Add utf8 support to the prometheus exporter (#5755)

### Changes

Disable sanitization when the UTF-8 support is enabled in the Prometheus
library.

### Usage

To enable UTF-8 support for the Prometheus exporter after this change,
set the following in your application:

```golang
import "github.com/prometheus/common/model"

func init() {
    model.NameValidationScheme = model.UTF8Validation
}
```

See `exporters/prometheus/testdata/counter_utf8.txt` for an example of
the text exposition format including names/labels with dots.

CHANGELOG.md

@@ -13,6 +13,7 @@ This project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.htm
 - Support `OTEL_EXPORTER_OTLP_LOGS_INSECURE` and `OTEL_EXPORTER_OTLP_INSECURE` environments in `go.opentelemetry.io/otel/exporters/otlp/otlplog/otlploggrpc`. (#5739)
 - The `WithResource` option for `NewMeterProvider` now merges the provided resources with the ones from environment variables. (#5773)
 - The `WithResource` option for `NewLoggerProvider` now merges the provided resources with the ones from environment variables. (#5773)
+- Add UTF-8 support to `go.opentelemetry.io/otel/exporters/prometheus`. (#5755)
 
 ### Fixed
 

exporters/prometheus/config.go

@@ -7,6 +7,7 @@ import (
 	"strings"
 
 	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/common/model"
 
 	"go.opentelemetry.io/otel/attribute"
 	"go.opentelemetry.io/otel/sdk/metric"
@@ -131,7 +132,10 @@ func WithoutScopeInfo() Option {
 // have special behavior based on their name.
 func WithNamespace(ns string) Option {
 	return optionFunc(func(cfg config) config {
-		ns = sanitizeName(ns)
+		if model.NameValidationScheme != model.UTF8Validation {
+			// Only sanitize if prometheus does not support UTF-8.
+			ns = model.EscapeName(ns, model.NameEscapingScheme)
+		}
 		if !strings.HasSuffix(ns, "_") {
 			// namespace and metric names should be separated with an underscore,
 			// adds a trailing underscore if there is not one already.

exporters/prometheus/exporter.go

@@ -11,11 +11,10 @@ import (
 	"slices"
 	"strings"
 	"sync"
-	"unicode"
-	"unicode/utf8"
 
 	"github.com/prometheus/client_golang/prometheus"
 	dto "github.com/prometheus/client_model/go"
+	"github.com/prometheus/common/model"
 	"google.golang.org/protobuf/proto"
 
 	"go.opentelemetry.io/otel"
@@ -298,28 +297,38 @@ func addGaugeMetric[N int64 | float64](ch chan<- prometheus.Metric, gauge metric
 }
 
 // getAttrs parses the attribute.Set to two lists of matching Prometheus-style
-// keys and values. It sanitizes invalid characters and handles duplicate keys
-// (due to sanitization) by sorting and concatenating the values following the spec.
+// keys and values.
 func getAttrs(attrs attribute.Set, ks, vs [2]string, resourceKV keyVals) ([]string, []string) {
-	keysMap := make(map[string][]string)
-	itr := attrs.Iter()
-	for itr.Next() {
-		kv := itr.Attribute()
-		key := strings.Map(sanitizeRune, string(kv.Key))
-		if _, ok := keysMap[key]; !ok {
-			keysMap[key] = []string{kv.Value.Emit()}
-		} else {
-			// if the sanitized key is a duplicate, append to the list of keys
-			keysMap[key] = append(keysMap[key], kv.Value.Emit())
-		}
-	}
-
 	keys := make([]string, 0, attrs.Len())
 	values := make([]string, 0, attrs.Len())
-	for key, vals := range keysMap {
-		keys = append(keys, key)
-		slices.Sort(vals)
-		values = append(values, strings.Join(vals, ";"))
+	itr := attrs.Iter()
+
+	if model.NameValidationScheme == model.UTF8Validation {
+		// Do not perform sanitization if prometheus supports UTF-8.
+		for itr.Next() {
+			kv := itr.Attribute()
+			keys = append(keys, string(kv.Key))
+			values = append(values, kv.Value.Emit())
+		}
+	} else {
+		// It sanitizes invalid characters and handles duplicate keys
+		// (due to sanitization) by sorting and concatenating the values following the spec.
+		keysMap := make(map[string][]string)
+		for itr.Next() {
+			kv := itr.Attribute()
+			key := model.EscapeName(string(kv.Key), model.NameEscapingScheme)
+			if _, ok := keysMap[key]; !ok {
+				keysMap[key] = []string{kv.Value.Emit()}
+			} else {
+				// if the sanitized key is a duplicate, append to the list of keys
+				keysMap[key] = append(keysMap[key], kv.Value.Emit())
+			}
+		}
+		for key, vals := range keysMap {
+			keys = append(keys, key)
+			slices.Sort(vals)
+			values = append(values, strings.Join(vals, ";"))
+		}
 	}
 
