wip

lib/regexutil/promregex.go

@@ -23,9 +23,11 @@ type PromRegex struct {
 	// For example, prefix="foo" for regex="foo(a|b)"
 	prefix string
 
-	// Suffix contains regex suffix left after removing the prefix.
-	// For example, suffix="a|b" for regex="foo(a|b)"
-	suffix string
+	// isSuffixDotStar is set to true if suffix is ".*"
+	isSuffixDotStar bool
+
+	// isSuffixDotPlus is set to true if suffix is ".+"
+	isSuffixDotPlus bool
 
 	// substrDotStar contains literal string for regex suffix=".*string.*"
 	substrDotStar string
@@ -48,8 +50,10 @@ func NewPromRegex(expr string) (*PromRegex, error) {
 	}
 	prefix, suffix := SimplifyPromRegex(expr)
 	orValues := GetOrValuesPromRegex(suffix)
-	substrDotStar := getSubstringLiteral(suffix, ".*")
-	substrDotPlus := getSubstringLiteral(suffix, ".+")
+	isSuffixDotStar := isDotOpRegexp(suffix, syntax.OpStar)
+	isSuffixDotPlus := isDotOpRegexp(suffix, syntax.OpPlus)
+	substrDotStar := getSubstringLiteral(suffix, syntax.OpStar)
+	substrDotPlus := getSubstringLiteral(suffix, syntax.OpPlus)
 	// It is expected that Optimize returns valid regexp in suffix, so use MustCompile here.
 	// Anchor suffix to the beginning and the end of the matching string.
 	suffixExpr := "^(?:" + suffix + ")$"
@@ -57,7 +61,8 @@ func NewPromRegex(expr string) (*PromRegex, error) {
 	reSuffixMatcher := bytesutil.NewFastStringMatcher(reSuffix.MatchString)
 	pr := &PromRegex{
 		prefix:          prefix,
-		suffix:          suffix,
+		isSuffixDotStar: isSuffixDotStar,
+		isSuffixDotPlus: isSuffixDotPlus,
 		substrDotStar:   substrDotStar,
 		substrDotPlus:   substrDotPlus,
 		orValues:        orValues,
@@ -71,19 +76,21 @@ func NewPromRegex(expr string) (*PromRegex, error) {
 // The pr is automatically anchored to the beginning and to the end
 // of the matching string with '^' and '$'.
 func (pr *PromRegex) MatchString(s string) bool {
-	if !strings.HasPrefix(s, pr.prefix) {
-		// Fast path - s has another prefix than pr.
-		return false
-	}
-	s = s[len(pr.prefix):]
-	if len(pr.orValues) > 0 {
-		// Fast path - pr contains only alternate strings such as 'foo|bar|baz'
-		for _, v := range pr.orValues {
-			if s == v {
-				return true
-			}
+	if len(pr.prefix) > 0 {
+		if !strings.HasPrefix(s, pr.prefix) {
+			// Fast path - s has another prefix than pr.
+			return false
 		}
-		return false
+		s = s[len(pr.prefix):]
+	}
+
+	if pr.isSuffixDotStar {
+		// Fast path - the pr contains "prefix.*"
+		return true
+	}
+	if pr.isSuffixDotPlus {
+		// Fast path - the pr contains "prefix.+"
+		return len(s) > 0
 	}
 	if pr.substrDotStar != "" {
 		// Fast path - pr contains ".*someText.*"
@@ -94,45 +101,17 @@ func (pr *PromRegex) MatchString(s string) bool {
 		n := strings.Index(s, pr.substrDotPlus)
 		return n > 0 && n+len(pr.substrDotPlus) < len(s)
 	}
-	switch pr.suffix {
-	case ".*":
-		// Fast path - the pr contains "prefix.*"
-		return true
-	case ".+":
-		// Fast path - the pr contains "prefix.+"
-		return len(s) > 0
+
+	if len(pr.orValues) > 0 {
+		// Fast path - pr contains only alternate strings such as 'foo|bar|baz'
+		for _, v := range pr.orValues {
+			if s == v {
+				return true
+			}
+		}
+		return false
 	}
+
 	// Fall back to slow path by matching the original regexp.
 	return pr.reSuffixMatcher.Match(s)
 }
-
-// getSubstringLiteral returns regex part from expr surrounded by prefixSuffix.
-//
-// For example, if expr=".+foo.+" and prefixSuffix=".+", then the function returns "foo".
-//
-// An empty string is returned if expr doesn't contain the given prefixSuffix prefix and suffix
-// or if the regex part surrounded by prefixSuffix contains alternate regexps.
-func getSubstringLiteral(expr, prefixSuffix string) string {
-	// Verify that the expr doesn't contain alternate regexps. In this case it is unsafe removing prefix and suffix.
-	sre, err := syntax.Parse(expr, syntax.Perl)
-	if err != nil {
-		return ""
-	}
-	if sre.Op == syntax.OpAlternate {
-		return ""
-	}
-
-	if !strings.HasPrefix(expr, prefixSuffix) {
-		return ""
-	}
-	expr = expr[len(prefixSuffix):]
-	if !strings.HasSuffix(expr, prefixSuffix) {
-		return ""
-	}
-	expr = expr[:len(expr)-len(prefixSuffix)]
-	prefix, suffix := SimplifyPromRegex(expr)
-	if suffix != "" {
-		return ""
-	}
-	return prefix
-}

