feat(prometheus): explicit label/metric name validation scheme

Author: juliusmh

prometheus/desc.go

@@ -95,7 +95,7 @@ func (v2) NewDesc(fqName, help string, variableLabels ConstrainableLabels, const
 		help:           help,
 		variableLabels: variableLabels.compile(),
 	}
-	if !model.IsValidMetricName(model.LabelValue(fqName)) {
+	if !model.IsValidMetricName(model.LabelValue(fqName), model.UTF8Validation) {
 		d.err = fmt.Errorf("%q is not a valid metric name", fqName)
 		return d
 	}
@@ -107,7 +107,7 @@ func (v2) NewDesc(fqName, help string, variableLabels ConstrainableLabels, const
 	labelNameSet := map[string]struct{}{}
 	// First add only the const label names and sort them...
 	for labelName := range constLabels {
-		if !checkLabelName(labelName) {
+		if !checkLabelName(labelName, model.UTF8Validation) {
 			d.err = fmt.Errorf("%q is not a valid label name for metric %q", labelName, fqName)
 			return d
 		}
@@ -129,7 +129,7 @@ func (v2) NewDesc(fqName, help string, variableLabels ConstrainableLabels, const
 	// cannot be in a regular label name. That prevents matching the label
 	// dimension with a different mix between preset and variable labels.
 	for _, label := range d.variableLabels.names {
-		if !checkLabelName(label) {
+		if !checkLabelName(label, model.UTF8Validation) {
 			d.err = fmt.Errorf("%q is not a valid label name for metric %q", label, fqName)
 			return d
 		}

prometheus/labels.go

@@ -183,6 +183,6 @@ func validateLabelValues(vals []string, expectedNumberOfValues int) error {
 	return nil
 }
 
-func checkLabelName(l string) bool {
-	return model.LabelName(l).IsValid() && !strings.HasPrefix(l, reservedLabelPrefix)
+func checkLabelName(l string, nameValidationSheme model.ValidationScheme) bool {
+	return model.LabelName(l).IsValid(nameValidationSheme) && !strings.HasPrefix(l, reservedLabelPrefix)
 }

prometheus/push/push.go

@@ -70,8 +70,9 @@ type HTTPDoer interface {
 type Pusher struct {
 	error error
 
-	url, job string
-	grouping map[string]string
+	url, job         string
+	grouping         map[string]string
+	validationScheme model.ValidationScheme
 
 	gatherers  prometheus.Gatherers
 	registerer prometheus.Registerer
@@ -84,11 +85,22 @@ type Pusher struct {
 	expfmt expfmt.Format
 }
 
+// Option used to create Pusher instances.
+type Option func(*Pusher)
+
+// WithValidationScheme sets the validation used for label and metric names.
+// Default is model.UTF8Validation.
+func WithValidationScheme(scheme model.ValidationScheme) Option {
+	return func(p *Pusher) {
+		p.validationScheme = scheme
+	}
+}
+
 // New creates a new Pusher to push to the provided URL with the provided job
 // name (which must not be empty). You can use just host:port or ip:port as url,
 // in which case “http://” is added automatically. Alternatively, include the
 // schema in the URL. However, do not include the “/metrics/jobs/…” part.
-func New(url, job string) *Pusher {
+func New(url, job string, opts ...Option) *Pusher {
 	var (
 		reg = prometheus.NewRegistry()
 		err error
@@ -101,16 +113,21 @@ func New(url, job string) *Pusher {
 	}
 	url = strings.TrimSuffix(url, "/")
 
-	return &Pusher{
-		error:      err,
-		url:        url,
-		job:        job,
-		grouping:   map[string]string{},
-		gatherers:  prometheus.Gatherers{reg},
-		registerer: reg,
-		client:     &http.Client{},
-		expfmt:     expfmt.NewFormat(expfmt.TypeProtoDelim),
+	pusher := &Pusher{
+		error:            err,
+		url:              url,
+		job:              job,
+		grouping:         map[string]string{},
+		gatherers:        prometheus.Gatherers{reg},
+		registerer:       reg,
+		client:           &http.Client{},
+		expfmt:           expfmt.NewFormat(expfmt.TypeProtoDelim),
+		validationScheme: model.UTF8Validation,
+	}
+	for _, opt := range opts {
+		opt(pusher)
 	}
+	return pusher
 }
 
