diff --git a/cmd/webhook/admission/admission_test.go b/cmd/webhook/admission/admission_test.go
index 424fb0b9a7..84b9879a56 100644
--- a/cmd/webhook/admission/admission_test.go
+++ b/cmd/webhook/admission/admission_test.go
@@ -107,7 +107,7 @@ func TestRouteGroupAdmitter(t *testing.T) {
 		{
 			name:      "invalid eskip filters",
 			inputFile: "rg-with-invalid-eskip-filters.json",
-			message:   "parse failed after token status, position 11: syntax error",
+			message:   "parse failed after token status, position 6: syntax error",
 		},
 		{
 			name:      "valid eskip predicates",
@@ -121,7 +121,7 @@ func TestRouteGroupAdmitter(t *testing.T) {
 		{
 			name:      "invalid eskip filters and predicates",
 			inputFile: "rg-with-invalid-eskip-filters-and-predicates.json",
-			message:   "parse failed after token status, position 11: syntax error\\nparse failed after token Method, position 6: syntax error",
+			message:   "parse failed after token status, position 6: syntax error\\nparse failed after token Method, position 6: syntax error",
 		},
 		{
 			name:      "invalid routgroup multiple filters per json/yaml array item",
@@ -185,12 +185,12 @@ func TestIngressAdmitter(t *testing.T) {
 		{
 			name:      "invalid eskip filters",
 			inputFile: "invalid-filters.json",
-			message:   `invalid \"zalando.org/skipper-filter\" annotation: parse failed after token this, position 9: syntax error`,
+			message:   `invalid \"zalando.org/skipper-filter\" annotation: parse failed after token this, position 4: syntax error`,
 		},
 		{
 			name:      "invalid eskip predicates",
 			inputFile: "invalid-predicates.json",
-			message:   `invalid \"zalando.org/skipper-predicate\" annotation: parse failed after token ), position 15: syntax error`,
+			message:   `invalid \"zalando.org/skipper-predicate\" annotation: parse failed after token ), position 15: unexpected token`,
 		},
 		{
 			name:      "invalid eskip routes",
@@ -200,7 +200,7 @@ func TestIngressAdmitter(t *testing.T) {
 		{
 			name:      "invalid eskip filters and predicates",
 			inputFile: "invalid-filters-and-predicates.json",
-			message:   `invalid \"zalando.org/skipper-filter\" annotation: parse failed after token this, position 9: syntax error\ninvalid \"zalando.org/skipper-predicate\" annotation: parse failed after token ), position 15: syntax error`,
+			message:   `invalid \"zalando.org/skipper-filter\" annotation: parse failed after token this, position 4: syntax error\ninvalid \"zalando.org/skipper-predicate\" annotation: parse failed after token ), position 15: unexpected token`,
 		},
 	} {
 		t.Run(tc.name, func(t *testing.T) {
diff --git a/dataclients/kubernetes/testdata/routegroups/convert/failing-filter.log b/dataclients/kubernetes/testdata/routegroups/convert/failing-filter.log
index b90ceb446a..ba158256c3 100644
--- a/dataclients/kubernetes/testdata/routegroups/convert/failing-filter.log
+++ b/dataclients/kubernetes/testdata/routegroups/convert/failing-filter.log
@@ -1 +1 @@
-\[routegroup\] parse failed after token foo, position 8: syntax error
+\[routegroup\] parse failed after token foo, position 3: syntax error
diff --git a/dataclients/kubernetes/testdata/routegroups/convert/no-catchall-for-failed-route-group.log b/dataclients/kubernetes/testdata/routegroups/convert/no-catchall-for-failed-route-group.log
index b90ceb446a..ba158256c3 100644
--- a/dataclients/kubernetes/testdata/routegroups/convert/no-catchall-for-failed-route-group.log
+++ b/dataclients/kubernetes/testdata/routegroups/convert/no-catchall-for-failed-route-group.log
@@ -1 +1 @@
-\[routegroup\] parse failed after token foo, position 8: syntax error
+\[routegroup\] parse failed after token foo, position 3: syntax error
diff --git a/eskip/eskip.go b/eskip/eskip.go
index 90cad17401..e68ab5d22c 100644
--- a/eskip/eskip.go
+++ b/eskip/eskip.go
@@ -588,8 +588,22 @@ type eskipLexParser struct {
 