lib/regexutil/regex.go

@@ -0,0 +1,165 @@
+package regexutil
+
+import (
+	"regexp"
+	"regexp/syntax"
+	"strings"
+)
+
+// Regex implements an optimized string matching for Go regex.
+//
+// The following regexs are optimized:
+//
+// - plain string such as "foobar"
+// - alternate strings such as "foo|bar|baz"
+// - prefix match such as "foo.*" or "foo.+"
+// - substring match such as ".*foo.*" or ".+bar.+"
+type Regex struct {
+	// prefix contains literal prefix for regex.
+	// For example, prefix="foo" for regex="foo(a|b)"
+	prefix string
+
+	// isSuffixDotStar is set to true if suffix is ".*"
+	isSuffixDotStar bool
+
+	// isSuffixDotPlus is set to true if suffix is ".+"
+	isSuffixDotPlus bool
+
+	// substrDotStar contains literal string for regex suffix=".*string.*"
+	substrDotStar string
+
+	// substrDotPlus contains literal string for regex suffix=".+string.+"
+	substrDotPlus string
+
+	// orValues contains or values for the suffix regex.
+	// For example, orValues contain ["foo","bar","baz"] for regex suffix="foo|bar|baz"
+	orValues []string
+
+	// re is the original regexp.
+	re *regexp.Regexp
+}
+
+// NewRegex returns Regex for the given expr.
+func NewRegex(expr string) (*Regex, error) {
+	if _, err := regexp.Compile(expr); err != nil {
+		return nil, err
+	}
+	prefix, suffix := SimplifyRegex(expr)
+	orValues := GetOrValuesRegex(suffix)
+	isSuffixDotStar := isDotOpRegexp(suffix, syntax.OpStar)
+	isSuffixDotPlus := isDotOpRegexp(suffix, syntax.OpPlus)
+	substrDotStar := getSubstringLiteral(suffix, syntax.OpStar)
+	substrDotPlus := getSubstringLiteral(suffix, syntax.OpPlus)
+
+	var re *regexp.Regexp
+	if len(orValues) == 0 && substrDotStar == "" && substrDotPlus == "" && suffix != ".*" && suffix != ".+" {
+		suffixAnchored := suffix
+		if len(prefix) > 0 {
+			suffixAnchored = "^(?:" + suffix + ")"
+		}
+		// The suffixAnchored must be properly compiled, since it has been already checked above.
+		// Otherwise it is a bug, which must be fixed.
+		re = regexp.MustCompile(suffixAnchored)
+	}
+	r := &Regex{
+		prefix:          prefix,
+		isSuffixDotStar: isSuffixDotStar,
+		isSuffixDotPlus: isSuffixDotPlus,
+		substrDotStar:   substrDotStar,
+		substrDotPlus:   substrDotPlus,
+		orValues:        orValues,
+		re:              re,
+	}
+	return r, nil
+}
+
+// MatchString returns true if s matches pr.
+func (r *Regex) MatchString(s string) bool {
+	if len(r.prefix) == 0 {
+		return r.matchStringNoPrefix(s)
+	}
+	return r.matchStringWithPrefix(s)
+}
+
+func (r *Regex) matchStringNoPrefix(s string) bool {
+	if r.isSuffixDotStar {
+		return true
+	}
+	if r.isSuffixDotPlus {
+		return len(s) > 0
+	}
+	if r.substrDotStar != "" {
+		// Fast path - r contains ".*someText.*"
+		return strings.Contains(s, r.substrDotStar)
+	}
+	if r.substrDotPlus != "" {
+		// Fast path - r contains ".+someText.+"
+		n := strings.Index(s, r.substrDotPlus)
+		return n > 0 && n+len(r.substrDotPlus) < len(s)
+	}
+
+	if len(r.orValues) == 0 {
+		// Fall back to slow path by matching the original regexp.
+		return r.re.MatchString(s)
+	}
+
+	// Fast path - compare s to pr.orValues
+	for _, v := range r.orValues {
+		if strings.Contains(s, v) {
+			return true
+		}
+	}
+	return false
+}
+
+func (r *Regex) matchStringWithPrefix(s string) bool {
+	n := strings.Index(s, r.prefix)
+	if n < 0 {
+		// Fast path - s doesn't contain the needed prefix
+		return false
+	}
+	sNext := s[n+1:]
+	s = s[n+len(r.prefix):]
+
+	if r.isSuffixDotStar {
+		return true
+	}
+	if r.isSuffixDotPlus {
+		return len(s) > 0
+	}
+	if r.substrDotStar != "" {
+		// Fast path - r contains ".*someText.*"
+		return strings.Contains(s, r.substrDotStar)
+	}
+	if r.substrDotPlus != "" {
+		// Fast path - r contains ".+someText.+"
+		n := strings.Index(s, r.substrDotPlus)
+		return n > 0 && n+len(r.substrDotPlus) < len(s)
+	}
+
+	for {
+		if len(r.orValues) == 0 {
+			// Fall back to slow path by matching the original regexp.
+			if r.re.MatchString(s) {
+				return true
+			}
+		} else {
+			// Fast path - compare s to pr.orValues
+			for _, v := range r.orValues {
+				if strings.HasPrefix(s, v) {
+					return true
+				}
+			}
+		}
+
+		// Mismatch. Try again starting from the next char.
+		s = sNext
+		n := strings.Index(s, r.prefix)
+		if n < 0 {
+			// Fast path - s doesn't contain the needed prefix
+			return false
+		}
+		sNext = s[n+1:]
+		s = s[n+len(r.prefix):]
+	}
+}

