Method Results

pkg/traceql/engine_metrics.go:1930–2008

()

}

func (h *HistogramAggregator) Results() SeriesSet {
	results := make(SeriesSet, len(h.ss)*len(h.qs))

	// Aggregate buckets across all series and time intervals for better quantile calculation
	aggregatedBuckets := make(map[float64]int) // bucketMax -> totalCount

	for _, in := range h.ss {
		// Aggregate bucket counts across all time intervals
		for _, hist := range in.hist {
			for _, bucket := range hist.Buckets {
				aggregatedBuckets[bucket.Max] += bucket.Count
			}
		}
	}

	// Calculate quantile values from aggregated distribution
	// Convert map to sorted slice
	buckets := make([]HistogramBucket, 0, len(aggregatedBuckets))
	for bucketMax, count := range aggregatedBuckets {
		buckets = append(buckets, HistogramBucket{
			Max:   bucketMax,
			Count: count,
		})
	}

	sort.Slice(buckets, func(i, j int) bool {
		return buckets[i].Max < buckets[j].Max
	})

	quantileValues := make([]float64, len(h.qs))
	for i, q := range h.qs {
		quantileValues[i] = Log2Quantile(q, buckets)
	}

	// Build results using the calculated quantile values
	for _, in := range h.ss {
		for i := range in.hist {
			if len(in.hist[i].Buckets) > 0 {
				// Sort interval buckets in place for performance
				sort.Slice(in.hist[i].Buckets, func(a, b int) bool {
					return in.hist[i].Buckets[a].Max < in.hist[i].Buckets[b].Max
				})
			}
		}

		// For each input series, we create a new series for each quantile.
		for qIdx, q := range h.qs {
			// Append label for the quantile
			labels := append((Labels)(nil), in.labels...)
			labels = append(labels, Label{"p", NewStaticFloat(q)})

			ts := TimeSeries{
				Labels: labels,
				Values: make([]float64, len(in.hist)),
			}

			for i := range in.hist {
				if len(in.hist[i].Buckets) == 0 {