 	if ks[0] != "" {
@@ -347,13 +356,6 @@ func createScopeInfoMetric(scope instrumentation.Scope) (prometheus.Metric, erro
 	return prometheus.NewConstMetric(desc, prometheus.GaugeValue, float64(1), scope.Name, scope.Version)
 }
 
-func sanitizeRune(r rune) rune {
-	if unicode.IsLetter(r) || unicode.IsDigit(r) || r == ':' || r == '_' {
-		return r
-	}
-	return '_'
-}
-
 var unitSuffixes = map[string]string{
 	// Time
 	"d":   "_days",
@@ -392,7 +394,11 @@ var unitSuffixes = map[string]string{
 
 // getName returns the sanitized name, prefixed with the namespace and suffixed with unit.
 func (c *collector) getName(m metricdata.Metrics, typ *dto.MetricType) string {
-	name := sanitizeName(m.Name)
+	name := m.Name
+	if model.NameValidationScheme != model.UTF8Validation {
+		// Only sanitize if prometheus does not support UTF-8.
+		name = model.EscapeName(name, model.NameEscapingScheme)
+	}
 	addCounterSuffix := !c.withoutCounterSuffixes && *typ == dto.MetricType_COUNTER
 	if addCounterSuffix {
 		// Remove the _total suffix here, as we will re-add the total suffix
@@ -411,59 +417,6 @@ func (c *collector) getName(m metricdata.Metrics, typ *dto.MetricType) string {
 	return name
 }
 
-func sanitizeName(n string) string {
-	// This algorithm is based on strings.Map from Go 1.19.
-	const replacement = '_'
-
-	valid := func(i int, r rune) bool {
-		// Taken from
-		// https://github.com/prometheus/common/blob/dfbc25bd00225c70aca0d94c3c4bb7744f28ace0/model/metric.go#L92-L102
-		if (r >= 'a' && r <= 'z') || (r >= 'A' && r <= 'Z') || r == '_' || r == ':' || (r >= '0' && r <= '9' && i > 0) {
-			return true
-		}
-		return false
-	}
-
-	// This output buffer b is initialized on demand, the first time a
-	// character needs to be replaced.
-	var b strings.Builder
-	for i, c := range n {
-		if valid(i, c) {
-			continue
-		}
-
-		if i == 0 && c >= '0' && c <= '9' {
-			// Prefix leading number with replacement character.
-			b.Grow(len(n) + 1)
-			_ = b.WriteByte(byte(replacement))
-			break
-		}
-		b.Grow(len(n))
-		_, _ = b.WriteString(n[:i])
-		_ = b.WriteByte(byte(replacement))
-		width := utf8.RuneLen(c)
-		n = n[i+width:]
-		break
-	}
-
-	// Fast path for unchanged input.
-	if b.Cap() == 0 { // b.Grow was not called above.
-		return n
-	}
-
-	for _, c := range n {
-		// Due to inlining, it is more performant to invoke WriteByte rather then
-		// WriteRune.
-		if valid(1, c) { // We are guaranteed to not be at the start.
-			_ = b.WriteByte(byte(c))
-		} else {
-			_ = b.WriteByte(byte(replacement))
-		}
-	}
-
-	return b.String()
-}
-
 func (c *collector) metricType(m metricdata.Metrics) *dto.MetricType {
 	switch v := m.Data.(type) {
 	case metricdata.Histogram[int64], metricdata.Histogram[float64]:

exporters/prometheus/exporter_test.go

@@ -14,6 +14,7 @@ import (
 	"github.com/prometheus/client_golang/prometheus"
 	"github.com/prometheus/client_golang/prometheus/testutil"
 	dto "github.com/prometheus/client_model/go"
+	"github.com/prometheus/common/model"
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
 