lib/regexutil/regex_test.go

@@ -0,0 +1,125 @@
+package regexutil
+
+import (
+	"testing"
+)
+
+func TestRegexMatchString(t *testing.T) {
+	f := func(regex, s string, resultExpected bool) {
+		t.Helper()
+
+		re, err := NewRegex(regex)
+		if err != nil {
+			t.Fatalf("cannot parse %q: %s", regex, err)
+		}
+		result := re.MatchString(s)
+		if result != resultExpected {
+			t.Fatalf("unexpected result when matching %q against regex=%q; got %v; want %v", s, regex, result, resultExpected)
+		}
+	}
+
+	f("", "", true)
+	f("", "foo", true)
+	f("foo", "", false)
+	f(".*", "", true)
+	f(".*", "foo", true)
+	f(".+", "", false)
+	f(".+", "foo", true)
+	f("foo.*", "bar", false)
+	f("foo.*", "foo", true)
+	f("foo.*", "a foo", true)
+	f("foo.*", "a foo a", true)
+	f("foo.*", "foobar", true)
+	f("foo.*", "a foobar", true)
+	f("foo.+", "bar", false)
+	f("foo.+", "foo", false)
+	f("foo.+", "a foo", false)
+	f("foo.+", "foobar", true)
+	f("foo.+", "a foobar", true)
+	f("foo|bar", "", false)
+	f("foo|bar", "a", false)
+	f("foo|bar", "foo", true)
+	f("foo|bar", "a foo", true)
+	f("foo|bar", "foo a", true)
+	f("foo|bar", "a foo a", true)
+	f("foo|bar", "bar", true)
+	f("foo|bar", "foobar", true)
+	f("foo(bar|baz)", "a", false)
+	f("foo(bar|baz)", "foobar", true)
+	f("foo(bar|baz)", "foobaz", true)
+	f("foo(bar|baz)", "foobaza", true)
+	f("foo(bar|baz)", "a foobaz a", true)
+	f("foo(bar|baz)", "foobal", false)
+	f("^foo|b(ar)$", "foo", true)
+	f("^foo|b(ar)$", "foo a", true)
+	f("^foo|b(ar)$", "a foo", false)
+	f("^foo|b(ar)$", "bar", true)
+	f("^foo|b(ar)$", "a bar", true)
+	f("^foo|b(ar)$", "barz", false)
+	f("^foo|b(ar)$", "ar", false)
+	f(".*foo.*", "foo", true)
+	f(".*foo.*", "afoobar", true)
+	f(".*foo.*", "abc", false)
+	f("foo.*bar.*", "foobar", true)
+	f("foo.*bar.*", "foo_bar_", true)
+	f("foo.*bar.*", "a foo bar baz", true)
+	f("foo.*bar.*", "foobaz", false)
