RepoMirrors
/
prometheus
mirror of https://github.com/prometheus/prometheus
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Factor out quantile fucntion.

This commit is contained in:
Brian Brazil 2016-07-08 13:33:20 +01:00
parent b0342ba9ec
commit 15f9fe0a45
2 changed files with 40 additions and 37 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
promql/functions.go
View File

@ -490,33 +490,13 @@ func funcQuantileOverTime(ev *evaluator, args Expressions) model.Value {
}
el.Metric.Del(model.MetricNameLabel)
var result float64
if q < 0 {
result = math.Inf(-1)
} else if q > 1 {
result = math.Inf(+1)
} else {
values := make([]float64, 0, len(el.Values))
for _, v := range el.Values {
values = append(values, float64(v.Value))
}
sort.Float64s(values)
n := float64(len(el.Values))
// When the quantile lies between two samples,
// we use a weighted average of the two samples.
rank := q * (n-1)
lowerIndex := math.Max(0, math.Floor(rank))
upperIndex := math.Min(n-1, lowerIndex+1)
weight := rank - math.Floor(rank)
result = values[int(lowerIndex)] * (1 - weight) + values[int(upperIndex)] * weight
}
values := make([]float64, 0, len(el.Values))
for _, v := range el.Values {
values = append(values, float64(v.Value))
}
resultVector = append(resultVector, &sample{
Metric: el.Metric,
Value: model.SampleValue(result),
Value: model.SampleValue(quantile(q, values)),
Timestamp: ev.Timestamp,
})
}
@ -750,7 +730,7 @@ func funcHistogramQuantile(ev *evaluator, args Expressions) model.Value {
for _, mb := range signatureToMetricWithBuckets {
outVec = append(outVec, &sample{
Metric: mb.metric,
Value: model.SampleValue(quantile(q, mb.buckets)),
Value: model.SampleValue(bucketQuantile(q, mb.buckets)),
Timestamp: ev.Timestamp,
})
}

45
promql/quantile.go
View File

@ -48,16 +48,16 @@ type metricWithBuckets struct {
buckets buckets
}
// quantile calculates the quantile 'q' based on the given buckets. The buckets
// will be sorted by upperBound by this function (i.e. no sorting needed before
// calling this function). The quantile value is interpolated assuming a linear
// distribution within a bucket. However, if the quantile falls into the highest
// bucket, the upper bound of the 2nd highest bucket is returned. A natural
// lower bound of 0 is assumed if the upper bound of the lowest bucket is
// greater 0. In that case, interpolation in the lowest bucket happens linearly
// between 0 and the upper bound of the lowest bucket. However, if the lowest
// bucket has an upper bound less or equal 0, this upper bound is returned if
// the quantile falls into the lowest bucket.
// bucketQuantile calculates the quantile 'q' based on the given buckets. The
// buckets will be sorted by upperBound by this function (i.e. no sorting
// needed before calling this function). The quantile value is interpolated
// assuming a linear distribution within a bucket. However, if the quantile
// falls into the highest bucket, the upper bound of the 2nd highest bucket is
// returned. A natural lower bound of 0 is assumed if the upper bound of the
// lowest bucket is greater 0. In that case, interpolation in the lowest bucket
// happens linearly between 0 and the upper bound of the lowest bucket.
// However, if the lowest bucket has an upper bound less or equal 0, this upper
// bound is returned if the quantile falls into the lowest bucket.
//
// There are a number of special cases (once we have a way to report errors
// happening during evaluations of AST functions, we should report those
@ -70,7 +70,7 @@ type metricWithBuckets struct {
// If q<0, -Inf is returned.
//
// If q>1, +Inf is returned.
func quantile(q model.SampleValue, buckets buckets) float64 {
func bucketQuantile(q model.SampleValue, buckets buckets) float64 {
if q < 0 {
return math.Inf(-1)
}
@ -106,3 +106,26 @@ func quantile(q model.SampleValue, buckets buckets) float64 {
}
return bucketStart + (bucketEnd-bucketStart)*float64(rank/count)
}
// qauntile calculates the given quantile of a slice of floats.
// The slice will be sorted.
func quantile(q float64, values []float64) float64 {
if q < 0 {
return math.Inf(-1)
}
if q > 1 {
return math.Inf(+1)
}
sort.Float64s(values)
n := float64(len(values))
// When the quantile lies between two samples,
// we use a weighted average of the two samples.
rank := q * (n - 1)
lowerIndex := math.Max(0, math.Floor(rank))
upperIndex := math.Min(n-1, lowerIndex+1)
weight := rank - math.Floor(rank)
return values[int(lowerIndex)]*(1-weight) + values[int(upperIndex)]*weight
}