mirror of
https://github.com/c9s/bbgo.git
synced 2024-11-14 02:53:50 +00:00
613 lines
10 KiB
Go
613 lines
10 KiB
Go
//go:build !dnum
|
|
|
|
package fixedpoint
|
|
|
|
import (
|
|
"bytes"
|
|
"database/sql/driver"
|
|
"errors"
|
|
"fmt"
|
|
"math"
|
|
"strconv"
|
|
"strings"
|
|
"sync/atomic"
|
|
)
|
|
|
|
const MaxPrecision = 12
|
|
const DefaultPrecision = 8
|
|
|
|
const DefaultPow = 1e8
|
|
|
|
type Value int64
|
|
|
|
const Zero = Value(0)
|
|
const One = Value(1e8)
|
|
const NegOne = Value(-1e8)
|
|
const PosInf = Value(math.MaxInt64)
|
|
const NegInf = Value(math.MinInt64)
|
|
|
|
type RoundingMode int
|
|
|
|
const (
|
|
Up RoundingMode = iota
|
|
Down
|
|
HalfUp
|
|
)
|
|
|
|
// Trunc returns the integer portion (truncating any fractional part)
|
|
func (v Value) Trunc() Value {
|
|
return NewFromFloat(math.Floor(v.Float64()))
|
|
}
|
|
|
|
func (v Value) Round(r int, mode RoundingMode) Value {
|
|
pow := math.Pow10(r)
|
|
f := v.Float64() * pow
|
|
switch mode {
|
|
case Up:
|
|
f = math.Ceil(f) / pow
|
|
case HalfUp:
|
|
f = math.Floor(f+0.5) / pow
|
|
case Down:
|
|
f = math.Floor(f) / pow
|
|
}
|
|
|
|
s := strconv.FormatFloat(f, 'f', r, 64)
|
|
return MustNewFromString(s)
|
|
}
|
|
|
|
func (v Value) Value() (driver.Value, error) {
|
|
return v.Float64(), nil
|
|
}
|
|
|
|
func (v *Value) Scan(src interface{}) error {
|
|
switch d := src.(type) {
|
|
case int64:
|
|
*v = NewFromInt(d)
|
|
return nil
|
|
|
|
case float64:
|
|
*v = NewFromFloat(d)
|
|
return nil
|
|
|
|
case []byte:
|
|
vv, err := NewFromString(string(d))
|
|
if err != nil {
|
|
return err
|
|
}
|
|
*v = vv
|
|
return nil
|
|
|
|
default:
|
|
|
|
}
|
|
|
|
return fmt.Errorf("fixedpoint.Value scan error, type: %T is not supported, value; %+v", src, src)
|
|
}
|
|
|
|
func (v Value) Float64() float64 {
|
|
if v == PosInf {
|
|
return math.Inf(1)
|
|
} else if v == NegInf {
|
|
return math.Inf(-1)
|
|
}
|
|
return float64(v) / DefaultPow
|
|
}
|
|
|
|
func (v Value) Abs() Value {
|
|
if v < 0 {
|
|
return -v
|
|
}
|
|
return v
|
|
}
|
|
|
|
func (v Value) String() string {
|
|
if v == PosInf {
|
|
return "inf"
|
|
} else if v == NegInf {
|
|
return "-inf"
|
|
}
|
|
return strconv.FormatFloat(float64(v)/DefaultPow, 'f', -1, 64)
|
|
}
|
|
|
|
func (v Value) FormatString(prec int) string {
|
|
if v == PosInf {
|
|
return "inf"
|
|
} else if v == NegInf {
|
|
return "-inf"
|
|
}
|
|
|
|
u := int64(v)
|
|
|
|
// trunc precision
|
|
precDiff := DefaultPrecision - prec
|
|
if precDiff > 0 {
|
|
powDiff := int64(math.Round(math.Pow10(precDiff)))
|
|
u = int64(v) / powDiff * powDiff
|
|
}
|
|
|
|
// check sign
|
|
sign := Value(u).Sign()
|
|
|
|
basePow := int64(DefaultPow)
|
|
a := u / basePow
|
|
b := u % basePow
|
|
|
|
if a < 0 {
|
|
a = -a
|
|
}
|
|
|
|
if b < 0 {
|
|
b = -b
|
|
}
|
|
|
|
str := strconv.FormatInt(a, 10)
|
|
if prec > 0 {
|
|
bStr := fmt.Sprintf(".%08d", b)
|
|
if prec <= DefaultPrecision {
|
|
bStr = bStr[0 : prec+1]
|
|
} else {
|
|
for i := prec - DefaultPrecision; i > 0; i-- {
|
|
bStr += "0"
|
|
}
|
|
}
|
|
|
|
str += bStr
|
|
}
|
|
|
|
if sign < 0 {