+	f("foo.*bar.*", "baz foo", false)
+	f(".+foo.+", "foo", false)
+	f(".+foo.+", "afoobar", true)
+	f(".+foo.+", "afoo", false)
+	f(".+foo.+", "abc", false)
+	f("foo.+bar.+", "foobar", false)
+	f("foo.+bar.+", "foo_bar_", true)
+	f("foo.+bar.+", "a foo_bar_", true)
+	f("foo.+bar.+", "foobaz", false)
+	f("foo.+bar.+", "abc", false)
+	f(".+foo.*", "foo", false)
+	f(".+foo.*", "afoo", true)
+	f(".+foo.*", "afoobar", true)
+	f(".*(a|b).*", "a", true)
+	f(".*(a|b).*", "ax", true)
+	f(".*(a|b).*", "xa", true)
+	f(".*(a|b).*", "xay", true)
+	f(".*(a|b).*", "xzy", false)
+	f("^(?:true)$", "true", true)
+	f("^(?:true)$", "false", false)
+
+	f(".+;|;.+", ";", false)
+	f(".+;|;.+", "foo", false)
+	f(".+;|;.+", "foo;bar", true)
+	f(".+;|;.+", "foo;", true)
+	f(".+;|;.+", ";foo", true)
+	f(".+foo|bar|baz.+", "foo", false)
+	f(".+foo|bar|baz.+", "afoo", true)
+	f(".+foo|bar|baz.+", "fooa", false)
+	f(".+foo|bar|baz.+", "afooa", true)
+	f(".+foo|bar|baz.+", "bar", true)
+	f(".+foo|bar|baz.+", "abar", true)
+	f(".+foo|bar|baz.+", "abara", true)
+	f(".+foo|bar|baz.+", "bara", true)
+	f(".+foo|bar|baz.+", "baz", false)
+	f(".+foo|bar|baz.+", "baza", true)
+	f(".+foo|bar|baz.+", "abaz", false)
+	f(".+foo|bar|baz.+", "abaza", true)
+	f(".+foo|bar|baz.+", "afoo|bar|baza", true)
+	f(".+(foo|bar|baz).+", "bar", false)
+	f(".+(foo|bar|baz).+", "bara", false)
+	f(".+(foo|bar|baz).+", "abar", false)
+	f(".+(foo|bar|baz).+", "abara", true)
+	f(".+(foo|bar|baz).+", "afooa", true)
+	f(".+(foo|bar|baz).+", "abaza", true)
+
+	f(".*;|;.*", ";", true)
+	f(".*;|;.*", "foo", false)
+	f(".*;|;.*", "foo;bar", true)
+	f(".*;|;.*", "foo;", true)
+	f(".*;|;.*", ";foo", true)
+
+	f("^bar", "foobarbaz", false)
+	f("^foo", "foobarbaz", true)
+	f("bar$", "foobarbaz", false)
+	f("baz$", "foobarbaz", true)
+	f("(bar$|^foo)", "foobarbaz", true)
+	f("(bar$^boo)", "foobarbaz", false)
+}