 // Push collects/gathers all metrics from all Collectors and Gatherers added to
@@ -182,7 +199,7 @@ func (p *Pusher) Error() error {
 // For convenience, this method returns a pointer to the Pusher itself.
 func (p *Pusher) Grouping(name, value string) *Pusher {
 	if p.error == nil {
-		if !model.LabelName(name).IsValid() {
+		if !model.LabelName(name).IsValid(p.validationScheme) {
 			p.error = fmt.Errorf("grouping label has invalid name: %s", name)
 			return p
 		}

prometheus/registry.go

@@ -31,6 +31,7 @@ import (
 	"github.com/cespare/xxhash/v2"
 	dto "github.com/prometheus/client_model/go"
 	"github.com/prometheus/common/expfmt"
+	"github.com/prometheus/common/model"
 	"google.golang.org/protobuf/proto"
 )
 
@@ -62,14 +63,27 @@ func init() {
 	MustRegister(NewGoCollector())
 }
 
+type RegistryOption func(*Registry)
+
+func WithNameValidationScheme(scheme model.ValidationScheme) RegistryOption {
+	return func(r *Registry) {
+		r.nameValidationScheme = scheme
+	}
+}
+
 // NewRegistry creates a new vanilla Registry without any Collectors
 // pre-registered.
-func NewRegistry() *Registry {
-	return &Registry{
-		collectorsByID:  map[uint64]Collector{},
-		descIDs:         map[uint64]struct{}{},
-		dimHashesByName: map[string]uint64{},
+func NewRegistry(opts ...RegistryOption) *Registry {
+	reg := &Registry{
+		collectorsByID:       map[uint64]Collector{},
+		descIDs:              map[uint64]struct{}{},
+		dimHashesByName:      map[string]uint64{},
+		nameValidationScheme: model.UTF8Validation,
+	}
+	for _, opt := range opts {
+		opt(reg)
 	}
+	return reg
 }
 
 // NewPedanticRegistry returns a registry that checks during collection if each
@@ -264,6 +278,7 @@ type Registry struct {
 	dimHashesByName       map[string]uint64
 	uncheckedCollectors   []Collector
 	pedanticChecksEnabled bool
+	nameValidationScheme  model.ValidationScheme
 }
 
 // Register implements Registerer.
@@ -503,6 +518,7 @@ func (r *Registry) Gather() ([]*dto.MetricFamily, error) {
 				metric, metricFamiliesByName,
 				metricHashes,
 				registeredDescIDs,
+				r.nameValidationScheme,
 			))
 		case metric, ok := <-umc:
 			if !ok {
@@ -513,6 +529,7 @@ func (r *Registry) Gather() ([]*dto.MetricFamily, error) {
 				metric, metricFamiliesByName,
 				metricHashes,
 				nil,
+				r.nameValidationScheme,
 			))
 		default:
 			if goroutineBudget <= 0 || len(checkedCollectors)+len(uncheckedCollectors) == 0 {
@@ -530,6 +547,7 @@ func (r *Registry) Gather() ([]*dto.MetricFamily, error) {
 						metric, metricFamiliesByName,
 						metricHashes,
 						registeredDescIDs,
+						r.nameValidationScheme,
 					))
 				case metric, ok := <-umc:
 					if !ok {
@@ -540,6 +558,7 @@ func (r *Registry) Gather() ([]*dto.MetricFamily, error) {
 						metric, metricFamiliesByName,
 						metricHashes,
 						nil,
+						r.nameValidationScheme,
 					))
 				}
 				break
@@ -622,6 +641,7 @@ func processMetric(
 	metricFamiliesByName map[string]*dto.MetricFamily,
 	metricHashes map[uint64]struct{},
 	registeredDescIDs map[uint64]struct{},
+	nameValidationScheme model.ValidationScheme,
 ) error {
 	desc := metric.Desc()
 	// Wrapped metrics collected by an unchecked Collector can have an
@@ -705,7 +725,7 @@ func processMetric(
 		}
 		metricFamiliesByName[desc.fqName] = metricFamily
 	}
-	if err := checkMetricConsistency(metricFamily, dtoMetric, metricHashes); err != nil {
+	if err := checkMetricConsistency(metricFamily, dtoMetric, metricHashes, nameValidationScheme); err != nil {
 		return err
 	}
 	if registeredDescIDs != nil {
@@ -791,7 +811,8 @@ func (gs Gatherers) Gather() ([]*dto.MetricFamily, error) {
 				metricFamiliesByName[mf.GetName()] = existingMF
 			}
 			for _, m := range mf.Metric {
-				if err := checkMetricConsistency(existingMF, m, metricHashes); err != nil {
+				// TODO(juliusmh): hardcoded UTF8 validation
+				if err := checkMetricConsistency(existingMF, m, metricHashes, model.UTF8Validation); err != nil {
 					errs = append(errs, err)
 					continue
 				}
@@ -870,6 +891,7 @@ func checkMetricConsistency(
 	metricFamily *dto.MetricFamily,
 	dtoMetric *dto.Metric,
 	metricHashes map[uint64]struct{},
+	nameValidationScheme model.ValidationScheme,
 ) error {
 	name := metricFamily.GetName()
 