|
|
str = "-" + str
|
|
}
|
|
|
|
return str
|
|
}
|
|
|
|
func (v Value) Percentage() string {
|
|
if v == 0 {
|
|
return "0"
|
|
}
|
|
if v == PosInf {
|
|
return "inf%"
|
|
} else if v == NegInf {
|
|
return "-inf%"
|
|
}
|
|
return strconv.FormatFloat(float64(v)/DefaultPow*100., 'f', -1, 64) + "%"
|
|
}
|
|
|
|
func (v Value) FormatPercentage(prec int) string {
|
|
if v == 0 {
|
|
return "0"
|
|
}
|
|
if v == PosInf {
|
|
return "inf%"
|
|
} else if v == NegInf {
|
|
return "-inf%"
|
|
}
|
|
pow := math.Pow10(prec)
|
|
result := strconv.FormatFloat(
|
|
math.Trunc(float64(v)/DefaultPow*pow*100.)/pow, 'f', prec, 64)
|
|
return result + "%"
|
|
}
|
|
|
|
func (v Value) SignedPercentage() string {
|
|
if v > 0 {
|
|
return "+" + v.Percentage()
|
|
}
|
|
return v.Percentage()
|
|
}
|
|
|
|
func (v Value) Int64() int64 {
|
|
return int64(v.Float64())
|
|
}
|
|
|
|
func (v Value) Int() int {
|
|
n := v.Int64()
|
|
if int64(int(n)) != n {
|
|
panic("unable to convert Value to int32")
|
|
}
|
|
return int(n)
|
|
}
|
|
|
|
func (v Value) Neg() Value {
|
|
return -v
|
|
}
|
|
|
|
// TODO inf
|
|
func (v Value) Sign() int {
|
|
if v > 0 {
|
|
return 1
|
|
} else if v == 0 {
|
|
return 0
|
|
} else {
|
|
return -1
|
|
}
|
|
}
|
|
|
|
func (v Value) IsZero() bool {
|
|
return v == 0
|
|
}
|
|
|
|
func Mul(x, y Value) Value {
|
|
return NewFromFloat(x.Float64() * y.Float64())
|
|
}
|
|
|
|
func (v Value) Mul(v2 Value) Value {
|
|
return NewFromFloat(v.Float64() * v2.Float64())
|
|
}
|
|
|
|
func Div(x, y Value) Value {
|
|
return NewFromFloat(x.Float64() / y.Float64())
|
|
}
|
|
|
|
func (v Value) Div(v2 Value) Value {
|
|
return NewFromFloat(v.Float64() / v2.Float64())
|
|
}
|
|
|
|
func (v Value) Floor() Value {
|
|
return NewFromFloat(math.Floor(v.Float64()))
|
|
}
|
|
|
|
func (v Value) Ceil() Value {
|
|
return NewFromFloat(math.Ceil(v.Float64()))
|
|
}
|
|
|
|
func (v Value) Sub(v2 Value) Value {
|
|
return Value(int64(v) - int64(v2))
|
|
}
|
|
|
|
func (v Value) Add(v2 Value) Value {
|
|
return Value(int64(v) + int64(v2))
|
|
}
|
|
|
|
func (v *Value) AtomicAdd(v2 Value) {
|
|
atomic.AddInt64((*int64)(v), int64(v2))
|
|
}
|
|
|
|
func (v *Value) AtomicLoad() Value {
|
|
i := atomic.LoadInt64((*int64)(v))
|
|
return Value(i)
|
|
}
|
|
|
|
func (v *Value) UnmarshalYAML(unmarshal func(a interface{}) error) (err error) {
|
|
var s string
|
|
if err = unmarshal(&s); err == nil {
|
|
nv, err2 := NewFromString(s)
|
|
if err2 == nil {
|
|
*v = nv
|
|
return
|
|
}
|
|
}
|
|
|
|
return err
|
|
}
|
|
|
|
func (v Value) MarshalYAML() (interface{}, error) {
|
|
return v.FormatString(DefaultPrecision), nil
|
|
}
|
|
|
|
func (v Value) MarshalJSON() ([]byte, error) {
|
|
if v.IsInf() {
|
|
return []byte("\"" + v.String() + "\""), nil
|
|
}
|
|
return []byte(v.FormatString(DefaultPrecision)), nil
|
|
}
|
|
|
|
func (v *Value) UnmarshalJSON(data []byte) error {
|
|
if bytes.Equal(data, []byte{'n', 'u', 'l', 'l'}) {
|
|
*v = Zero
|
|
return nil
|
|
}
|
|
if len(data) == 0 {
|
|
*v = Zero
|
|
return nil
|
|
}
|
|
var err error
|
|
if data[0] == '"' {
|
|
