Optimize .+ regex line filter (#8646)

Interpolated variables from Grafana dashboards may result in the regex
matcher `.+`. Filters with `+` are already being rewritten to be non
greedy but there is a further optimization to make replace the regex
altogether with a filter that just matchers `len(line) > 0`

pkg/logql/log/filter.go

@@ -33,6 +33,20 @@ func (trueFilter) ToStage() Stage       { return NoopStage }
 // TrueFilter is a filter that returns and matches all log lines whatever their content.
 var TrueFilter = trueFilter{}
 
+type existsFilter struct{}
+
+func (existsFilter) Filter(_ []byte) bool { return true }
+func (existsFilter) ToStage() Stage {
+	return StageFunc{
+		process: func(_ int64, line []byte, _ *LabelsBuilder) ([]byte, bool) {
+			return line, len(line) > 0
+		},
+	}
+}
+
+// ExistsFilter is a filter that returns and matches when a line has any characters.
+var ExistsFilter = existsFilter{}
+
 type notFilter struct {
 	Filterer
 }
@@ -425,6 +439,10 @@ func simplify(reg *syntax.Regexp) (Filterer, bool) {
 		if reg.Sub[0].Op == syntax.OpAnyCharNotNL {
 			return TrueFilter, true
 		}
+	case syntax.OpPlus:
+		if len(reg.Sub) == 1 && reg.Sub[0].Op == syntax.OpAnyCharNotNL { // simplify ".+"
+			return ExistsFilter, true
+		}
 	case syntax.OpEmptyMatch:
 		return TrueFilter, true
 	}

pkg/logql/log/filter_test.go

@@ -61,6 +61,7 @@ func Test_SimplifiedRegex(t *testing.T) {
 		{"(?i)f|foo|foobar", true, newOrFilter(newContainsFilter([]byte("F"), true), newOrFilter(newContainsFilter([]byte("FOO"), true), newContainsFilter([]byte("FOOBAR"), true))), true},
 		{"(?i)f|fatal|e.*", true, newOrFilter(newOrFilter(newContainsFilter([]byte("F"), true), newContainsFilter([]byte("FATAL"), true)), newContainsFilter([]byte("E"), true)), true},
 		{"(?i).*foo.*", true, newContainsFilter([]byte("FOO"), true), true},
+		{".+", true, ExistsFilter, true},
 
 		// regex we are not supporting.
 		{"[a-z]+foo", true, nil, false},
@@ -77,6 +78,7 @@ func Test_SimplifiedRegex(t *testing.T) {
 		{`(\w\d+)`, true, nil, false},
 		{`.*f.*oo|fo{1,2}`, true, nil, false},
 		{"f|f(?i)oo", true, nil, false},
+		{".foo+", true, nil, false},
 	} {
 		t.Run(test.re, func(t *testing.T) {
 			d, err := newRegexpFilter(test.re, test.match)