lib/regexutil/regexutil.go

@@ -18,6 +18,16 @@ func RemoveStartEndAnchors(expr string) string {
 	return expr
 }
 
+// GetOrValuesRegex returns "or" values from the given regexp expr.
+//
+// It returns ["foo", "bar"] for "foo|bar" regexp.
+// It returns ["foo"] for "foo" regexp.
+// It returns [""] for "" regexp.
+// It returns an empty list if it is impossible to extract "or" values from the regexp.
+func GetOrValuesRegex(expr string) []string {
+	return getOrValuesRegex(expr, true)
+}
+
 // GetOrValuesPromRegex returns "or" values from the given Prometheus-like regexp expr.
 //
 // It ignores start and end anchors ('^') and ('$') at the start and the end of expr.
@@ -27,15 +37,19 @@ func RemoveStartEndAnchors(expr string) string {
 // It returns an empty list if it is impossible to extract "or" values from the regexp.
 func GetOrValuesPromRegex(expr string) []string {
 	expr = RemoveStartEndAnchors(expr)
-	prefix, tailExpr := SimplifyPromRegex(expr)
+	return getOrValuesRegex(expr, false)
+}
+
+func getOrValuesRegex(expr string, keepAnchors bool) []string {
+	prefix, tailExpr := simplifyRegex(expr, keepAnchors)
 	if tailExpr == "" {
 		return []string{prefix}
 	}
-	sre, err := syntax.Parse(tailExpr, syntax.Perl)
+	sre, err := syntax.Parse(tailExpr, regexParseFlags)
 	if err != nil {
 		panic(fmt.Errorf("BUG: unexpected error when parsing verified tailExpr=%q: %w", tailExpr, err))
 	}
-	orValues := getOrValuesExt(sre)
+	orValues := getOrValues(sre)
 
 	// Sort orValues for faster index seek later
 	sort.Strings(orValues)
@@ -50,10 +64,10 @@ func GetOrValuesPromRegex(expr string) []string {
 	return orValues
 }
 
-func getOrValuesExt(sre *syntax.Regexp) []string {
+func getOrValues(sre *syntax.Regexp) []string {
 	switch sre.Op {
 	case syntax.OpCapture:
-		return getOrValuesExt(sre.Sub[0])
+		return getOrValues(sre.Sub[0])
 	case syntax.OpLiteral:
 		if !isLiteral(sre) {
 			return nil
@@ -64,7 +78,7 @@ func getOrValuesExt(sre *syntax.Regexp) []string {
 	case syntax.OpAlternate:
 		a := make([]string, 0, len(sre.Sub))
 		for _, reSub := range sre.Sub {
-			ca := getOrValuesExt(reSub)
+			ca := getOrValues(reSub)
 			if len(ca) == 0 {
 				return nil
 			}
@@ -94,7 +108,7 @@ func getOrValuesExt(sre *syntax.Regexp) []string {
 		if len(sre.Sub) < 1 {
 			return []string{""}
 		}
-		prefixes := getOrValuesExt(sre.Sub[0])
+		prefixes := getOrValues(sre.Sub[0])
 		if len(prefixes) == 0 {
 			return nil
 		}
@@ -102,7 +116,7 @@ func getOrValuesExt(sre *syntax.Regexp) []string {
 			return prefixes
 		}
 		sre.Sub = sre.Sub[1:]
-		suffixes := getOrValuesExt(sre)
+		suffixes := getOrValues(sre)
 		if len(suffixes) == 0 {
 			return nil
 		}
@@ -132,21 +146,33 @@ func isLiteral(sre *syntax.Regexp) bool {
 