data = data[1 : len(data)-1]
|
|
}
|
|
if *v, err = NewFromString(string(data)); err != nil {
|
|
return err
|
|
}
|
|
return nil
|
|
}
|
|
|
|
var ErrPrecisionLoss = errors.New("precision loss")
|
|
|
|
func Parse(input string) (num int64, numDecimalPoints int, err error) {
|
|
length := len(input)
|
|
isPercentage := input[length-1] == '%'
|
|
if isPercentage {
|
|
length -= 1
|
|
input = input[0:length]
|
|
}
|
|
|
|
var neg int64 = 1
|
|
var digit int64
|
|
for i := 0; i < length; i++ {
|
|
c := input[i]
|
|
if c == '-' {
|
|
neg = -1
|
|
} else if c >= '0' && c <= '9' {
|
|
digit, err = strconv.ParseInt(string(c), 10, 64)
|
|
if err != nil {
|
|
return
|
|
}
|
|
|
|
num = num*10 + digit
|
|
} else if c == '.' {
|
|
i++
|
|
if i > len(input)-1 {
|
|
err = fmt.Errorf("expect fraction numbers after dot")
|
|
return
|
|
}
|
|
|
|
for j := i; j < len(input); j++ {
|
|
fc := input[j]
|
|
if fc >= '0' && fc <= '9' {
|
|
digit, err = strconv.ParseInt(string(fc), 10, 64)
|
|
if err != nil {
|
|
return
|
|
}
|
|
|
|
numDecimalPoints++
|
|
num = num*10 + digit
|
|
|
|
if numDecimalPoints >= MaxPrecision {
|
|
return num, numDecimalPoints, ErrPrecisionLoss
|
|
}
|
|
} else {
|
|
err = fmt.Errorf("expect digit, got %c", fc)
|
|
return
|
|
}
|
|
}
|
|
break
|
|
} else {
|
|
err = fmt.Errorf("unexpected char %c", c)
|
|
return
|
|
}
|
|
}
|
|
|
|
num = num * neg
|
|
if isPercentage {
|
|
numDecimalPoints += 2
|
|
}
|
|
|
|
return num, numDecimalPoints, nil
|
|
}
|
|
|
|
func NewFromString(input string) (Value, error) {
|
|
length := len(input)
|
|
|
|
if length == 0 {
|
|
return 0, nil
|
|
}
|
|
|
|
isPercentage := input[length-1] == '%'
|
|
if isPercentage {
|
|
input = input[0 : length-1]
|
|
}
|
|
dotIndex := -1
|
|
hasDecimal := false
|
|
decimalCount := 0
|
|
// if is decimal, we don't need this
|
|
hasScientificNotion := false
|
|
hasIChar := false
|
|
scIndex := -1
|
|
for i, c := range input {
|
|
if hasDecimal {
|
|
if c <= '9' && c >= '0' {
|
|
decimalCount++
|
|
} else {
|
|
break
|
|
}
|
|
|
|
} else if c == '.' {
|
|
dotIndex = i
|
|
hasDecimal = true
|
|
}
|
|
if c == 'e' || c == 'E' {
|
|
hasScientificNotion = true
|
|
scIndex = i
|
|
break
|
|
}
|
|
if c == 'i' || c == 'I' {
|
|
hasIChar = true
|
|
break
|
|
}
|
|
}
|
|
if hasDecimal {
|
|
after := input[dotIndex+1:]
|
|
if decimalCount >= 8 {
|
|
after = after[0:8] + "." + after[8:]
|
|
} else {
|
|
after = after[0:decimalCount] + strings.Repeat("0", 8-decimalCount) + after[decimalCount:]
|
|
}
|
|
input = input[0:dotIndex] + after
|
|
v, err := strconv.ParseFloat(input, 64)
|
|
if err != nil {
|
|
return 0, err
|
|
}
|
|
|
|
if isPercentage {
|
|
v = v * 0.01
|
|
}
|
|
|
|
return Value(int64(math.Trunc(v))), nil
|
|
|
|
} else if hasScientificNotion {
|
|
exp, err := strconv.ParseInt(input[scIndex+1:], 10, 32)
|
|
if err != nil {
|
|
return 0, err
|
|
}
|
|
v, err := strconv.ParseFloat(input[0:scIndex+1]+strconv.FormatInt(exp+8, 10), 64)
|
|
if err != nil {
|
|
return 0, err
|
|
}
|
|
return Value(int64(math.Trunc(v))), nil
|
|
} else if hasIChar {
|
|
if floatV, err := strconv.ParseFloat(input, 64); nil != err {