 var parserPool sync.Pool
 
-// executes the parser.
-func parse(code string) ([]*parsedRoute, error) {
+func parseDocument(code string) ([]*parsedRoute, error) {
+	routes, _, _, err := parse(start_document, code)
+	return routes, err
+}
+
+func parsePredicates(code string) ([]*Predicate, error) {
+	_, predicates, _, err := parse(start_predicates, code)
+	return predicates, err
+}
+
+func parseFilters(code string) ([]*Filter, error) {
+	_, _, filters, err := parse(start_filters, code)
+	return filters, err
+}
+
+func parse(start int, code string) ([]*parsedRoute, []*Predicate, []*Filter, error) {
 	lp, ok := parserPool.Get().(*eskipLexParser)
 	if !ok {
 		lp = &eskipLexParser{}
@@ -598,16 +612,19 @@ func parse(code string) ([]*parsedRoute, error) {
 	}
 	defer parserPool.Put(lp)
 
-	lp.lexer.init(code)
+	lexer := &lp.lexer
+	lexer.init(start, code)
 
-	lp.parser.Parse(&lp.lexer)
+	lp.parser.Parse(lexer)
 
-	return lp.lexer.routes, lp.lexer.err
+	// Do not return lexer to avoid reading lexer fields after returning eskipLexParser to the pool.
+	// Let the caller decide which of return values to use based on the start token.
+	return lexer.routes, lexer.predicates, lexer.filters, lexer.err
 }
 
 // Parses a route expression or a routing document to a set of route definitions.
 func Parse(code string) ([]*Route, error) {
-	parsedRoutes, err := parse(code)
+	parsedRoutes, err := parseDocument(code)
 	if err != nil {
 		return nil, err
 	}
@@ -662,12 +679,7 @@ func ParseFilters(f string) ([]*Filter, error) {
 		return nil, nil
 	}
 
-	rs, err := parse("* -> " + f + " -> <shunt>")
-	if err != nil {
-		return nil, err
-	}
-
-	return rs[0].filters, nil
+	return parseFilters(f)
 }
 
 // ParsePredicates parses a set of predicates (combined by '&&') into
@@ -678,7 +690,7 @@ func ParsePredicates(p string) ([]*Predicate, error) {
 		return nil, nil
 	}
 
-	rs, err := parse(p + " -> <shunt>")
+	rs, err := parsePredicates(p)
 	if err != nil {
 		return nil, err
 	}
@@ -687,9 +699,8 @@ func ParsePredicates(p string) ([]*Predicate, error) {
 		return nil, nil
 	}
 