 const maxOrValues = 100
 
+// SimplifyRegex simplifies the given regexp expr.
+//
+// It returns plaintext pefix and the remaining regular expression
+// without capturing parens.
+func SimplifyRegex(expr string) (string, string) {
+	return simplifyRegex(expr, true)
+}
+
 // SimplifyPromRegex simplifies the given Prometheus-like expr.
 //
 // It returns plaintext prefix and the remaining regular expression
-// with dropped '^' and '$' anchors at the beginning and the end
+// with dropped '^' and '$' anchors at the beginning and at the end
 // of the regular expression.
 //
 // The function removes capturing parens from the expr,
 // so it cannot be used when capturing parens are necessary.
 func SimplifyPromRegex(expr string) (string, string) {
-	sre, err := syntax.Parse(expr, syntax.Perl)
+	return simplifyRegex(expr, false)
+}
+
+func simplifyRegex(expr string, keepAnchors bool) (string, string) {
+	sre, err := syntax.Parse(expr, regexParseFlags)
 	if err != nil {
 		// Cannot parse the regexp. Return it all as prefix.
 		return expr, ""
 	}
-	sre = simplifyRegexp(sre, false)
+	sre = simplifyRegexp(sre, keepAnchors, keepAnchors)
 	if sre == emptyRegexp {
 		return "", ""
 	}
@@ -162,7 +188,7 @@ func SimplifyPromRegex(expr string) (string, string) {
 			if len(sre.Sub) == 0 {
 				return prefix, ""
 			}
-			sre = simplifyRegexp(sre, true)
+			sre = simplifyRegexp(sre, true, keepAnchors)
 		}
 	}
 	if _, err := syntax.Compile(sre); err != nil {
@@ -171,17 +197,19 @@ func SimplifyPromRegex(expr string) (string, string) {
 	}
 	s := sre.String()
 	s = strings.ReplaceAll(s, "(?:)", "")
-	s = strings.ReplaceAll(s, "(?-s:.)", ".")
-	s = strings.ReplaceAll(s, "(?-m:$)", "$")
+	s = strings.ReplaceAll(s, "(?s:.)", ".")
+	s = strings.ReplaceAll(s, "(?m:$)", "$")
 	return prefix, s
 }
 
-func simplifyRegexp(sre *syntax.Regexp, hasPrefix bool) *syntax.Regexp {
+func simplifyRegexp(sre *syntax.Regexp, keepBeginOp, keepEndOp bool) *syntax.Regexp {
 	s := sre.String()
 	for {
-		sre = simplifyRegexpExt(sre, hasPrefix, false)
+		sre = simplifyRegexpExt(sre, keepBeginOp, keepEndOp)
 		sre = sre.Simplify()
-		if sre.Op == syntax.OpBeginText || sre.Op == syntax.OpEndText {
+		if !keepBeginOp && sre.Op == syntax.OpBeginText {
+			sre = emptyRegexp
+		} else if !keepEndOp && sre.Op == syntax.OpEndText {
 			sre = emptyRegexp
 		}
 		sNew := sre.String()
@@ -189,7 +217,7 @@ func simplifyRegexp(sre *syntax.Regexp, hasPrefix bool) *syntax.Regexp {
 			return sre
 		}
 		var err error
-		sre, err = syntax.Parse(sNew, syntax.Perl)
+		sre, err = syntax.Parse(sNew, regexParseFlags)
 		if err != nil {
 			panic(fmt.Errorf("BUG: cannot parse simplified regexp %q: %w", sNew, err))
 		}
@@ -197,18 +225,18 @@ func simplifyRegexp(sre *syntax.Regexp, hasPrefix bool) *syntax.Regexp {
 	}
 }
 