|
|
return 0, err
|
|
} else if math.IsInf(floatV, 1) {
|
|
return PosInf, nil
|
|
} else if math.IsInf(floatV, -1) {
|
|
return NegInf, nil
|
|
} else {
|
|
return 0, fmt.Errorf("fixedpoint.Value parse error, invalid input string %s", input)
|
|
}
|
|
} else {
|
|
v, err := strconv.ParseInt(input, 10, 64)
|
|
if err != nil {
|
|
return 0, err
|
|
}
|
|
if isPercentage {
|
|
v = v * DefaultPow / 100
|
|
} else {
|
|
v = v * DefaultPow
|
|
}
|
|
return Value(v), nil
|
|
}
|
|
}
|
|
|
|
func MustNewFromString(input string) Value {
|
|
v, err := NewFromString(input)
|
|
if err != nil {
|
|
panic(fmt.Errorf("can not parse %s into fixedpoint, error: %s", input, err.Error()))
|
|
}
|
|
return v
|
|
}
|
|
|
|
func NewFromBytes(input []byte) (Value, error) {
|
|
return NewFromString(string(input))
|
|
}
|
|
|
|
func MustNewFromBytes(input []byte) (v Value) {
|
|
var err error
|
|
if v, err = NewFromString(string(input)); err != nil {
|
|
return Zero
|
|
}
|
|
return v
|
|
}
|
|
|
|
func Must(v Value, err error) Value {
|
|
if err != nil {
|
|
panic(err)
|
|
}
|
|
return v
|
|
}
|
|
|
|
func NewFromFloat(val float64) Value {
|
|
if math.IsInf(val, 1) {
|
|
return PosInf
|
|
} else if math.IsInf(val, -1) {
|
|
return NegInf
|
|
}
|
|
return Value(int64(math.Trunc(val * DefaultPow)))
|
|
}
|
|
|
|
func NewFromInt(val int64) Value {
|
|
return Value(val * DefaultPow)
|
|
}
|
|
|
|
func (a Value) IsInf() bool {
|
|
return a == PosInf || a == NegInf
|
|
}
|
|
|
|
func (a Value) MulExp(exp int) Value {
|
|
return Value(int64(float64(a) * math.Pow(10, float64(exp))))
|
|
}
|
|
|
|
func (a Value) NumIntDigits() int {
|
|
digits := 0
|
|
target := int64(a)
|
|
for pow := int64(DefaultPow); pow <= target; pow *= 10 {
|
|
digits++
|
|
}
|
|
return digits
|
|
}
|
|
|
|
// TODO: speedup
|
|
func (a Value) NumFractionalDigits() int {
|
|
if a == 0 {
|
|
return 0
|
|
}
|
|
numPow := 0
|
|
for pow := int64(DefaultPow); pow%10 != 1; pow /= 10 {
|
|
numPow++
|
|
}
|
|
numZeros := 0
|
|
for v := int64(a); v%10 == 0; v /= 10 {
|
|
numZeros++
|
|
}
|
|
return numPow - numZeros
|
|
}
|
|
|
|
func Compare(x, y Value) int {
|
|
if x > y {
|
|
return 1
|
|
} else if x == y {
|
|
return 0
|
|
} else {
|
|
return -1
|
|
}
|
|
}
|
|
|
|
func (x Value) Compare(y Value) int {
|
|
if x > y {
|
|
return 1
|
|
} else if x == y {
|
|
return 0
|
|
} else {
|
|
return -1
|
|
}
|
|
}
|
|
|
|
func Min(a, b Value) Value {
|
|
if a < b {
|
|
return a
|
|
}
|
|
|
|
return b
|
|
}
|
|
|
|
func Max(a, b Value) Value {
|
|
if a > b {
|
|
return a
|
|
}
|
|
|
|
return b
|
|
}
|
|
|
|
func Equal(x, y Value) bool {
|
|
return x == y
|
|
}
|
|
|
|
func (x Value) Eq(y Value) bool {
|
|
return x == y
|
|
}
|
|
|
|
func Abs(a Value) Value {
|
|
if a < 0 {
|
|
return -a
|
|
}
|
|
return a
|
|
}
|
|
|
|
func Clamp(x, min, max Value) Value {
|
|
if x < min {
|
|
return min
|
|
}
|
|
if x > max {
|
|
return max
|
|
}
|
|
return x
|
|
}
|
|
|
|
func (x Value) Clamp(min, max Value) Value {
|
|
if x < min {
|
|
return min
|
|
}
|
|
if x > max {
|
|
return max
|
|
}
|
|
return x
|
|
}
|