@@ -34,6 +35,7 @@ func TestPrometheusExporter(t *testing.T) {
 		recordMetrics       func(ctx context.Context, meter otelmetric.Meter)
 		options             []Option
 		expectedFile        string
+		enableUTF8          bool
 	}{
 		{
 			name:         "counter",
@@ -399,10 +401,47 @@ func TestPrometheusExporter(t *testing.T) {
 				counter.Add(ctx, 5.3, opt)
 			},
 		},
+		{
+			name:         "counter utf-8",
+			expectedFile: "testdata/counter_utf8.txt",
+			enableUTF8:   true,
+			recordMetrics: func(ctx context.Context, meter otelmetric.Meter) {
+				opt := otelmetric.WithAttributes(
+					attribute.Key("A.G").String("B"),
+					attribute.Key("C.H").String("D"),
+					attribute.Key("E.I").Bool(true),
+					attribute.Key("F.J").Int(42),
+				)
+				counter, err := meter.Float64Counter(
+					"foo.things",
+					otelmetric.WithDescription("a simple counter"),
+					otelmetric.WithUnit("s"),
+				)
+				require.NoError(t, err)
+				counter.Add(ctx, 5, opt)
+				counter.Add(ctx, 10.3, opt)
+				counter.Add(ctx, 9, opt)
+
+				attrs2 := attribute.NewSet(
+					attribute.Key("A.G").String("D"),
+					attribute.Key("C.H").String("B"),
+					attribute.Key("E.I").Bool(true),
+					attribute.Key("F.J").Int(42),
+				)
+				counter.Add(ctx, 5, otelmetric.WithAttributeSet(attrs2))
+			},
+		},
 	}
 
 	for _, tc := range testCases {
 		t.Run(tc.name, func(t *testing.T) {
+			if tc.enableUTF8 {
+				model.NameValidationScheme = model.UTF8Validation
+				defer func() {
+					// Reset to defaults
+					model.NameValidationScheme = model.LegacyValidation
+				}()
+			}
 			ctx := context.Background()
 			registry := prometheus.NewRegistry()
 			exporter, err := New(append(tc.options, WithRegisterer(registry))...)
@@ -452,36 +491,6 @@ func TestPrometheusExporter(t *testing.T) {
 	}
 }
 
-func TestSantitizeName(t *testing.T) {
-	tests := []struct {
-		input string
-		want  string
-	}{
-		{"name€_with_4_width_rune", "name__with_4_width_rune"},
-		{"`", "_"},
-		{
-			`! "#$%&'()*+,-./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWKYZ[]\^_abcdefghijklmnopqrstuvwkyz{|}~`,
-			`________________0123456789:______ABCDEFGHIJKLMNOPQRSTUVWKYZ_____abcdefghijklmnopqrstuvwkyz____`,
-		},
-
-		// Test cases taken from
-		// https://github.com/prometheus/common/blob/dfbc25bd00225c70aca0d94c3c4bb7744f28ace0/model/metric_test.go#L85-L136
-		{"Avalid_23name", "Avalid_23name"},
-		{"_Avalid_23name", "_Avalid_23name"},
-		{"1valid_23name", "_1valid_23name"},
-		{"avalid_23name", "avalid_23name"},
-		{"Ava:lid_23name", "Ava:lid_23name"},
-		{"a lid_23name", "a_lid_23name"},
-		{":leading_colon", ":leading_colon"},
-		{"colon:in:the:middle", "colon:in:the:middle"},
-		{"", ""},
-	}
-
-	for _, test := range tests {
-		require.Equalf(t, test.want, sanitizeName(test.input), "input: %q", test.input)
-	}
-}
-
 func TestMultiScopes(t *testing.T) {
 	ctx := context.Background()
 	registry := prometheus.NewRegistry()

exporters/prometheus/go.mod

@@ -5,6 +5,7 @@ go 1.22
 require (
 	github.com/prometheus/client_golang v1.20.3
 	github.com/prometheus/client_model v0.6.1
+	github.com/prometheus/common v0.59.1
 	github.com/stretchr/testify v1.9.0
 	go.opentelemetry.io/otel v1.29.0
 	go.opentelemetry.io/otel/metric v1.29.0
@@ -25,7 +26,6 @@ require (
 	github.com/kylelemons/godebug v1.1.0 // indirect
 	github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 // indirect
 	github.com/pmezard/go-difflib v1.0.0 // indirect
-	github.com/prometheus/common v0.59.1 // indirect
 	github.com/prometheus/procfs v0.15.1 // indirect
 	golang.org/x/sys v0.25.0 // indirect
 	gopkg.in/yaml.v3 v3.0.1 // indirect

