556 lines
12 KiB
Go
556 lines
12 KiB
Go
// Copyright 2016, Circonus, Inc. All rights reserved.
|
|
// See the LICENSE file.
|
|
|
|
// Package circllhist provides an implementation of Circonus' fixed log-linear
|
|
// histogram data structure. This allows tracking of histograms in a
|
|
// composable way such that accurate error can be reasoned about.
|
|
package circonusllhist
|
|
|
|
import (
|
|
"bytes"
|
|
"errors"
|
|
"fmt"
|
|
"math"
|
|
"sync"
|
|
)
|
|
|
|
const (
|
|
DEFAULT_HIST_SIZE = int16(100)
|
|
)
|
|
|
|
var power_of_ten = [...]float64{
|
|
1, 10, 100, 1000, 10000, 100000, 1e+06, 1e+07, 1e+08, 1e+09, 1e+10,
|
|
1e+11, 1e+12, 1e+13, 1e+14, 1e+15, 1e+16, 1e+17, 1e+18, 1e+19, 1e+20,
|
|
1e+21, 1e+22, 1e+23, 1e+24, 1e+25, 1e+26, 1e+27, 1e+28, 1e+29, 1e+30,
|
|
1e+31, 1e+32, 1e+33, 1e+34, 1e+35, 1e+36, 1e+37, 1e+38, 1e+39, 1e+40,
|
|
1e+41, 1e+42, 1e+43, 1e+44, 1e+45, 1e+46, 1e+47, 1e+48, 1e+49, 1e+50,
|
|
1e+51, 1e+52, 1e+53, 1e+54, 1e+55, 1e+56, 1e+57, 1e+58, 1e+59, 1e+60,
|
|
1e+61, 1e+62, 1e+63, 1e+64, 1e+65, 1e+66, 1e+67, 1e+68, 1e+69, 1e+70,
|
|
1e+71, 1e+72, 1e+73, 1e+74, 1e+75, 1e+76, 1e+77, 1e+78, 1e+79, 1e+80,
|
|
1e+81, 1e+82, 1e+83, 1e+84, 1e+85, 1e+86, 1e+87, 1e+88, 1e+89, 1e+90,
|
|
1e+91, 1e+92, 1e+93, 1e+94, 1e+95, 1e+96, 1e+97, 1e+98, 1e+99, 1e+100,
|
|
1e+101, 1e+102, 1e+103, 1e+104, 1e+105, 1e+106, 1e+107, 1e+108, 1e+109,
|
|
1e+110, 1e+111, 1e+112, 1e+113, 1e+114, 1e+115, 1e+116, 1e+117, 1e+118,
|
|
1e+119, 1e+120, 1e+121, 1e+122, 1e+123, 1e+124, 1e+125, 1e+126, 1e+127,
|
|
1e-128, 1e-127, 1e-126, 1e-125, 1e-124, 1e-123, 1e-122, 1e-121, 1e-120,
|
|
1e-119, 1e-118, 1e-117, 1e-116, 1e-115, 1e-114, 1e-113, 1e-112, 1e-111,
|
|
1e-110, 1e-109, 1e-108, 1e-107, 1e-106, 1e-105, 1e-104, 1e-103, 1e-102,
|
|
1e-101, 1e-100, 1e-99, 1e-98, 1e-97, 1e-96,
|
|
1e-95, 1e-94, 1e-93, 1e-92, 1e-91, 1e-90, 1e-89, 1e-88, 1e-87, 1e-86,
|
|
1e-85, 1e-84, 1e-83, 1e-82, 1e-81, 1e-80, 1e-79, 1e-78, 1e-77, 1e-76,
|
|
1e-75, 1e-74, 1e-73, 1e-72, 1e-71, 1e-70, 1e-69, 1e-68, 1e-67, 1e-66,
|
|
1e-65, 1e-64, 1e-63, 1e-62, 1e-61, 1e-60, 1e-59, 1e-58, 1e-57, 1e-56,
|
|
1e-55, 1e-54, 1e-53, 1e-52, 1e-51, 1e-50, 1e-49, 1e-48, 1e-47, 1e-46,
|
|
1e-45, 1e-44, 1e-43, 1e-42, 1e-41, 1e-40, 1e-39, 1e-38, 1e-37, 1e-36,
|
|
1e-35, 1e-34, 1e-33, 1e-32, 1e-31, 1e-30, 1e-29, 1e-28, 1e-27, 1e-26,
|
|
1e-25, 1e-24, 1e-23, 1e-22, 1e-21, 1e-20, 1e-19, 1e-18, 1e-17, 1e-16,
|
|
1e-15, 1e-14, 1e-13, 1e-12, 1e-11, 1e-10, 1e-09, 1e-08, 1e-07, 1e-06,
|
|
1e-05, 0.0001, 0.001, 0.01, 0.1,
|
|
}
|
|
|
|
// A Bracket is a part of a cumulative distribution.
