envcar: refine normalization

CHANGELOG.md

@@ -12,6 +12,10 @@ This project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.htm
 
 - Configuration file can now be set via `OTEL_CONFIG_FILE` in `go.opentelemetry.io/contrib/otelconf`. (#8639)
 
+### Changed
+
+- Prepend `_` to the normalized environment variable name when the key starts with a digit in `go.opentelemetry.io/contrib/propagators/envcar`, ensuring POSIX compliance. (#8678)
+
 ### Fixed
 
 - Limit the request body size at 1MB in `go.opentelemetry.io/contrib/zpages`. (#8656)

propagators/envcar/carrier.go

@@ -11,23 +11,6 @@ import (
 	"go.opentelemetry.io/otel/propagation"
 )
 
-// upperWithUnderscores converts a string so that A-Z and 0-9 and _ are kept
-// as-is, a-z is uppercased, and all other characters are replaced with _.
-func upperWithUnderscores(s string) string {
-	b := make([]byte, 0, len(s))
-	for _, r := range s {
-		switch {
-		case r >= 'A' && r <= 'Z', r >= '0' && r <= '9', r == '_':
-			b = append(b, byte(r)) //nolint:gosec // G115: overflow is already checked.
-		case r >= 'a' && r <= 'z':
-			b = append(b, byte(r+'A'-'a'))
-		default:
-			b = append(b, '_')
-		}
-	}
-	return string(b)
-}
-
 // Carrier is a TextMapCarrier that uses the environment variables as a
 // storage medium for propagated key-value pairs. The keys are normalised
 // before being used to access the environment variables.
@@ -71,7 +54,7 @@ func (c *Carrier) fetch() {
 // environment and all future reads will be from that store.
 func (c *Carrier) Get(key string) string {
 	c.fetch()
-	return c.values[upperWithUnderscores(key)]
+	return c.values[normalize(key)]
 }
 
 // Set stores the key-value pair in the environment variable.
@@ -82,7 +65,7 @@ func (c *Carrier) Set(key, value string) {
 	if c.SetEnvFunc == nil {
 		return
 	}
-	k := upperWithUnderscores(key)
+	k := normalize(key)
 	c.SetEnvFunc(k, value)
 }
 

propagators/envcar/normalize.go

@@ -0,0 +1,48 @@
+// Copyright The OpenTelemetry Authors
+// SPDX-License-Identifier: Apache-2.0
+
+package envcar // import "go.opentelemetry.io/contrib/propagators/envcar"
+
+import (
+	"unicode/utf8"
+)
+
+// normalize converts s to a valid POSIX environment variable name.
+// The conversion rules are:
+//   - A–Z, 0–9, and _ are kept as-is.
+//   - a–z are uppercased.
+//   - All other characters are replaced with _.
+//   - If the result would start with a digit, an underscore is prepended.
+func normalize(s string) string {
+	if s == "" {
+		return ""
+	}
+
+	// Pre-allocate the exact output length. If the first byte is a digit,
+	// the name must be prefixed with '_', so allocate one extra byte.
+	var b []byte
+	i := 0
+	if s[0] >= '0' && s[0] <= '9' {
+		b = make([]byte, utf8.RuneCountInString(s)+1)
+		b[0] = '_'
+		i = 1
+	} else {
+		b = make([]byte, utf8.RuneCountInString(s))
+	}
+
+	for _, r := range s {
+		switch {
+		case r >= 'A' && r <= 'Z', r >= '0' && r <= '9', r == '_':
+			// Uppercase letters, digits, and underscores are valid as-is.
+			b[i] = byte(r) //nolint:gosec // G115: overflow is not possible.
+		case r >= 'a' && r <= 'z':
+			// Lowercase letters are converted to uppercase.
+			b[i] = byte(r + 'A' - 'a')
+		default:
+			// All other characters (including non-ASCII runes) become underscores.
+			b[i] = '_'
+		}
+		i++
+	}
+	return string(b)
+}

propagators/envcar/normalize_test.go

@@ -0,0 +1,49 @@
+// Copyright The OpenTelemetry Authors
+// SPDX-License-Identifier: Apache-2.0
+
+package envcar
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+var normalizeCases = []struct {
+	in, want string
+}{
+	{"", ""},
+	{"ABC", "ABC"},
+	{"abc", "ABC"},
+	{"01239", "_01239"},
+	{"0abc", "_0ABC"},
+	{"9", "_9"},
+	{"a_b_c", "A_B_C"},
+	{"hello-world", "HELLO_WORLD"},
+	{"foo.bar", "FOO_BAR"},
+	{"Content-Type", "CONTENT_TYPE"},
+	{"traceparent", "TRACEPARENT"},
+	{"key with spaces", "KEY_WITH_SPACES"},
+	{"MiXeD_123!", "MIXED_123_"},
+	{"🧳", "_"},
+	{"Mój Bagaż", "M_J_BAGA_"},
+}
+
+func TestNormalize(t *testing.T) {
+	for _, tc := range normalizeCases {
+		t.Run(tc.in, func(t *testing.T) {
+			assert.Equal(t, tc.want, normalize(tc.in))
+		})
+	}
+}
+
+func BenchmarkNormalize(b *testing.B) {
+	for _, tc := range normalizeCases {
+		b.Run(tc.in, func(b *testing.B) {
+			b.ReportAllocs()
+			for b.Loop() {
+				normalize(tc.in)
+			}
+		})
+	}
+}