Exemplar: revert wrong implementation. (#1067)

* Exemplar: move to metricdata.

* Remove AttachmentExtractor.

* More cleanup.

* Add a TODO for next PR.

trace/exemplar.go → metric/metricdata/exemplar.go

@@ -12,32 +12,22 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 
-package trace
+package metricdata
 
 import (
-	"context"
-	"encoding/hex"
-
-	"go.opencensus.io/exemplar"
+	"time"
 )
 
-func init() {
-	exemplar.RegisterAttachmentExtractor(attachSpanContext)
+// Exemplar is an example data point associated with each bucket of a
+// distribution type aggregation.
+//
+// Their purpose is to provide an example of the kind of thing
+// (request, RPC, trace span, etc.) that resulted in that measurement.
+type Exemplar struct {
+	Value       float64     // the value that was recorded
+	Timestamp   time.Time   // the time the value was recorded
+	Attachments Attachments // attachments (if any)
 }
 
-func attachSpanContext(ctx context.Context, a exemplar.Attachments) exemplar.Attachments {
-	span := FromContext(ctx)
-	if span == nil {
-		return a
-	}
-	sc := span.SpanContext()
-	if !sc.IsSampled() {
-		return a
-	}
-	if a == nil {
-		a = make(exemplar.Attachments)
-	}
-	a[exemplar.KeyTraceID] = hex.EncodeToString(sc.TraceID[:])
-	a[exemplar.KeySpanID] = hex.EncodeToString(sc.SpanID[:])
-	return a
-}
+// Attachments is a map of extra values associated with a recorded data point.
+type Attachments map[string]string

metric/metricdata/point.go

@@ -16,8 +16,6 @@ package metricdata
 
 import (
 	"time"
-
-	"go.opencensus.io/exemplar"
 )
 
 // Point is a single data point of a time series.
@@ -144,7 +142,7 @@ type Bucket struct {
 	// bucket_bounds.
 	Count int64
 	// Exemplar associated with this bucket (if any).
-	Exemplar *exemplar.Exemplar
+	Exemplar *Exemplar
 }
 
 // Summary is a representation of percentiles.

stats/record.go

@@ -18,7 +18,6 @@ package stats
 import (
 	"context"
 
-	"go.opencensus.io/exemplar"
 	"go.opencensus.io/stats/internal"
 	"go.opencensus.io/tag"
 )
@@ -51,7 +50,8 @@ func Record(ctx context.Context, ms ...Measurement) {
 	if !record {
 		return
 	}
-	recorder(tag.FromContext(ctx), ms, exemplar.AttachmentsFromContext(ctx))
+	// TODO(songy23): fix attachments.
+	recorder(tag.FromContext(ctx), ms, map[string]string{})
 }
 
 // RecordWithTags records one or multiple measurements at once.

stats/view/aggregation_data.go

@@ -18,15 +18,15 @@ package view
 import (
 	"math"
 
-	"go.opencensus.io/exemplar"
+	"go.opencensus.io/metric/metricdata"
 )
 
 // AggregationData represents an aggregated value from a collection.
 // They are reported on the view data during exporting.
 // Mosts users won't directly access aggregration data.
 type AggregationData interface {
 	isAggregationData() bool
-	addSample(e *exemplar.Exemplar)
+	addSample(v float64)
 	clone() AggregationData
 	equal(other AggregationData) bool
 }
@@ -43,7 +43,7 @@ type CountData struct {
 
 func (a *CountData) isAggregationData() bool { return true }
 
-func (a *CountData) addSample(_ *exemplar.Exemplar) {
+func (a *CountData) addSample(_ float64) {
 	a.Value = a.Value + 1
 }
 