exporters/prometheus/testdata/counter_utf8.txt

@@ -0,0 +1,10 @@
+# HELP "foo.things_seconds_total" a simple counter
+# TYPE "foo.things_seconds_total" counter
+{"foo.things_seconds_total","A.G"="B","C.H"="D","E.I"="true","F.J"="42",otel_scope_name="testmeter",otel_scope_version="v0.1.0"} 24.3
+{"foo.things_seconds_total","A.G"="D","C.H"="B","E.I"="true","F.J"="42",otel_scope_name="testmeter",otel_scope_version="v0.1.0"} 5
+# HELP otel_scope_info Instrumentation Scope metadata
+# TYPE otel_scope_info gauge
+otel_scope_info{otel_scope_name="testmeter",otel_scope_version="v0.1.0"} 1
+# HELP target_info Target metadata
+# TYPE target_info gauge
+target_info{"service.name"="prometheus_test","telemetry.sdk.language"="go","telemetry.sdk.name"="opentelemetry","telemetry.sdk.version"="latest"} 1

exporters/prometheus/testdata/sanitized_names.txt

@@ -1,9 +1,9 @@
 # HELP bar a fun little gauge
 # TYPE bar gauge
 bar{A="B",C="D",otel_scope_name="testmeter",otel_scope_version="v0.1.0"} 75
-# HELP _0invalid_counter_name_total a counter with an invalid name
-# TYPE _0invalid_counter_name_total counter
-_0invalid_counter_name_total{A="B",C="D",otel_scope_name="testmeter",otel_scope_version="v0.1.0"} 100
+# HELP _invalid_counter_name_total a counter with an invalid name
+# TYPE _invalid_counter_name_total counter
+_invalid_counter_name_total{A="B",C="D",otel_scope_name="testmeter",otel_scope_version="v0.1.0"} 100
 # HELP invalid_gauge_name a gauge with an invalid name
 # TYPE invalid_gauge_name gauge
 invalid_gauge_name{A="B",C="D",otel_scope_name="testmeter",otel_scope_version="v0.1.0"} 100

internal/tools/go.mod

@@ -51,7 +51,7 @@ require (
 	github.com/butuzov/mirror v1.2.0 // indirect
 	github.com/catenacyber/perfsprint v0.7.1 // indirect
 	github.com/ccojocar/zxcvbn-go v1.0.2 // indirect
-	github.com/cespare/xxhash/v2 v2.2.0 // indirect
+	github.com/cespare/xxhash/v2 v2.3.0 // indirect
 	github.com/charithe/durationcheck v0.0.10 // indirect
 	github.com/chavacava/garif v0.1.0 // indirect
 	github.com/ckaznocha/intrange v0.1.2 // indirect
@@ -133,6 +133,7 @@ require (
 	github.com/mitchellh/go-homedir v1.1.0 // indirect
 	github.com/mitchellh/mapstructure v1.5.0 // indirect
 	github.com/moricho/tparallel v0.3.2 // indirect
+	github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 // indirect
 	github.com/nakabonne/nestif v0.3.1 // indirect
 	github.com/nishanths/exhaustive v0.12.0 // indirect
 	github.com/nishanths/predeclared v0.2.2 // indirect
@@ -142,10 +143,10 @@ require (
 	github.com/pjbgf/sha1cd v0.3.0 // indirect
 	github.com/pmezard/go-difflib v1.0.1-0.20181226105442-5d4384ee4fb2 // indirect
 	github.com/polyfloyd/go-errorlint v1.6.0 // indirect
-	github.com/prometheus/client_golang v1.19.0 // indirect
+	github.com/prometheus/client_golang v1.20.2 // indirect
 	github.com/prometheus/client_model v0.6.1 // indirect
-	github.com/prometheus/common v0.48.0 // indirect
-	github.com/prometheus/procfs v0.12.0 // indirect
+	github.com/prometheus/common v0.57.0 // indirect
+	github.com/prometheus/procfs v0.15.1 // indirect
 	github.com/quasilyte/go-ruleguard v0.4.2 // indirect
 	github.com/quasilyte/go-ruleguard/dsl v0.3.22 // indirect
 	github.com/quasilyte/gogrep v0.5.0 // indirect