-	r := rs[0]
-	ps := make([]*Predicate, 0, len(r.predicates))
-	for _, p := range r.predicates {
+	ps := make([]*Predicate, 0, len(rs))
+	for _, p := range rs {
 		if p.Name != "*" {
 			ps = append(ps, p)
 		}
diff --git a/eskip/eskip_test.go b/eskip/eskip_test.go
index 4a5f72c93e..bbc832dec0 100644
--- a/eskip/eskip_test.go
+++ b/eskip/eskip_test.go
@@ -239,19 +239,39 @@ func TestParseFilters(t *testing.T) {
 		"error",
 		"trallala",
 		nil,
-		// TODO: fix position
-		"parse failed after token ->, position 16: syntax error",
+		"parse failed after token trallala, position 8: syntax error",
 	}, {
 		"error 2",
 		"foo-bar",
 		nil,
-		// TODO: fix position
-		"parse failed after token foo, position 8: syntax error",
+		"parse failed after token foo, position 3: syntax error",
 	}, {
 		"success",
 		`filter1(3.14) -> filter2("key", 42)`,
 		[]*Filter{{Name: "filter1", Args: []interface{}{3.14}}, {Name: "filter2", Args: []interface{}{"key", float64(42)}}},
 		"",
+	}, {
+		"a comment produces empty filters without error",
+		"// a comment",
+		nil,
+		"",
+	}, {
+		"a trailing comment is ignored",
+		`filter1(3.14) -> filter2("key", 42) // a trailing comment`,
+		[]*Filter{{Name: "filter1", Args: []interface{}{3.14}}, {Name: "filter2", Args: []interface{}{"key", float64(42)}}},
+		"",
+	}, {
+		"a comment before is ignored",
+		`// a comment on a separate line
+		filter1(3.14) -> filter2("key", 42)`,
+		[]*Filter{{Name: "filter1", Args: []interface{}{3.14}}, {Name: "filter2", Args: []interface{}{"key", float64(42)}}},
+		"",
+	}, {
+		"a comment after is ignored",
+		`filter1(3.14) -> filter2("key", 42)
+		// a comment on a separate line`,
+		[]*Filter{{Name: "filter1", Args: []interface{}{3.14}}, {Name: "filter2", Args: []interface{}{"key", float64(42)}}},
+		"",
 	}} {
 		t.Run(ti.msg, func(t *testing.T) {
 			fs, err := ParseFilters(ti.expression)
@@ -306,6 +326,29 @@ func TestParsePredicates(t *testing.T) {
 		title:    "comment fuzz 2", // "\x2f\x2f..." == "//..."
 		input:    "\x2f\x2f\x00\x00\x00\xe6\xfe\x00\x00\x2f\x00\x00\x00\x00\x00\x00\x00\xe6\xfe\x00\x00\x2f\x00\x00\x00\x00",
 		expected: nil,
+	}, {
+		title: "a trailing comment is ignored",
+		input: `Foo("bar") && Baz("qux") // a trailing comment`,
+		expected: []*Predicate{
+			{Name: "Foo", Args: []interface{}{"bar"}},
+			{Name: "Baz", Args: []interface{}{"qux"}},
+		},
+	}, {
+		title: "a comment before is ignored",
+		input: `// a comment on a separate line
+		Foo("bar") && Baz("qux")`,
+		expected: []*Predicate{
+			{Name: "Foo", Args: []interface{}{"bar"}},
+			{Name: "Baz", Args: []interface{}{"qux"}},
+		},
+	}, {
+		title: "a comment after is ignored",
+		input: `Foo("bar") && Baz("qux")
+		// a comment on a separate line`,
+		expected: []*Predicate{
+			{Name: "Foo", Args: []interface{}{"bar"}},
+			{Name: "Baz", Args: []interface{}{"qux"}},
+		},
 	}} {
 		t.Run(test.title, func(t *testing.T) {
 			ps, err := ParsePredicates(test.input)
diff --git a/eskip/lexer.go b/eskip/lexer.go
index c49942f190..875aafd806 100644
--- a/eskip/lexer.go
+++ b/eskip/lexer.go
@@ -14,12 +14,15 @@ type token struct {
 type charPredicate func(byte) bool
 
 type eskipLex struct {
+	start         int
 	code          string
 	lastToken     string
 	lastRouteID   string
 	err           error
 	initialLength int
 	routes        []*parsedRoute
+	predicates    []*Predicate
+	filters       []*Filter
 }
 
 type fixedScanner token
@@ -62,7 +65,8 @@ func (fs *fixedScanner) scan(code string) (t token, rest string, err error) {
 	return token(*fs), code[len(fs.val):], nil
 }
 
-func (l *eskipLex) init(code string) {
+func (l *eskipLex) init(start int, code string) {
+	l.start = start
 	l.code = code
 	l.initialLength = len(code)
 }
@@ -352,6 +356,13 @@ func (l *eskipLex) next() (token, error) {
 }
 
 func (l *eskipLex) Lex(lval *eskipSymType) int {
+	// first emit the start token
+	if l.start != 0 {
+		start := l.start
+		l.start = 0
+		return start
+	}
+
 	t, err := l.next()
 	if err == eof {
 		return -1
diff --git a/eskip/parser.go b/eskip/parser.go
index 569ff7f323..52de78b4a5 100644
--- a/eskip/parser.go
+++ b/eskip/parser.go
@@ -31,8 +31,6 @@ type eskipSymType struct {
 	dynamic     bool
 	lbBackend   bool
 	numval      float64
-	stringval   string
-	regexpval   string
 	stringvals  []string
 	lbAlgorithm string
 	lbEndpoints []string
@@ -55,6 +53,9 @@ const stringliteral = 57359
 const symbol = 57360
 const openarrow = 57361
 const closearrow = 57362
+const start_document = 57363
+const start_predicates = 57364
+const start_filters = 57365
 
 var eskipToknames = [...]string{
 	"$end",
@@ -77,6 +78,9 @@ var eskipToknames = [...]string{
 	"symbol",
 	"openarrow",
 	"closearrow",
+	"start_document",
+	"start_predicates",
+	"start_filters",
 }
 
 var eskipStatenames = [...]string{}
@@ -93,62 +97,65 @@ var eskipExca = [...]int8{
 
