Move registry hashing to xxhash

This is a much stronger hash function than fnv64a and comparably fast
(with super-fast assembly implementation for amd64).

Performance is not critical here anyway.

The old fnv64a is kept for vectors, where collision detection is in
place and the weakness of the hashing doesn't matter that much. I
implemented a vector version with xxhash and found that xxhash is
slower in all cases except very very high cardinality (where it is
only slightly faster). Also, ``xxhash.New`` comes with an allocation
of 80 bytes. Thus, to keep vectors alloc-free, we needed to add a
`sync.Pool`, which would have an additional performance overhead.

Signed-off-by: beorn7 <[email protected]>

Author: beorn7

go.mod

@@ -2,6 +2,7 @@ module github.com/prometheus/client_golang
 
 require (
 	github.com/beorn7/perks v1.0.1
+	github.com/cespare/xxhash/v2 v2.1.0
 	github.com/golang/protobuf v1.3.2
 	github.com/json-iterator/go v1.1.7
 	github.com/prometheus/client_model v0.0.0-20190812154241-14fe0d1b01d4

go.sum

@@ -8,6 +8,8 @@ github.com/beorn7/perks v1.0.0 h1:HWo1m869IqiPhD389kmkxeTalrjNbbJTC8LXupb+sl0=
 github.com/beorn7/perks v1.0.0/go.mod h1:KWe93zE9D1o94FZ5RNwFwVgaQK1VOXiVxmqh+CedLV8=
 github.com/beorn7/perks v1.0.1 h1:VlbKKnNfV8bJzeqoa4cOKqO6bYr3WgKZxO8Z16+hsOM=
 github.com/beorn7/perks v1.0.1/go.mod h1:G2ZrVWU2WbWT9wwq4/hrbKbnv/1ERSJQ0ibhJ6rlkpw=
+github.com/cespare/xxhash/v2 v2.1.0 h1:yTUvW7Vhb89inJ+8irsUqiWjh8iT6sQPZiQzI6ReGkA=
+github.com/cespare/xxhash/v2 v2.1.0/go.mod h1:dgIUBU3pDso/gPgZ1osOZ0iQf77oPR28Tjxl5dIMyVM=
 github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
 github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=

prometheus/desc.go

@@ -19,6 +19,7 @@ import (
 	"sort"
 	"strings"
 
+	"github.com/cespare/xxhash/v2"
 	"github.com/golang/protobuf/proto"
 	"github.com/prometheus/common/model"
 
@@ -126,24 +127,24 @@ func NewDesc(fqName, help string, variableLabels []string, constLabels Labels) *
 		return d
 	}
 
-	vh := hashNew()
+	xxh := xxhash.New()
 	for _, val := range labelValues {
-		vh = hashAdd(vh, val)
-		vh = hashAddByte(vh, separatorByte)
+		xxh.WriteString(val)
+		xxh.Write(separatorByteSlice)
 	}
-	d.id = vh
+	d.id = xxh.Sum64()
 	// Sort labelNames so that order doesn't matter for the hash.
 	sort.Strings(labelNames)
 	// Now hash together (in this order) the help string and the sorted
 	// label names.
-	lh := hashNew()
-	lh = hashAdd(lh, help)
-	lh = hashAddByte(lh, separatorByte)
+	xxh.Reset()
+	xxh.WriteString(help)
+	xxh.Write(separatorByteSlice)
 	for _, labelName := range labelNames {
-		lh = hashAdd(lh, labelName)
-		lh = hashAddByte(lh, separatorByte)
+		xxh.WriteString(labelName)
+		xxh.Write(separatorByteSlice)
 	}
-	d.dimHash = lh
+	d.dimHash = xxh.Sum64()
 
 	d.constLabelPairs = make([]*dto.LabelPair, 0, len(constLabels))
 	for n, v := range constLabels {

prometheus/metric.go

@@ -24,6 +24,8 @@ import (
 
 const separatorByte byte = 255
 
+var separatorByteSlice = []byte{255} // For convenient use with xxhash.
+
 // A Metric models a single sample value with its meta data being exported to
 // Prometheus. Implementations of Metric in this package are Gauge, Counter,
 // Histogram, Summary, and Untyped.

prometheus/registry.go

@@ -25,6 +25,7 @@ import (
 	"sync"
 	"unicode/utf8"
 
+	"github.com/cespare/xxhash/v2"
 	"github.com/golang/protobuf/proto"
 	"github.com/prometheus/common/expfmt"
 
@@ -875,9 +876,9 @@ func checkMetricConsistency(
 	}
 
 	// Is the metric unique (i.e. no other metric with the same name and the same labels)?
-	h := hashNew()
-	h = hashAdd(h, name)
-	h = hashAddByte(h, separatorByte)
+	h := xxhash.New()
+	h.WriteString(name)
+	h.Write(separatorByteSlice)
 	// Make sure label pairs are sorted. We depend on it for the consistency
 	// check.
 	if !sort.IsSorted(labelPairSorter(dtoMetric.Label)) {
@@ -888,18 +889,19 @@ func checkMetricConsistency(
 		dtoMetric.Label = copiedLabels
 	}
 	for _, lp := range dtoMetric.Label {
-		h = hashAdd(h, lp.GetName())
-		h = hashAddByte(h, separatorByte)
-		h = hashAdd(h, lp.GetValue())
-		h = hashAddByte(h, separatorByte)
+		h.WriteString(lp.GetName())
+		h.Write(separatorByteSlice)
+		h.WriteString(lp.GetValue())
+		h.Write(separatorByteSlice)
 	}
-	if _, exists := metricHashes[h]; exists {
+	hSum := h.Sum64()
+	if _, exists := metricHashes[hSum]; exists {
 		return fmt.Errorf(
 			"collected metric %q { %s} was collected before with the same name and label values",
 			name, dtoMetric,
 		)
 	}
-	metricHashes[h] = struct{}{}
+	metricHashes[hSum] = struct{}{}
 	return nil
 }