-func simplifyRegexpExt(sre *syntax.Regexp, hasPrefix, hasSuffix bool) *syntax.Regexp {
+func simplifyRegexpExt(sre *syntax.Regexp, keepBeginOp, keepEndOp bool) *syntax.Regexp {
 	switch sre.Op {
 	case syntax.OpCapture:
 		// Substitute all the capture regexps with non-capture regexps.
 		sre.Op = syntax.OpAlternate
-		sre.Sub[0] = simplifyRegexpExt(sre.Sub[0], hasPrefix, hasSuffix)
+		sre.Sub[0] = simplifyRegexpExt(sre.Sub[0], keepBeginOp, keepEndOp)
 		if sre.Sub[0] == emptyRegexp {
 			return emptyRegexp
 		}
 		return sre
 	case syntax.OpStar, syntax.OpPlus, syntax.OpQuest, syntax.OpRepeat:
-		sre.Sub[0] = simplifyRegexpExt(sre.Sub[0], hasPrefix, hasSuffix)
+		sre.Sub[0] = simplifyRegexpExt(sre.Sub[0], keepBeginOp, keepEndOp)
 		if sre.Sub[0] == emptyRegexp {
 			return emptyRegexp
 		}
@@ -216,13 +244,13 @@ func simplifyRegexpExt(sre *syntax.Regexp, hasPrefix, hasSuffix bool) *syntax.Re
 	case syntax.OpAlternate:
 		// Do not remove empty captures from OpAlternate, since this may break regexp.
 		for i, sub := range sre.Sub {
-			sre.Sub[i] = simplifyRegexpExt(sub, hasPrefix, hasSuffix)
+			sre.Sub[i] = simplifyRegexpExt(sub, keepBeginOp, keepEndOp)
 		}
 		return sre
 	case syntax.OpConcat:
 		subs := sre.Sub[:0]
 		for i, sub := range sre.Sub {
-			sub = simplifyRegexpExt(sub, hasPrefix || len(subs) > 0, hasSuffix || i+1 < len(sre.Sub))
+			sub = simplifyRegexpExt(sub, keepBeginOp || len(subs) > 0, keepEndOp || i+1 < len(sre.Sub))
 			if sub != emptyRegexp {
 				subs = append(subs, sub)
 			}
@@ -230,12 +258,12 @@ func simplifyRegexpExt(sre *syntax.Regexp, hasPrefix, hasSuffix bool) *syntax.Re
 		sre.Sub = subs
 		// Remove anchros from the beginning and the end of regexp, since they
 		// will be added later.
-		if !hasPrefix {
+		if !keepBeginOp {
 			for len(sre.Sub) > 0 && sre.Sub[0].Op == syntax.OpBeginText {
 				sre.Sub = sre.Sub[1:]
 			}
 		}
-		if !hasSuffix {
+		if !keepEndOp {
 			for len(sre.Sub) > 0 && sre.Sub[len(sre.Sub)-1].Op == syntax.OpEndText {
 				sre.Sub = sre.Sub[:len(sre.Sub)-1]
 			}
@@ -254,6 +282,47 @@ func simplifyRegexpExt(sre *syntax.Regexp, hasPrefix, hasSuffix bool) *syntax.Re
 	}
 }
 
