[i] = base + float64(i)*0.5 + math.Sin(float64(i)/3.0)*0.7
+	}
+	return out
+}
+
+func almostEqualDecFloat(dec decimal.Decimal, f float64, tol float64) bool {
+	df, _ := dec.Float64()
+	return math.Abs(df-f) <= tol
+}
+
+// 注意：使用 NewEMAForTalib(period)（First-Price seed）來比對 TA-Lib
+func TestEMA_MatchesGoTalib(t *testing.T) {
+	prices := genPrices(300)
+	tol := 1e-7 // decimal<->float64 轉換微誤差
+
+	periods := []uint{3, 5, 12, 26}
+	for _, p := range periods {
+		t.Run("EMA_p="+decimal.NewFromInt(int64(p)).String(), func(t *testing.T) {
+			ema := NewEMA(p) // 重點：用 TA-Lib 兼容 seed
+
+			our := make([]decimal.Decimal, len(prices))
+			ready := make([]bool, len(prices))
+			for i, px := range prices {
+				out := ema.Update(dv(px))
+				our[i] = out.Value
+				ready[i] = out.Ready
+			}
+			ref := talib.Ema(prices, int(p))
+
+			start := p // talib 第一個有效值在 index = period-1
+
+			for i := start; i < uint(len(prices)); i++ {
+				if !ready[i] {
+					t.Fatalf("i=%d: our not ready but talib has value", i)
+				}
+				if !almostEqualDecFloat(our[i], ref[i], tol) {
+					t.Fatalf("i=%d: EMA mismatch: our=%s ref=%f", i, our[i], ref[i])
+				}
+			}
+		})
+	}
+}
+
+func TestSMA_MatchesGoTalib(t *testing.T) {
+	prices := genPrices(300)
+	tol := 1e-9
+
+	windows := []uint{3, 5, 20, 50}
+	for _, w := range windows {
+		t.Run("SMA_w="+decimal.NewFromInt(int64(w)).String(), func(t *testing.T) {
+			sma := NewSMA(w)
+
+			our := make([]decimal.Decimal, len(prices))
+			ready := make([]bool, len(prices))
+			for i, px := range prices {
+				out := sma.Update(dv(px))
+				our[i] = out.Value
+				ready[i] = out.Ready
+			}
+
+			ref := talib.Sma(prices, int(w))
+			start := w
+			for i := start; i < uint(len(prices)); i++ {
+				if !ready[i] {
+					t.Fatalf("i=%d: our not ready but talib has value", i)
+				}
+				if !almostEqualDecFloat(our[i], ref[i], tol) {
+					t.Fatalf("i=%d: SMA mismatch: our=%s ref=%f", i, our[i], ref[i])
+				}
+			}
+		})
+	}
+}