|
|
type Bin struct {
|
|
val int8
|
|
exp int8
|
|
count uint64
|
|
}
|
|
|
|
func NewBinRaw(val int8, exp int8, count uint64) *Bin {
|
|
return &Bin{
|
|
val: val,
|
|
exp: exp,
|
|
count: count,
|
|
}
|
|
}
|
|
func NewBin() *Bin {
|
|
return NewBinRaw(0, 0, 0)
|
|
}
|
|
func NewBinFromFloat64(d float64) *Bin {
|
|
hb := NewBinRaw(0, 0, 0)
|
|
hb.SetFromFloat64(d)
|
|
return hb
|
|
}
|
|
func (hb *Bin) SetFromFloat64(d float64) *Bin {
|
|
hb.val = -1
|
|
if math.IsInf(d, 0) || math.IsNaN(d) {
|
|
return hb
|
|
}
|
|
if d == 0.0 {
|
|
hb.val = 0
|
|
return hb
|
|
}
|
|
sign := 1
|
|
if math.Signbit(d) {
|
|
sign = -1
|
|
}
|
|
d = math.Abs(d)
|
|
big_exp := int(math.Floor(math.Log10(d)))
|
|
hb.exp = int8(big_exp)
|
|
if int(hb.exp) != big_exp { //rolled
|
|
hb.exp = 0
|
|
if big_exp < 0 {
|
|
hb.val = 0
|
|
}
|
|
return hb
|
|
}
|
|
d = d / hb.PowerOfTen()
|
|
d = d * 10
|
|
hb.val = int8(sign * int(math.Floor(d+1e-13)))
|
|
if hb.val == 100 || hb.val == -100 {
|
|
if hb.exp < 127 {
|
|
hb.val = hb.val / 10
|
|
hb.exp++
|
|
} else {
|
|
hb.val = 0
|
|
hb.exp = 0
|
|
}
|
|
}
|
|
if hb.val == 0 {
|
|
hb.exp = 0
|
|
return hb
|
|
}
|
|
if !((hb.val >= 10 && hb.val < 100) ||
|
|
(hb.val <= -10 && hb.val > -100)) {
|
|
hb.val = -1
|
|
hb.exp = 0
|
|
}
|
|
return hb
|
|
}
|
|
func (hb *Bin) PowerOfTen() float64 {
|
|
idx := int(hb.exp)
|
|
if idx < 0 {
|
|
idx = 256 + idx
|
|
}
|
|
return power_of_ten[idx]
|
|
}
|
|
|
|
func (hb *Bin) IsNaN() bool {
|
|
if hb.val > 99 || hb.val < -99 {
|
|
return true
|
|
}
|
|
return false
|
|
}
|
|
func (hb *Bin) Val() int8 {
|
|
return hb.val
|
|
}
|
|
func (hb *Bin) Exp() int8 {
|
|
return hb.exp
|
|
}
|
|
func (hb *Bin) Count() uint64 {
|
|
return hb.count
|
|
}
|
|
func (hb *Bin) Value() float64 {
|
|
if hb.IsNaN() {
|
|
return math.NaN()
|
|
}
|
|
if hb.val < 10 && hb.val > -10 {
|
|
return 0.0
|
|
}
|
|
return (float64(hb.val) / 10.0) * hb.PowerOfTen()
|
|
}
|
|
func (hb *Bin) BinWidth() float64 {
|
|
if hb.IsNaN() {
|
|
return math.NaN()
|
|
}
|
|
if hb.val < 10 && hb.val > -10 {
|
|
return 0.0
|
|
}
|
|
return hb.PowerOfTen() / 10.0
|
|
}
|
|
func (hb *Bin) Midpoint() float64 {
|
|
if hb.IsNaN() {
|
|
return math.NaN()
|
|
}
|
|
out := hb.Value()
|
|
if out == 0 {
|
|
return 0
|
|
}
|
|
interval := hb.BinWidth()
|
|
if out < 0 {
|
|
interval = interval * -1
|
|
}
|
|
return out + interval/2.0
|
|
}
|
|
func (hb *Bin) Left() float64 {
|
|
if hb.IsNaN() {
|
|
return math.NaN()
