Remove empty match within concat of regex (#6026)

* Remove empty match within concat of regex

* add back tests

pkg/logql/log/filter.go

@@ -471,6 +471,16 @@ func simplifyAlternate(reg *syntax.Regexp) (Filterer, bool) {
 // Anything else is rejected.
 func simplifyConcat(reg *syntax.Regexp, baseLiteral []byte) (Filterer, bool) {
 	clearCapture(reg.Sub...)
+	// remove empty match as we don't need them for filtering
+	i := 0
+	for _, r := range reg.Sub {
+		if r.Op == syntax.OpEmptyMatch {
+			continue
+		}
+		reg.Sub[i] = r
+		i++
+	}
+	reg.Sub = reg.Sub[:i]
 	// we support only simplication of concat operation with 3 sub expressions.
 	// for instance .*foo.*bar contains 4 subs (.*+foo+.*+bar) and can't be simplified.
 	if len(reg.Sub) > 3 {

pkg/logql/log/filter_test.go

@@ -56,6 +56,7 @@ func Test_SimplifiedRegex(t *testing.T) {
 		{"(?i)foo", true, newContainsFilter([]byte("foo"), true), true},
 		{"(?i)界", true, newContainsFilter([]byte("界"), true), true},
 		{"(?i)ïB", true, newContainsFilter([]byte("ïB"), true), true},
+		{"(?:)foo|fatal|exception", true, newOrFilter(newOrFilter(newContainsFilter([]byte("foo"), false), newContainsFilter([]byte("fatal"), false)), newContainsFilter([]byte("exception"), false)), true},
 
 		// regex we are not supporting.
 		{"[a-z]+foo", true, nil, false},
@@ -160,6 +161,7 @@ func Benchmark_LineFilter(b *testing.B) {
 		{"(node:24) buzz*"},
 		{"(HTTP/.*\\\"|HEAD|GET) (2..|5..)"},
 		{"\"@l\":\"(Warning|Error|Fatal)\""},
+		{"(?:)foo|fatal|exception"},
 	} {
 		benchmarkRegex(b, test.re, logline, true)
 		benchmarkRegex(b, test.re, logline, false)