@@ -70,8 +70,8 @@ type SumData struct {
 
 func (a *SumData) isAggregationData() bool { return true }
 
-func (a *SumData) addSample(e *exemplar.Exemplar) {
-	a.Value += e.Value
+func (a *SumData) addSample(v float64) {
+	a.Value += v
 }
 
 func (a *SumData) clone() AggregationData {
@@ -101,16 +101,16 @@ type DistributionData struct {
 	SumOfSquaredDev float64 // sum of the squared deviation from the mean
 	CountPerBucket  []int64 // number of occurrences per bucket
 	// ExemplarsPerBucket is slice the same length as CountPerBucket containing
-	// an exemplar for the associated bucket, or nil.
-	ExemplarsPerBucket []*exemplar.Exemplar
+	// an metricdata for the associated bucket, or nil.
+	ExemplarsPerBucket []*metricdata.Exemplar
 	bounds             []float64 // histogram distribution of the values
 }
 
 func newDistributionData(bounds []float64) *DistributionData {
 	bucketCount := len(bounds) + 1
 	return &DistributionData{
 		CountPerBucket:     make([]int64, bucketCount),
-		ExemplarsPerBucket: make([]*exemplar.Exemplar, bucketCount),
+		ExemplarsPerBucket: make([]*metricdata.Exemplar, bucketCount),
 		bounds:             bounds,
 		Min:                math.MaxFloat64,
 		Max:                math.SmallestNonzeroFloat64,
@@ -129,64 +129,45 @@ func (a *DistributionData) variance() float64 {
 
 func (a *DistributionData) isAggregationData() bool { return true }
 
-func (a *DistributionData) addSample(e *exemplar.Exemplar) {
-	f := e.Value
-	if f < a.Min {
-		a.Min = f
+// TODO(songy23): support exemplar attachments.
+func (a *DistributionData) addSample(v float64) {
+	if v < a.Min {
+		a.Min = v
 	}
-	if f > a.Max {
-		a.Max = f
+	if v > a.Max {
+		a.Max = v
 	}
 	a.Count++
-	a.addToBucket(e)
+	a.addToBucket(v)
 
 	if a.Count == 1 {
-		a.Mean = f
+		a.Mean = v
 		return
 	}
 
 	oldMean := a.Mean
-	a.Mean = a.Mean + (f-a.Mean)/float64(a.Count)
-	a.SumOfSquaredDev = a.SumOfSquaredDev + (f-oldMean)*(f-a.Mean)
+	a.Mean = a.Mean + (v-a.Mean)/float64(a.Count)
+	a.SumOfSquaredDev = a.SumOfSquaredDev + (v-oldMean)*(v-a.Mean)
 }
 
-func (a *DistributionData) addToBucket(e *exemplar.Exemplar) {
+func (a *DistributionData) addToBucket(v float64) {
 	var count *int64
-	var ex **exemplar.Exemplar
 	for i, b := range a.bounds {
-		if e.Value < b {
+		if v < b {
 			count = &a.CountPerBucket[i]
-			ex = &a.ExemplarsPerBucket[i]
 			break
 		}
 	}
-	if count == nil {
+	if count == nil { // Last bucket.
 		count = &a.CountPerBucket[len(a.bounds)]
-		ex = &a.ExemplarsPerBucket[len(a.bounds)]
 	}
 	*count++
-	*ex = maybeRetainExemplar(*ex, e)
-}
-
-func maybeRetainExemplar(old, cur *exemplar.Exemplar) *exemplar.Exemplar {
-	if old == nil {
-		return cur
-	}
-
-	// Heuristic to pick the "better" exemplar: first keep the one with a
-	// sampled trace attachment, if neither have a trace attachment, pick the
-	// one with more attachments.
-	_, haveTraceID := cur.Attachments[exemplar.KeyTraceID]
-	if haveTraceID || len(cur.Attachments) >= len(old.Attachments) {
-		return cur
-	}
-	return old
 }
 
 func (a *DistributionData) clone() AggregationData {
 	c := *a
 	c.CountPerBucket = append([]int64(nil), a.CountPerBucket...)
-	c.ExemplarsPerBucket = append([]*exemplar.Exemplar(nil), a.ExemplarsPerBucket...)
+	c.ExemplarsPerBucket = append([]*metricdata.Exemplar(nil), a.ExemplarsPerBucket...)
 	return &c
 }
 