|
|
}
|
|
out := hb.Value()
|
|
if out >= 0 {
|
|
return out
|
|
}
|
|
return out - hb.BinWidth()
|
|
}
|
|
|
|
func (h1 *Bin) Compare(h2 *Bin) int {
|
|
if h1.val == h2.val && h1.exp == h2.exp {
|
|
return 0
|
|
}
|
|
if h1.val == -1 {
|
|
return 1
|
|
}
|
|
if h2.val == -1 {
|
|
return -1
|
|
}
|
|
if h1.val == 0 {
|
|
if h2.val > 0 {
|
|
return 1
|
|
}
|
|
return -1
|
|
}
|
|
if h2.val == 0 {
|
|
if h1.val < 0 {
|
|
return 1
|
|
}
|
|
return -1
|
|
}
|
|
if h1.val < 0 && h2.val > 0 {
|
|
return 1
|
|
}
|
|
if h1.val > 0 && h2.val < 0 {
|
|
return -1
|
|
}
|
|
if h1.exp == h2.exp {
|
|
if h1.val < h2.val {
|
|
return 1
|
|
}
|
|
return -1
|
|
}
|
|
if h1.exp > h2.exp {
|
|
if h1.val < 0 {
|
|
return 1
|
|
}
|
|
return -1
|
|
}
|
|
if h1.exp < h2.exp {
|
|
if h1.val < 0 {
|
|
return -1
|
|
}
|
|
return 1
|
|
}
|
|
return 0
|
|
}
|
|
|
|
// This histogram structure tracks values are two decimal digits of precision
|
|
// with a bounded error that remains bounded upon composition
|
|
type Histogram struct {
|
|
mutex sync.Mutex
|
|
bvs []Bin
|
|
used int16
|
|
allocd int16
|
|
}
|
|
|
|
// New returns a new Histogram
|
|
func New() *Histogram {
|
|
return &Histogram{
|
|
allocd: DEFAULT_HIST_SIZE,
|
|
used: 0,
|
|
bvs: make([]Bin, DEFAULT_HIST_SIZE),
|
|
}
|
|
}
|
|
|
|
// Max returns the approximate maximum recorded value.
|
|
func (h *Histogram) Max() float64 {
|
|
return h.ValueAtQuantile(1.0)
|
|
}
|
|
|
|
// Min returns the approximate minimum recorded value.
|
|
func (h *Histogram) Min() float64 {
|
|
return h.ValueAtQuantile(0.0)
|
|
}
|
|
|
|
// Mean returns the approximate arithmetic mean of the recorded values.
|
|
func (h *Histogram) Mean() float64 {
|
|
return h.ApproxMean()
|
|
}
|
|
|
|
// Reset forgets all bins in the histogram (they remain allocated)
|
|
func (h *Histogram) Reset() {
|
|
h.mutex.Lock()
|
|
h.used = 0
|
|
h.mutex.Unlock()
|
|
}
|
|
|
|
// RecordValue records the given value, returning an error if the value is out
|
|
// of range.
|
|
func (h *Histogram) RecordValue(v float64) error {
|
|
return h.RecordValues(v, 1)
|
|
}
|
|
|
|
// RecordCorrectedValue records the given value, correcting for stalls in the
|
|
// recording process. This only works for processes which are recording values
|
|
// at an expected interval (e.g., doing jitter analysis). Processes which are
|
|
// recording ad-hoc values (e.g., latency for incoming requests) can't take
|
|
// advantage of this.