internal/lib/strategy/rsi.go

@@ -0,0 +1,62 @@
+package strategy
+
+import "github.com/shopspring/decimal"
+
+// RSI 使用 Wilder 的平均漲跌（非簡單平均），更貼近交易軟體常見計法
+type RSI struct {
+	n         int
+	prevC     decimal.Decimal // 前一根收盤價
+	initCount int             // 初始化用：先累積前 n 根的總漲/總跌
+	avgGain   decimal.Decimal // 平滑後的平均上漲
+	avgLoss   decimal.Decimal // 平滑後的平均下跌
+	ok        bool            // 是否有 prevC
+}
+
+func NewRSI(n int) *RSI { return &RSI{n: n} }
+
+// Push 餵入一根K線，回傳 (RSI值, 是否就緒)
+// 注意：前 n 根會回傳就緒=false；之後才可信
+func (r *RSI) Push(c CandleForStrategy) (decimal.Decimal, bool) {
+	if !r.ok {
+		r.prevC = c.C
+		r.ok = true
+		return decimal.Zero, false
+	}
+	// 價差
+	chg := c.C.Sub(r.prevC)
+	r.prevC = c.C
+
+	// 區分上漲與下跌
+	gain := decimal.Max(chg, decimal.Zero)
+	loss := decimal.Max(chg.Neg(), decimal.Zero)
+
+	// 初始化階段：先把前 n 根的平均值建好
+	if r.initCount < r.n {
+		r.avgGain = r.avgGain.Add(gain)
+		r.avgLoss = r.avgLoss.Add(loss)
+		r.initCount++
+		if r.initCount == r.n {
+			r.avgGain = r.avgGain.Div(decimal.NewFromInt(int64(r.n)))
+			r.avgLoss = r.avgLoss.Div(decimal.NewFromInt(int64(r.n)))
+			return r.calc(), true
+		}
+		return decimal.Zero, false
+	}
+
+	// Wilder 平滑：新的平均 = (舊平均*(n-1) + 當期值) / n
+	nDec := decimal.NewFromInt(int64(r.n))
+	r.avgGain = (r.avgGain.Mul(nDec.Sub(decimal.NewFromInt(1))).Add(gain)).Div(nDec)
+	r.avgLoss = (r.avgLoss.Mul(nDec.Sub(decimal.NewFromInt(1))).Add(loss)).Div(nDec)
+
+	return r.calc(), true
+}
+
+func (r *RSI) calc() decimal.Decimal {
+	if r.avgLoss.IsZero() {
+		return decimal.NewFromInt(100) // 沒有下跌時，RSI=100
+	}
+	rs := r.avgGain.Div(r.avgLoss)
+	one := decimal.NewFromInt(1)
+	hundred := decimal.NewFromInt(100)
+	return hundred.Sub(hundred.Div(one.Add(rs)))
+}

internal/lib/strategy/rsi_test.go

@@ -0,0 +1,71 @@
+package strategy
+
+import (
+	"github.com/shopspring/decimal"
+	"testing"
+)
+
+func TestRSI_WarmupAndRange(t *testing.T) {
+	n := 2
+	r := NewRSI(n)
+
+	type step struct {
+		close     float64
+		wantReady bool
+	}
+	steps := []step{
+		{10, false}, // 第一根，只建立 prevC
+		{11, false}, // 第二根，累積
+		{12, true},  // 第三根，完成 seed -> ready
+		{13, true},
+		{12, true},
+		{11, true},
+		{12, true},
+		{13, true},
+	}
+
+	for i, st := range steps {
+		val, ready := r.Push(CandleForStrategy{C: dv(st.close)})
+		if ready != st.wantReady {
+			t.Fatalf("step %d: ready got=%v want=%v", i, ready, st.wantReady)
+		}
+		if ready {
+			// RSI 應落在 0..100
+			if val.LessThan(decimal.Zero) || val.GreaterThan(decimal.NewFromInt(100)) {
+				t.Fatalf("step %d: RSI out of [0,100], got %s", i, val)
+			}
+		}
+	}
+}
+
+func TestRSI_PerfectUpAndDown(t *testing.T) {
+	n := 2
+	r := NewRSI(n)
+
+	// 先 seed
+	_, _ = r.Push(CandleForStrategy{C: d(10)})
+	_, _ = r.Push(CandleForStrategy{C: d(11)})
+	val, ready := r.Push(CandleForStrategy{C: d(12)})
+	if !ready {
+		t.Fatalf("should be ready after %d candles", n)
+	}
+
+	// 連續上漲：理想情況 avgLoss -> 0，RSI 逼近 100
+	val, ready = r.Push(CandleForStrategy{C: d(13)})
+	if !ready || !val.LessThanOrEqual(decimal.NewFromInt(100)) {
+		t.Fatalf("uptrend: ready=%v val=%s", ready, val)
+	}
+
+	// 連續下跌：理想情況 avgGain -> 0，RSI 逼近 0
+	r = NewRSI(n)
+	_, _ = r.Push(CandleForStrategy{C: d(13)})
+	_, _ = r.Push(CandleForStrategy{C: d(12)})
+	val, ready = r.Push(CandleForStrategy{C: d(11)})
+	if !ready {
+		t.Fatalf("should be ready after %d candles", n)
+	}
+	val, ready = r.Push(CandleForStrategy{C: d(10)})
+	if !ready || !val.GreaterThanOrEqual(decimal.Zero) {
+		t.Fatalf("downtrend: ready=%v val=%s", ready, val)
+	}
+}