stats

Documentation

Index

• func ParsePercentiles(percentiles string) ([]float64, error)
• func Round(v float64) float64
• func RoundToDigits(v float64, digits int) float64
• type Bucket
• type Counter
  • func (c *Counter) Avg() float64
  • func (c *Counter) Log(msg string)
  • func (c *Counter) Print(out io.Writer, msg string)
  • func (c *Counter) Record(v float64)
  • func (c *Counter) RecordN(v float64, n int)
  • func (c *Counter) Reset()
  • func (c *Counter) StdDev() float64
  • func (c *Counter) Transfer(src *Counter)
• type Histogram
  • func Merge(h1 *Histogram, h2 *Histogram) *Histogram
  • func NewHistogram(Offset float64, Divider float64) *Histogram
  • func (h *Histogram) Clone() *Histogram
  • func (h *Histogram) CopyFrom(src *Histogram)
  • func (h *Histogram) Export() *HistogramData
  • func (h *Histogram) Log(msg string, percentiles []float64)
  • func (h *Histogram) Print(out io.Writer, msg string, percentiles []float64)
  • func (h *Histogram) Record(v float64)
  • func (h *Histogram) RecordN(v float64, n int)
  • func (h *Histogram) Reset()
  • func (h *Histogram) Transfer(src *Histogram)
• type HistogramData
  • func (e *HistogramData) CalcPercentile(percentile float64) float64
  • func (e *HistogramData) CalcPercentiles(percentiles []float64) *HistogramData
  • func (e *HistogramData) Print(out io.Writer, msg string)
• type Interval
• type Percentile

Functions

func ParsePercentiles

func ParsePercentiles(percentiles string) ([]float64, error)

ParsePercentiles extracts the percentiles from string (flag).

func Round

func Round(v float64) float64

Round rounds to 4 digits after the decimal point.

func RoundToDigits

func RoundToDigits(v float64, digits int) float64

RoundToDigits rounds the input to digits number of digits after decimal point. Note this incorrectly rounds the last digit of negative numbers.

Types

type Bucket

type Bucket struct {
	Interval
	Percent float64 // Cumulative percentile
	Count   int64   // How many in this bucket
}

Bucket is the data for 1 bucket: an Interval and the occurrence Count for that interval.

type Counter

type Counter struct {
	Count int64
	Min   float64
	Max   float64
	Sum   float64
	// contains filtered or unexported fields
}

Counter is a type whose instances record values and calculate stats (count,average,min,max,stddev).

func (*Counter) Avg

func (c *Counter) Avg() float64

Avg returns the average.

func (*Counter) Log

func (c *Counter) Log(msg string)

Log outputs the stats to the logger.

func (*Counter) Print

func (c *Counter) Print(out io.Writer, msg string)

Print prints stats.

func (*Counter) Record

func (c *Counter) Record(v float64)

Record records a data point.

func (*Counter) RecordN

func (c *Counter) RecordN(v float64, n int)

RecordN efficiently records the same value N times

func (*Counter) Reset

func (c *Counter) Reset()

Reset clears the counter to reset it to original 'no data' state.

func (*Counter) StdDev

func (c *Counter) StdDev() float64

StdDev returns the standard deviation.

func (*Counter) Transfer

func (c *Counter) Transfer(src *Counter)

Transfer merges the data from src into this Counter and clears src.

type Histogram

type Histogram struct {
	Counter
	Offset  float64 // offset applied to data before fitting into buckets
	Divider float64 // divider applied to data before fitting into buckets
	// Don't access directly (outside of this package):
	Hdata []int32 // numValues buckets (one more than values, for last one)
}

Histogram extends Counter and adds an histogram. Must be created using NewHistogram or anotherHistogram.Clone() and not directly.

func Merge

func Merge(h1 *Histogram, h2 *Histogram) *Histogram

Merge two different histogram with different scale parameters Lowest offset and highest divider value will be selected on new Histogram as scale parameters

func NewHistogram

func NewHistogram(Offset float64, Divider float64) *Histogram

NewHistogram creates a new histogram (sets up the buckets). Divider value can not be zero, otherwise returns zero

func (*Histogram) Clone

func (h *Histogram) Clone() *Histogram

Clone returns a copy of the histogram.

func (*Histogram) CopyFrom

func (h *Histogram) CopyFrom(src *Histogram)

CopyFrom sets the content of this object to a copy of the src.

func (*Histogram) Export

func (h *Histogram) Export() *HistogramData

Export translate the internal representation of the histogram data in an externally usable one. Calculates the request Percentiles.

func (*Histogram) Log

func (h *Histogram) Log(msg string, percentiles []float64)

Log Logs the histogram to the counter.

func (*Histogram) Print

func (h *Histogram) Print(out io.Writer, msg string, percentiles []float64)

Print dumps the histogram (and counter) to the provided writer. Also calculates the percentiles. Use Export() once and Print if you are going to need the Export results too.

func (*Histogram) Record

func (h *Histogram) Record(v float64)

Record records a data point.

func (*Histogram) RecordN

func (h *Histogram) RecordN(v float64, n int)

RecordN efficiently records a data point N times.

func (*Histogram) Reset

func (h *Histogram) Reset()

Reset clears the data. Reset it to NewHistogram state.

func (*Histogram) Transfer

func (h *Histogram) Transfer(src *Histogram)

Transfer merges the data from src into this Histogram and clears src.

type HistogramData

type HistogramData struct {
	Count       int64
	Min         float64
	Max         float64
	Sum         float64
	Avg         float64
	StdDev      float64
	Data        []Bucket
	Percentiles []Percentile
}

HistogramData is the exported Histogram data, a sorted list of intervals covering [Min, Max]. Pure data, so Counter for instance is flattened

func (*HistogramData) CalcPercentile

func (e *HistogramData) CalcPercentile(percentile float64) float64

CalcPercentile returns the value for an input percentile e.g. for 90. as input returns an estimate of the original value threshold where 90.0% of the data is below said threshold. with 3 data points 10, 20, 30; p0-p33.33 == 10, p 66.666 = 20, p100 = 30 p33.333 - p66.666 = linear between 10 and 20; so p50 = 15 TODO: consider spreading the count of the bucket evenly from start to end so the % grows by at least to 1/N on start of range, and for last range when start == end we should get to that % faster

func (*HistogramData) CalcPercentiles

func (e *HistogramData) CalcPercentiles(percentiles []float64) *HistogramData

CalcPercentiles calculates the requested percentile and add them to the HistogramData. Potential TODO: sort or assume sorting and calculate all the percentiles in 1 pass (greater and greater values).

func (*HistogramData) Print

func (e *HistogramData) Print(out io.Writer, msg string)

Print dumps the histogram (and counter) to the provided writer. Also calculates the percentile.

type Interval

type Interval struct {
	Start float64
	End   float64
}

Interval is a range from start to end. Interval are left closed, open right expect the last one which includes Max. ie [Start, End[ with the next one being [PrevEnd, NextEnd[.

type Percentile

type Percentile struct {
	Percentile float64 // For this Percentile
	Value      float64 // value at that Percentile
}

Percentile value for the percentile