|
|
// CH Compat
|
|
func (h *Histogram) RecordCorrectedValue(v, expectedInterval int64) error {
|
|
if err := h.RecordValue(float64(v)); err != nil {
|
|
return err
|
|
}
|
|
|
|
if expectedInterval <= 0 || v <= expectedInterval {
|
|
return nil
|
|
}
|
|
|
|
missingValue := v - expectedInterval
|
|
for missingValue >= expectedInterval {
|
|
if err := h.RecordValue(float64(missingValue)); err != nil {
|
|
return err
|
|
}
|
|
missingValue -= expectedInterval
|
|
}
|
|
|
|
return nil
|
|
}
|
|
|
|
// find where a new bin should go
|
|
func (h *Histogram) InternalFind(hb *Bin) (bool, int16) {
|
|
if h.used == 0 {
|
|
return false, 0
|
|
}
|
|
rv := -1
|
|
idx := int16(0)
|
|
l := int16(0)
|
|
r := h.used - 1
|
|
for l < r {
|
|
check := (r + l) / 2
|
|
rv = h.bvs[check].Compare(hb)
|
|
if rv == 0 {
|
|
l = check
|
|
r = check
|
|
} else if rv > 0 {
|
|
l = check + 1
|
|
} else {
|
|
r = check - 1
|
|
}
|
|
}
|
|
if rv != 0 {
|
|
rv = h.bvs[l].Compare(hb)
|
|
}
|
|
idx = l
|
|
if rv == 0 {
|
|
return true, idx
|
|
}
|
|
if rv < 0 {
|
|
return false, idx
|
|
}
|
|
idx++
|
|
return false, idx
|
|
}
|
|
|
|
func (h *Histogram) InsertBin(hb *Bin, count int64) uint64 {
|
|
h.mutex.Lock()
|
|
defer h.mutex.Unlock()
|
|
if count == 0 {
|
|
return 0
|
|
}
|
|
found, idx := h.InternalFind(hb)
|
|
if !found {
|
|
if h.used == h.allocd {
|
|
new_bvs := make([]Bin, h.allocd+DEFAULT_HIST_SIZE)
|
|
if idx > 0 {
|
|
copy(new_bvs[0:], h.bvs[0:idx])
|
|
}
|
|
if idx < h.used {
|
|
copy(new_bvs[idx+1:], h.bvs[idx:])
|
|
}
|
|
h.allocd = h.allocd + DEFAULT_HIST_SIZE
|
|
h.bvs = new_bvs
|
|
} else {
|
|
copy(h.bvs[idx+1:], h.bvs[idx:h.used])
|
|
}
|
|
h.bvs[idx].val = hb.val
|
|
h.bvs[idx].exp = hb.exp
|
|
h.bvs[idx].count = uint64(count)
|
|
h.used++
|
|
return h.bvs[idx].count
|
|
}
|
|
var newval uint64
|
|
if count < 0 {
|
|
newval = h.bvs[idx].count - uint64(-count)
|
|
} else {
|
|
newval = h.bvs[idx].count + uint64(count)
|
|
}
|
|
if newval < h.bvs[idx].count { //rolled
|
|
newval = ^uint64(0)
|
|
}
|
|
h.bvs[idx].count = newval
|
|
return newval - h.bvs[idx].count
|
|
}
|
|
|
|
// RecordValues records n occurrences of the given value, returning an error if
|
|
// the value is out of range.
|
|
func (h *Histogram) RecordValues(v float64, n int64) error {
|
|
var hb Bin
|
|
hb.SetFromFloat64(v)
|
|
h.InsertBin(&hb, n)
|
|
return nil
|
|
}
|
|
|
|
// Approximate mean
|
|
func (h *Histogram) ApproxMean() float64 {
|
|
h.mutex.Lock()
|
|
defer h.mutex.Unlock()
|
|
divisor := 0.0
|
|
sum := 0.0
|
|
for i := int16(0); i < h.used; i++ {
|
|
midpoint := h.bvs[i].Midpoint()
|
|
cardinality := float64(h.bvs[i].count)
|
|
divisor += cardinality
|
|
sum += midpoint * cardinality
|
|
}
|
|
if divisor == 0.0 {
|
|
return math.NaN()
|
|
}
|
|
return sum / divisor
|
|
}
|
|
|
|
// Approximate sum
|
|
func (h *Histogram) ApproxSum() float64 {
|
|
h.mutex.Lock()
|
|
defer h.mutex.Unlock()
|
|
sum := 0.0
|
|
for i := int16(0); i < h.used; i++ {
|
|
midpoint := h.bvs[i].Midpoint()
|
|
cardinality := float64(h.bvs[i].count)
|
|
sum += midpoint * cardinality
|
|
}
|
|
return sum
|
|
}
|
|
|
|
func (h *Histogram) ApproxQuantile(q_in []float64) ([]float64, error) {
|
|
h.mutex.Lock()
|
|
defer h.mutex.Unlock()
|
|
q_out := make([]float64, len(q_in))
|
|
i_q, i_b := 0, int16(0)
|
|
total_cnt, bin_width, bin_left, lower_cnt, upper_cnt := 0.0, 0.0, 0.0, 0.0, 0.0
