Fix for resampler

This commit is contained in:
Aditya Ujeniya 2024-08-25 16:12:43 +02:00
parent b186dca79d
commit 8cefabac7f
2 changed files with 27 additions and 6 deletions

View File

@ -2,6 +2,7 @@ package resampler
import ( import (
"errors" "errors"
"fmt"
"math" "math"
"github.com/ClusterCockpit/cc-metric-store/internal/util" "github.com/ClusterCockpit/cc-metric-store/internal/util"
@ -34,14 +35,14 @@ func SimpleResampler(data []util.Float, old_frequency int64, new_frequency int64
// Inspired by one of the algorithms from https://skemman.is/bitstream/1946/15343/3/SS_MSthesis.pdf // Inspired by one of the algorithms from https://skemman.is/bitstream/1946/15343/3/SS_MSthesis.pdf
// Adapted from https://github.com/haoel/downsampling/blob/master/core/lttb.go // Adapted from https://github.com/haoel/downsampling/blob/master/core/lttb.go
func LargestTriangleThreeBucket(data []util.Float, old_frequency int64, new_frequency int64) ([]util.Float, int64, error) { func LargestTriangleThreeBucket(data []util.Float, old_frequency int, new_frequency int) ([]util.Float, int, error) {
if old_frequency == 0 || new_frequency == 0 { if old_frequency == 0 || new_frequency == 0 {
return data, old_frequency, nil return data, old_frequency, nil
} }
if new_frequency%old_frequency != 0 { if new_frequency%old_frequency != 0 {
return nil, 0, errors.New("new sampling frequency should be multiple of the old frequency") return nil, 0, errors.New(fmt.Sprintf("new sampling frequency : %d should be multiple of the old frequency : %d", new_frequency, old_frequency))
} }
var step int = int(new_frequency / old_frequency) var step int = int(new_frequency / old_frequency)
@ -89,6 +90,7 @@ func LargestTriangleThreeBucket(data []util.Float, old_frequency int64, new_freq
maxArea := -1.0 maxArea := -1.0
var maxAreaPoint int var maxAreaPoint int
flag_ := 0
for ; currBucketStart < currBucketEnd; currBucketStart++ { for ; currBucketStart < currBucketEnd; currBucketStart++ {
area := calculateTriangleArea(util.Float(pointX), pointY, avgPointX, avgPointY, util.Float(currBucketStart), data[currBucketStart]) area := calculateTriangleArea(util.Float(pointX), pointY, avgPointX, avgPointY, util.Float(currBucketStart), data[currBucketStart])
@ -96,10 +98,19 @@ func LargestTriangleThreeBucket(data []util.Float, old_frequency int64, new_freq
maxArea = area maxArea = area
maxAreaPoint = currBucketStart maxAreaPoint = currBucketStart
} }
if math.IsNaN(float64(avgPointY)) {
flag_ = 1
}
} }
new_data = append(new_data, data[maxAreaPoint]) // Pick this point from the bucket if flag_ == 1 {
prevMaxAreaPoint = maxAreaPoint // This MaxArea point is the next's prevMAxAreaPoint new_data = append(new_data, util.NaN) // Pick this point from the bucket
} else {
new_data = append(new_data, data[maxAreaPoint]) // Pick this point from the bucket
}
prevMaxAreaPoint = maxAreaPoint // This MaxArea point is the next's prevMAxAreaPoint
//move to the next window //move to the next window
bucketLow = bucketMiddle bucketLow = bucketMiddle

View File

@ -12,14 +12,24 @@ func calculateTriangleArea(paX, paY, pbX, pbY, pcX, pcY util.Float) float64 {
} }
func calculateAverageDataPoint(points []util.Float, xStart int64) (avgX util.Float, avgY util.Float) { func calculateAverageDataPoint(points []util.Float, xStart int64) (avgX util.Float, avgY util.Float) {
flag := 0
for _, point := range points { for _, point := range points {
avgX += util.Float(xStart) avgX += util.Float(xStart)
avgY += point avgY += point
xStart++ xStart++
if math.IsNaN(float64(point)) {
flag = 1
}
} }
l := util.Float(len(points)) l := util.Float(len(points))
avgX /= l avgX /= l
avgY /= l avgY /= l
return avgX, avgY
if flag == 1 {
return avgX, util.NaN
} else {
return avgX, avgY
}
} }