+// getSubstringLiteral returns regex part from expr surrounded by .+ or .* depending on the prefixSuffixOp.
+//
+// For example, if expr=".+foo.+" and prefixSuffix=syntax.OpPlus, then the function returns "foo".
+//
+// An empty string is returned if expr doesn't contain the given prefixSuffix prefix and suffix
+// or if the regex part surrounded by prefixSuffix contains alternate regexps.
+func getSubstringLiteral(expr string, prefixSuffixOp syntax.Op) string {
+	// Verify that the expr doesn't contain alternate regexps. In this case it is unsafe removing prefix and suffix.
+	sre, err := syntax.Parse(expr, regexParseFlags)
+	if err != nil {
+		return ""
+	}
+	if sre.Op != syntax.OpConcat {
+		return ""
+	}
+	if len(sre.Sub) != 3 {
+		return ""
+	}
+	if !isDotOp(sre.Sub[0], prefixSuffixOp) || !isDotOp(sre.Sub[2], prefixSuffixOp) || !isLiteral(sre.Sub[1]) {
+		return ""
+	}
+	return string(sre.Sub[1].Rune)
+}
+
+func isDotOpRegexp(expr string, op syntax.Op) bool {
+	sre, err := syntax.Parse(expr, regexParseFlags)
+	if err != nil {
+		return false
+	}
+	return isDotOp(sre, op)
+}
+
+func isDotOp(sre *syntax.Regexp, op syntax.Op) bool {
+	if sre.Op != op {
+		return false
+	}
+	return sre.Sub[0].Op == syntax.OpAnyChar
+}
+
 var emptyRegexp = &syntax.Regexp{
 	Op: syntax.OpEmptyMatch,
 }
+
+const regexParseFlags = syntax.Perl | syntax.DotNL

lib/regexutil/regexutil_test.go

@@ -77,7 +77,7 @@ func TestSimplifyPromRegex(t *testing.T) {
 	f("^foobar|foobaz", "fooba", "[rz]")
 	f("^foobar|^foobaz$", "fooba", "[rz]")
 	f("foobar|foobaz", "fooba", "[rz]")
-	f("(?:^foobar|^foobaz)aa.*", "fooba", "(?-s:[rz]aa.*)")
+	f("(?:^foobar|^foobaz)aa.*", "fooba", "(?s:[rz]aa.*)")
 	f("foo[bar]+", "foo", "[abr]+")
 	f("foo[a-z]+", "foo", "[a-z]+")
 	f("foo[bar]*", "foo", "[abr]*")
@@ -88,12 +88,12 @@ func TestSimplifyPromRegex(t *testing.T) {
 	f("foo[^x]*", "foo", "[^x]*")
 	f("foo[x]*bar", "foo", "x*bar")
 	f("fo\\Bo[x]*bar?", "fo", "\\Box*bar?")
-	f("foo.+bar", "foo", "(?-s:.+bar)")
-	f("a(b|c.*).+", "a", "(?-s:(?:b|c.*).+)")
+	f("foo.+bar", "foo", "(?s:.+bar)")
+	f("a(b|c.*).+", "a", "(?s:(?:b|c.*).+)")
 	f("ab|ac", "a", "[bc]")
 	f("(?i)xyz", "", "(?i:XYZ)")
 	f("(?i)foo|bar", "", "(?i:FOO|BAR)")
-	f("(?i)up.+x", "", "(?i-s:UP.+X)")
+	f("(?i)up.+x", "", "(?is:UP.+X)")
 	f("(?smi)xy.*z$", "", "(?ims:XY.*Z$)")
 
 	// test invalid regexps
@@ -111,12 +111,12 @@ func TestSimplifyPromRegex(t *testing.T) {
 	f("(foo|bar$)x*", "", "(?-m:(?:foo|bar$)x*)")
 
 	// See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/5297
-	f(".+;|;.+", "", "(?-s:.+;|;.+)")
-	f("^(.+);|;(.+)$", "", "(?-s:.+;|;.+)")
-	f("^(.+);$|^;(.+)$", "", "(?-s:.+;|;.+)")
-	f(".*;|;.*", "", "(?-s:.*;|;.*)")
-	f("^(.*);|;(.*)$", "", "(?-s:.*;|;.*)")
-	f("^(.*);$|^;(.*)$", "", "(?-s:.*;|;.*)")
+	f(".+;|;.+", "", "(?s:.+;|;.+)")
+	f("^(.+);|;(.+)$", "", "(?s:.+;|;.+)")
+	f("^(.+);$|^;(.+)$", "", "(?s:.+;|;.+)")
+	f(".*;|;.*", "", "(?s:.*;|;.*)")
+	f("^(.*);|;(.*)$", "", "(?s:.*;|;.*)")
+	f("^(.*);$|^;(.*)$", "", "(?s:.*;|;.*)")
 }
 
 func TestRemoveStartEndAnchors(t *testing.T) {