|
|
if len(q_in) == 0 {
|
|
return q_out, nil
|
|
}
|
|
// Make sure the requested quantiles are in order
|
|
for i_q = 1; i_q < len(q_in); i_q++ {
|
|
if q_in[i_q-1] > q_in[i_q] {
|
|
return nil, errors.New("out of order")
|
|
}
|
|
}
|
|
// Add up the bins
|
|
for i_b = 0; i_b < h.used; i_b++ {
|
|
if !h.bvs[i_b].IsNaN() {
|
|
total_cnt += float64(h.bvs[i_b].count)
|
|
}
|
|
}
|
|
if total_cnt == 0.0 {
|
|
return nil, errors.New("empty_histogram")
|
|
}
|
|
|
|
for i_q = 0; i_q < len(q_in); i_q++ {
|
|
if q_in[i_q] < 0.0 || q_in[i_q] > 1.0 {
|
|
return nil, errors.New("out of bound quantile")
|
|
}
|
|
q_out[i_q] = total_cnt * q_in[i_q]
|
|
}
|
|
|
|
for i_b = 0; i_b < h.used; i_b++ {
|
|
if h.bvs[i_b].IsNaN() {
|
|
continue
|
|
}
|
|
bin_width = h.bvs[i_b].BinWidth()
|
|
bin_left = h.bvs[i_b].Left()
|
|
lower_cnt = upper_cnt
|
|
upper_cnt = lower_cnt + float64(h.bvs[i_b].count)
|
|
break
|
|
}
|
|
for i_q = 0; i_q < len(q_in); i_q++ {
|
|
for i_b < (h.used-1) && upper_cnt < q_out[i_q] {
|
|
i_b++
|
|
bin_width = h.bvs[i_b].BinWidth()
|
|
bin_left = h.bvs[i_b].Left()
|
|
lower_cnt = upper_cnt
|
|
upper_cnt = lower_cnt + float64(h.bvs[i_b].count)
|
|
}
|
|
if lower_cnt == q_out[i_q] {
|
|
q_out[i_q] = bin_left
|
|
} else if upper_cnt == q_out[i_q] {
|
|
q_out[i_q] = bin_left + bin_width
|
|
} else {
|
|
if bin_width == 0 {
|
|
q_out[i_q] = bin_left
|
|
} else {
|
|
q_out[i_q] = bin_left + (q_out[i_q]-lower_cnt)/(upper_cnt-lower_cnt)*bin_width
|
|
}
|
|
}
|
|
}
|
|
return q_out, nil
|
|
}
|
|
|
|
// ValueAtQuantile returns the recorded value at the given quantile (0..1).
|
|
func (h *Histogram) ValueAtQuantile(q float64) float64 {
|
|
h.mutex.Lock()
|
|
defer h.mutex.Unlock()
|
|
q_in := make([]float64, 1)
|
|
q_in[0] = q
|
|
q_out, err := h.ApproxQuantile(q_in)
|
|
if err == nil && len(q_out) == 1 {
|
|
return q_out[0]
|
|
}
|
|
return math.NaN()
|
|
}
|
|
|
|
// SignificantFigures returns the significant figures used to create the
|
|
// histogram
|
|
// CH Compat
|
|
func (h *Histogram) SignificantFigures() int64 {
|
|
return 2
|
|
}
|
|
|
|
// Equals returns true if the two Histograms are equivalent, false if not.
|
|
func (h *Histogram) Equals(other *Histogram) bool {
|
|
h.mutex.Lock()
|
|
other.mutex.Lock()
|
|
defer h.mutex.Unlock()
|
|
defer other.mutex.Unlock()
|
|
switch {
|
|
case
|
|
h.used != other.used:
|
|
return false
|
|
default:
|
|
for i := int16(0); i < h.used; i++ {
|
|
if h.bvs[i].Compare(&other.bvs[i]) != 0 {
|
|
return false
|
|
}
|
|
if h.bvs[i].count != other.bvs[i].count {
|
|
return false
|
|
}
|
|
}
|
|
}
|
|
return true
|
|
}
|
|
|
|
func (h *Histogram) CopyAndReset() *Histogram {
|
|
h.mutex.Lock()
|
|
defer h.mutex.Unlock()
|
|
newhist := &Histogram{
|
|
allocd: h.allocd,
|
|
used: h.used,
|
|
bvs: h.bvs,
|
|
}
|
|
h.allocd = DEFAULT_HIST_SIZE
|
|
h.bvs = make([]Bin, DEFAULT_HIST_SIZE)
|
|
h.used = 0
|
|
return newhist
|
|
}
|
|
func (h *Histogram) DecStrings() []string {
|
|
h.mutex.Lock()
|
|
defer h.mutex.Unlock()
|
|
out := make([]string, h.used)
|
|
for i, bin := range h.bvs[0:h.used] {
|
|
var buffer bytes.Buffer
|
|
buffer.WriteString("H[")
|
|
buffer.WriteString(fmt.Sprintf("%3.1e", bin.Value()))
|
|
buffer.WriteString("]=")
|
|
buffer.WriteString(fmt.Sprintf("%v", bin.count))
|
|
out[i] = buffer.String()
|
|
}
|
|
return out
|
|
}
|