@@ -218,8 +199,8 @@ func (l *LastValueData) isAggregationData() bool {
 	return true
 }
 
-func (l *LastValueData) addSample(e *exemplar.Exemplar) {
-	l.Value = e.Value
+func (l *LastValueData) addSample(v float64) {
+	l.Value = v
 }
 
 func (l *LastValueData) clone() AggregationData {

stats/view/aggregation_data_test.go

@@ -18,11 +18,10 @@ package view
 import (
 	"reflect"
 	"testing"
-	"time"
 
 	"github.com/google/go-cmp/cmp"
 	"github.com/google/go-cmp/cmp/cmpopts"
-	"go.opencensus.io/exemplar"
+	"go.opencensus.io/metric/metricdata"
 )
 
 func TestDataClone(t *testing.T) {
@@ -67,21 +66,12 @@ func TestDataClone(t *testing.T) {
 
 func TestDistributionData_addSample(t *testing.T) {
 	dd := newDistributionData([]float64{1, 2})
-	t1, _ := time.Parse("Mon Jan 2 15:04:05 -0700 MST 2006", "Mon Jan 2 15:04:05 -0700 MST 2006")
-	e1 := &exemplar.Exemplar{
-		Attachments: exemplar.Attachments{
-			"tag:X": "Y",
-			"tag:A": "B",
-		},
-		Timestamp: t1,
-		Value:     0.5,
-	}
-	dd.addSample(e1)
+	dd.addSample(0.5)
 
 	want := &DistributionData{
 		Count:              1,
 		CountPerBucket:     []int64{1, 0, 0},
-		ExemplarsPerBucket: []*exemplar.Exemplar{e1, nil, nil},
+		ExemplarsPerBucket: []*metricdata.Exemplar{nil, nil, nil},
 		Max:                0.5,
 		Min:                0.5,
 		Mean:               0.5,
@@ -91,21 +81,13 @@ func TestDistributionData_addSample(t *testing.T) {
 		t.Fatalf("Unexpected DistributionData -got +want: %s", diff)
 	}
 
-	t2 := t1.Add(time.Microsecond)
-	e2 := &exemplar.Exemplar{
-		Attachments: exemplar.Attachments{
-			"tag:X": "Y",
-		},
-		Timestamp: t2,
-		Value:     0.7,
-	}
-	dd.addSample(e2)
+	dd.addSample(0.7)
 
 	// Previous exemplar should be preserved, since it has more annotations.
 	want = &DistributionData{
 		Count:              2,
 		CountPerBucket:     []int64{2, 0, 0},
-		ExemplarsPerBucket: []*exemplar.Exemplar{e1, nil, nil},
+		ExemplarsPerBucket: []*metricdata.Exemplar{nil, nil, nil},
 		Max:                0.7,
 		Min:                0.5,
 		Mean:               0.6,
@@ -115,21 +97,13 @@ func TestDistributionData_addSample(t *testing.T) {
 		t.Fatalf("Unexpected DistributionData -got +want: %s", diff)
 	}
 
-	t3 := t2.Add(time.Microsecond)
-	e3 := &exemplar.Exemplar{
-		Attachments: exemplar.Attachments{
-			exemplar.KeyTraceID: "abcd",
-		},
-		Timestamp: t3,
-		Value:     0.2,
-	}
-	dd.addSample(e3)
+	dd.addSample(0.2)
 
 	// Exemplar should be replaced since it has a trace_id.
 	want = &DistributionData{
 		Count:              3,
 		CountPerBucket:     []int64{3, 0, 0},
-		ExemplarsPerBucket: []*exemplar.Exemplar{e3, nil, nil},
+		ExemplarsPerBucket: []*metricdata.Exemplar{nil, nil, nil},
 		Max:                0.7,
 		Min:                0.2,
 		Mean:               0.4666666666666667,