 const eskipPrivate = 57344
 
-const eskipLast = 64
+const eskipLast = 68
 
 var eskipAct = [...]int8{
-	34, 40, 32, 31, 26, 19, 22, 23, 24, 27,
-	29, 28, 21, 48, 36, 9, 37, 27, 41, 9,
-	18, 27, 27, 7, 10, 49, 43, 15, 14, 42,
-	16, 16, 8, 55, 4, 45, 30, 50, 44, 21,
-	45, 15, 3, 47, 46, 17, 38, 51, 52, 13,
-	53, 42, 54, 12, 25, 11, 39, 35, 33, 20,
-	5, 6, 2, 1,
+	48, 39, 17, 29, 38, 2, 3, 4, 32, 33,
+	34, 31, 18, 36, 11, 55, 16, 43, 8, 42,
+	50, 49, 13, 18, 41, 50, 13, 28, 44, 59,
+	23, 45, 19, 24, 26, 15, 37, 30, 27, 12,
+	24, 7, 53, 51, 52, 52, 56, 57, 23, 44,
+	54, 21, 22, 20, 58, 46, 25, 21, 60, 9,
+	35, 47, 40, 14, 10, 6, 5, 1,
 }
 
 var eskipPact = [...]int16{
-	14, -1000, 11, -1000, -1000, 49, 10, -1000, 19, -1000,
-	2, -8, 10, -1000, 20, -1000, 4, -1000, 33, -1000,
-	40, -1000, -1000, -1000, -1000, -1000, -1000, -1000, 0, 15,
-	-1000, 31, -1000, -1000, -1000, -1000, -1000, -1000, -8, -7,
-	16, 28, -1000, 4, -1000, 4, -1000, -1000, -1000, 5,
-	5, 26, -1000, -1000, 16, -1000,
+	-16, -1000, 21, 17, 5, -1000, 19, -1000, -1000, 47,
+	17, -1000, 22, -1000, 53, 29, 50, -1000, 23, 9,
+	-6, 17, -1000, -1000, 7, 5, 7, -1000, 40, -1000,
+	49, -1000, -1000, -1000, -1000, -1000, 3, -1000, 36, -1000,
+	-1000, -1000, -1000, -1000, -1000, 35, -6, -5, 37, 38,
+	-1000, -1000, 7, -1000, -1000, -1000, 12, 8, -1000, -1000,
+	37,
 }
 
 var eskipPgo = [...]int8{
-	0, 63, 62, 42, 34, 61, 60, 5, 59, 23,
-	3, 4, 2, 58, 0, 57, 1, 56, 54,
+	0, 67, 66, 59, 16, 65, 41, 18, 64, 3,
+	14, 4, 2, 1, 62, 0, 61, 60,
 }
 
 var eskipR1 = [...]int8{
-	0, 1, 1, 2, 2, 2, 2, 4, 5, 3,
-	3, 6, 6, 9, 9, 8, 8, 11, 10, 10,
-	10, 12, 12, 12, 16, 16, 17, 17, 18, 7,
-	7, 7, 7, 7, 13, 14, 15,
+	0, 1, 1, 1, 1, 1, 2, 2, 5, 5,
+	5, 5, 7, 8, 6, 6, 3, 3, 10, 10,
+	4, 4, 12, 11, 11, 11, 13, 13, 13, 15,
+	15, 16, 16, 17, 9, 9, 9, 9, 9, 14,
 }
 
 var eskipR2 = [...]int8{
-	0, 1, 1, 0, 1, 3, 2, 2, 2, 3,
-	5, 1, 3, 1, 4, 1, 3, 4, 0, 1,
-	3, 1, 1, 1, 1, 3, 1, 3, 3, 1,
-	1, 1, 1, 1, 1, 1, 1,
+	0, 2, 1, 2, 1, 2, 1, 1, 0, 1,
+	3, 2, 2, 2, 3, 5, 1, 3, 1, 4,
+	1, 3, 4, 0, 1, 3, 1, 1, 1, 1,
+	3, 1, 3, 3, 1, 1, 1, 1, 1, 1,
 }
 