@@ -894,7 +916,7 @@ func checkMetricConsistency(
 				name, dtoMetric, labelName,
 			)
 		}
-		if !checkLabelName(labelName) {
+		if !checkLabelName(labelName, nameValidationScheme) {
 			return fmt.Errorf(
 				"collected metric %q { %s} has a label with an invalid name: %s",
 				name, dtoMetric, labelName,

prometheus/testutil/lint.go

@@ -18,29 +18,30 @@ import (
 
 	"github.com/prometheus/client_golang/prometheus"
 	"github.com/prometheus/client_golang/prometheus/testutil/promlint"
+	"github.com/prometheus/common/model"
 )
 
 // CollectAndLint registers the provided Collector with a newly created pedantic
 // Registry. It then calls GatherAndLint with that Registry and with the
 // provided metricNames.
-func CollectAndLint(c prometheus.Collector, metricNames ...string) ([]promlint.Problem, error) {
+func CollectAndLint(c prometheus.Collector, validationScheme model.ValidationScheme, metricNames ...string) ([]promlint.Problem, error) {
 	reg := prometheus.NewPedanticRegistry()
 	if err := reg.Register(c); err != nil {
 		return nil, fmt.Errorf("registering collector failed: %w", err)
 	}
-	return GatherAndLint(reg, metricNames...)
+	return GatherAndLint(reg, validationScheme, metricNames...)
 }
 
 // GatherAndLint gathers all metrics from the provided Gatherer and checks them
 // with the linter in the promlint package. If any metricNames are provided,
 // only metrics with those names are checked.
-func GatherAndLint(g prometheus.Gatherer, metricNames ...string) ([]promlint.Problem, error) {
+func GatherAndLint(g prometheus.Gatherer, validationScheme model.ValidationScheme, metricNames ...string) ([]promlint.Problem, error) {
 	got, err := g.Gather()
 	if err != nil {
 		return nil, fmt.Errorf("gathering metrics failed: %w", err)
 	}
 	if metricNames != nil {
 		got = filterMetrics(got, metricNames)
 	}
-	return promlint.NewWithMetricFamilies(got).Lint()
+	return promlint.NewWithMetricFamilies(got).Lint(validationScheme)
 }

prometheus/testutil/promlint/promlint.go

@@ -21,6 +21,7 @@ import (
 
 	dto "github.com/prometheus/client_model/go"
 	"github.com/prometheus/common/expfmt"
+	"github.com/prometheus/common/model"
 )
 
 // A Linter is a Prometheus metrics linter.  It identifies issues with metric
@@ -64,15 +65,15 @@ func (l *Linter) AddCustomValidations(vs ...Validation) {
 // Lint performs a linting pass, returning a slice of Problems indicating any
 // issues found in the metrics stream. The slice is sorted by metric name
 // and issue description.
-func (l *Linter) Lint() ([]Problem, error) {
+func (l *Linter) Lint(nameValidationScheme model.ValidationScheme) ([]Problem, error) {
 	var problems []Problem
 
 	if l.r != nil {
 		d := expfmt.NewDecoder(l.r, expfmt.NewFormat(expfmt.TypeTextPlain))
 
 		mf := &dto.MetricFamily{}
 		for {
-			if err := d.Decode(mf); err != nil {
+			if err := d.Decode(mf, nameValidationScheme); err != nil {
 				if errors.Is(err, io.EOF) {
 					break
 				}

prometheus/value.go

@@ -23,6 +23,7 @@ import (
 	"github.com/prometheus/client_golang/prometheus/internal"
 
 	dto "github.com/prometheus/client_model/go"
+	"github.com/prometheus/common/model"
 	"google.golang.org/protobuf/proto"
 	"google.golang.org/protobuf/types/known/timestamppb"
 )
@@ -253,7 +254,8 @@ func newExemplar(value float64, ts time.Time, l Labels) (*dto.Exemplar, error) {
 	labelPairs := make([]*dto.LabelPair, 0, len(l))
 	var runes int
 	for name, value := range l {
-		if !checkLabelName(name) {
+		// TODO(juliusmh): hardcoded UTF8 validation
+		if !checkLabelName(name, model.UTF8Validation) {
 			return nil, fmt.Errorf("exemplar label name %q is invalid", name)
 		}
 		runes += utf8.RuneCountInString(name)