lib/protoparser/prometheus: properly unescape label values in Prometheus exposition format

Unescape only `\n`, `\"` and `\\` sequences as Prometheus does. Other escape sequences shouldn't be unescaped.

docs/CHANGELOG.md

@@ -31,6 +31,7 @@
 * BUGFIX: fix arm64 builds due to the issue in `github.com/golang/snappy`. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1074
 * BUGFIX: fix `index out of range [1024819115206086200] with length 27` panic, which could occur when `1e-9` value is passed to VictoriaMetrics histogram. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1096
 * BUGFIX: fix parsing for Graphite line with empty tags such as `foo; 123 456`. See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1100
+* BUGFIX: unescape only `\\`, `\n` and `\"` in label names when parsing Prometheus text exposition format as Prometheus does. Previously other escape sequences could be improperly unescaped.
 
 
 # [v1.54.1](https://github.com/VictoriaMetrics/VictoriaMetrics/releases/tag/v1.54.1)

lib/protoparser/prometheus/parser.go

@@ -2,7 +2,6 @@ package prometheus
 
 import (
 	"fmt"
-	"strconv"
 	"strings"
 
 	"github.com/VictoriaMetrics/VictoriaMetrics/lib/logger"
@@ -264,11 +263,7 @@ func unmarshalTags(dst []Tag, s string, noEscapes bool) (string, []Tag, error) {
 			if n < 0 {
 				return s, dst, fmt.Errorf("missing closing quote for tag value %q", s)
 			}
-			var err error
-			value, err = unescapeValue(s[:n+1])
-			if err != nil {
-				return s, dst, fmt.Errorf("cannot unescape value %q for tag %q: %w", s[:n+1], key, err)
-			}
+			value = unescapeValue(s[1:n])
 			s = s[n+1:]
 		}
 		if len(key) > 0 {
@@ -324,16 +319,41 @@ func findClosingQuote(s string) int {
 	}
 }
 
-func unescapeValue(s string) (string, error) {
-	if len(s) < 2 || s[0] != '"' || s[len(s)-1] != '"' {
-		return "", fmt.Errorf("unexpected tag value: %q", s)
-	}
+func unescapeValue(s string) string {
 	n := strings.IndexByte(s, '\\')
 	if n < 0 {
 		// Fast path - nothing to unescape
-		return s[1 : len(s)-1], nil
+		return s
 	}
-	return strconv.Unquote(s)
+	b := make([]byte, 0, len(s))
+	for {
+		b = append(b, s[:n]...)
+		s = s[n+1:]
+		if len(s) == 0 {
+			b = append(b, '\\')
+			break
+		}
+		// label_value can be any sequence of UTF-8 characters, but the backslash (\), double-quote ("),
+		// and line feed (\n) characters have to be escaped as \\, \", and \n, respectively.
+		// See https://github.com/prometheus/docs/blob/master/content/docs/instrumenting/exposition_formats.md
+		switch s[0] {
+		case '\\':
+			b = append(b, '\\')
+		case '"':
+			b = append(b, '"')
+		case 'n':
+			b = append(b, '\n')
+		default:
+			b = append(b, '\\', s[0])
+		}
+		s = s[1:]
+		n = strings.IndexByte(s, '\\')
+		if n < 0 {
+			b = append(b, s...)
+			break
+		}
+	}
+	return string(b)
 }
 
 func prevBackslashesCount(s string) int {

lib/protoparser/prometheus/parser_test.go

@@ -46,41 +46,22 @@ func TestFindClosingQuote(t *testing.T) {
 	f(`"foo\"bar\"baz"`, 14)
 }
 
-func TestUnescapeValueFailure(t *testing.T) {
-	f := func(s string) {
-		t.Helper()
-		ss, err := unescapeValue(s)
-		if err == nil {
-			t.Fatalf("expecting error")
-		}
-		if ss != "" {
-			t.Fatalf("expecting empty string; got %q", ss)
-		}
-	}
-	f(``)
-	f(`foobar`)
-	f(`"foobar`)
-	f(`foobar"`)
-	f(`"foobar\"`)
-	f(` "foobar"`)
-	f(`"foobar" `)
-}
-
-func TestUnescapeValueSuccess(t *testing.T) {
+func TestUnescapeValue(t *testing.T) {
 	f := func(s, resultExpected string) {
 		t.Helper()
-		result, err := unescapeValue(s)
-		if err != nil {
-			t.Fatalf("unexpected error: %s", err)
-		}
+		result := unescapeValue(s)
 		if result != resultExpected {
 			t.Fatalf("unexpected result; got %q; want %q", result, resultExpected)
 		}
 	}
-	f(`""`, "")
-	f(`"f"`, "f")
-	f(`"foobar"`, "foobar")
-	f(`"\"\n\t"`, "\"\n\t")
+	f(``, "")
+	f(`f`, "f")
+	f(`foobar`, "foobar")
+	f(`\"\n\t`, "\"\n\\t")
+
+	// Edge cases
+	f(`foo\bar`, "foo\\bar")
+	f(`foo\`, "foo\\")
 }
 
 func TestRowsUnmarshalFailure(t *testing.T) {
@@ -203,6 +184,24 @@ cassandra_token_ownership_ratio 78.9`, &Rows{
 		}},
 	})
 
+	// Incorrectly escaped backlash. This is real-world case, which must be supported.
+	f(`mssql_sql_server_active_transactions_sec{loginname="domain\somelogin",env="develop"} 56`, &Rows{
+		Rows: []Row{{
+			Metric: "mssql_sql_server_active_transactions_sec",
+			Tags: []Tag{
+				{
+					Key: "loginname",
+					Value: "domain\\somelogin",
+				},
+				{
+					Key: "env",
+					Value: "develop",
+				},
+			},
+			Value: 56,
+		}},
+	})
+
 	// Exemplars - see https://github.com/OpenObservability/OpenMetrics/blob/master/OpenMetrics.md#exemplars-1
 	f(`foo_bucket{le="10",a="#b"} 17 # {trace_id="oHg5SJ#YRHA0"} 9.8 1520879607.789
 	   abc 123 456 # foobar