 var eskipChk = [...]int16{
-	-1000, -1, -2, -3, -4, -6, -5, -9, 18, 5,
-	13, 6, 4, -3, 18, 8, 11, -4, 18, -7,
-	-8, -14, 14, 15, 16, -18, -11, 17, 19, 18,
-	-9, -10, -12, -13, -14, -15, 10, 12, 6, -17,
-	-16, 18, -14, 11, 7, 9, -7, -11, 20, 9,
-	9, -10, -12, -14, -16, 7,
+	-1000, -1, 21, 22, 23, -2, -5, -6, -7, -3,
+	-8, -10, 18, 5, -3, 18, -4, -12, 18, 13,
+	6, 4, -6, 8, 11, 6, 11, -7, 18, -9,
+	-4, 17, 14, 15, 16, -17, 19, -10, -11, -13,
+	-14, 17, 12, 10, -12, -11, 6, -16, -15, 18,
+	17, 7, 9, 7, -9, 20, 9, 9, -13, 17,
+	-15,
 }
 
 var eskipDef = [...]int8{
-	3, -2, 1, 2, 4, 0, 0, 11, 0, 13,
-	6, 0, 0, 7, 0, 8, 18, 5, 0, 9,
-	0, 29, 30, 31, 32, 33, 15, 35, 0, 0,
-	12, 0, 19, 21, 22, 23, 34, 36, 0, 0,
-	26, 0, 24, 18, 14, 0, 10, 16, 28, 0,
-	0, 0, 20, 25, 27, 17,
+	0, -2, 8, 2, 4, 1, 6, 7, 9, 0,
+	0, 16, 0, 18, 3, 0, 5, 20, 0, 11,
+	0, 0, 12, 13, 23, 0, 23, 10, 0, 14,
+	0, 34, 35, 36, 37, 38, 0, 17, 0, 24,
+	26, 27, 28, 39, 21, 0, 0, 0, 31, 0,
+	29, 19, 0, 22, 15, 33, 0, 0, 25, 30,
+	32,
 }
 
 var eskipTok1 = [...]int8{
@@ -157,7 +164,8 @@ var eskipTok1 = [...]int8{
 
 var eskipTok2 = [...]int8{
 	2, 3, 4, 5, 6, 7, 8, 9, 10, 11,
-	12, 13, 14, 15, 16, 17, 18, 19, 20,
+	12, 13, 14, 15, 16, 17, 18, 19, 20, 21,
+	22, 23,
 }
 
 var eskipTok3 = [...]int8{
@@ -500,47 +508,72 @@ eskipdefault:
 	switch eskipnt {
 
 	case 1:
+		eskipDollar = eskipS[eskippt-2 : eskippt+1]
+		{
+			eskiplex.(*eskipLex).routes = eskipDollar[2].routes
+		}
+	case 2:
+		eskipDollar = eskipS[eskippt-1 : eskippt+1]
+		{
+			// allow empty or comments only
+			eskiplex.(*eskipLex).predicates = nil
+		}
+	case 3:
+		eskipDollar = eskipS[eskippt-2 : eskippt+1]
+		{
+			eskiplex.(*eskipLex).predicates = eskipDollar[2].predicates
+		}
+	case 4:
+		eskipDollar = eskipS[eskippt-1 : eskippt+1]
+		{
+			// allow empty or comments only
+			eskiplex.(*eskipLex).filters = nil
+		}
+	case 5:
+		eskipDollar = eskipS[eskippt-2 : eskippt+1]
+		{
+			eskiplex.(*eskipLex).filters = eskipDollar[2].filters
+		}
+	case 6:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
 			eskipVAL.routes = eskipDollar[1].routes
-			eskiplex.(*eskipLex).routes = eskipVAL.routes
 		}
-	case 2:
+	case 7:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
 			eskipVAL.routes = []*parsedRoute{eskipDollar[1].route}
-			eskiplex.(*eskipLex).routes = eskipVAL.routes
 		}
-	case 4:
+	case 9:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
 			eskipVAL.routes = []*parsedRoute{eskipDollar[1].route}
 		}
-	case 5:
+	case 10:
 		eskipDollar = eskipS[eskippt-3 : eskippt+1]
 		{
 			eskipVAL.routes = eskipDollar[1].routes
 			eskipVAL.routes = append(eskipVAL.routes, eskipDollar[3].route)
 		}
-	case 6:
+	case 11:
 		eskipDollar = eskipS[eskippt-2 : eskippt+1]
 		{
 			eskipVAL.routes = eskipDollar[1].routes
 		}
-	case 7:
+	case 12:
 		eskipDollar = eskipS[eskippt-2 : eskippt+1]
 		{
 			eskipVAL.route = eskipDollar[2].route
 			eskipVAL.route.id = eskipDollar[1].token
 		}
-	case 8:
+	case 13:
 		eskipDollar = eskipS[eskippt-2 : eskippt+1]
 		{
 			// match symbol and colon to get route id early even if route parsing fails later
 			eskipVAL.token = eskipDollar[1].token
 			eskiplex.(*eskipLex).lastRouteID = eskipDollar[1].token
 		}
-	case 9:
+	case 14:
 		eskipDollar = eskipS[eskippt-3 : eskippt+1]
 		{
 			eskipVAL.route = &parsedRoute{
@@ -556,7 +589,7 @@ eskipdefault:
 			eskipDollar[1].predicates = nil
 			eskipDollar[3].lbEndpoints = nil
 		}
-	case 10:
+	case 15:
 		eskipDollar = eskipS[eskippt-5 : eskippt+1]
 		{
 			eskipVAL.route = &parsedRoute{
@@ -574,40 +607,40 @@ eskipdefault:
 			eskipDollar[3].filters = nil
 			eskipDollar[5].lbEndpoints = nil
 		}
-	case 11:
+	case 16:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
 			eskipVAL.predicates = []*Predicate{eskipDollar[1].predicate}
 		}
-	case 12:
+	case 17:
 		eskipDollar = eskipS[eskippt-3 : eskippt+1]
 		{
 			eskipVAL.predicates = eskipDollar[1].predicates
 			eskipVAL.predicates = append(eskipVAL.predicates, eskipDollar[3].predicate)
 		}
-	case 13:
+	case 18:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
 			eskipVAL.predicate = &Predicate{"*", nil}
 		}
-	case 14:
+	case 19:
 		eskipDollar = eskipS[eskippt-4 : eskippt+1]
 		{
 			eskipVAL.predicate = &Predicate{eskipDollar[1].token, eskipDollar[3].args}
 			eskipDollar[3].args = nil
 		}
-	case 15:
+	case 20:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
 			eskipVAL.filters = []*Filter{eskipDollar[1].filter}
 		}
-	case 16:
+	case 21:
 		eskipDollar = eskipS[eskippt-3 : eskippt+1]
 		{
 			eskipVAL.filters = eskipDollar[1].filters
 			eskipVAL.filters = append(eskipVAL.filters, eskipDollar[3].filter)
 		}
-	case 17:
+	case 22:
 		eskipDollar = eskipS[eskippt-4 : eskippt+1]
 		{
 			eskipVAL.filter = &Filter{
@@ -615,70 +648,70 @@ eskipdefault:
 				Args: eskipDollar[3].args}
 			eskipDollar[3].args = nil
 		}
-	case 19:
+	case 24:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
 			eskipVAL.args = []interface{}{eskipDollar[1].arg}
 		}
-	case 20:
+	case 25:
 		eskipDollar = eskipS[eskippt-3 : eskippt+1]
 		{
 			eskipVAL.args = eskipDollar[1].args
 			eskipVAL.args = append(eskipVAL.args, eskipDollar[3].arg)
 		}
-	case 21:
+	case 26:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
 			eskipVAL.arg = eskipDollar[1].numval
 		}
-	case 22:
+	case 27:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
-			eskipVAL.arg = eskipDollar[1].stringval
+			eskipVAL.arg = eskipDollar[1].token
 		}
-	case 23:
+	case 28:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
-			eskipVAL.arg = eskipDollar[1].regexpval
+			eskipVAL.arg = eskipDollar[1].token
 		}
-	case 24:
+	case 29:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
-			eskipVAL.stringvals = []string{eskipDollar[1].stringval}
+			eskipVAL.stringvals = []string{eskipDollar[1].token}
 		}
-	case 25:
+	case 30:
 		eskipDollar = eskipS[eskippt-3 : eskippt+1]
 		{
 			eskipVAL.stringvals = eskipDollar[1].stringvals
-			eskipVAL.stringvals = append(eskipVAL.stringvals, eskipDollar[3].stringval)
+			eskipVAL.stringvals = append(eskipVAL.stringvals, eskipDollar[3].token)
 		}
-	case 26:
+	case 31:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
 			eskipVAL.lbEndpoints = eskipDollar[1].stringvals
 		}
-	case 27:
+	case 32:
 		eskipDollar = eskipS[eskippt-3 : eskippt+1]
 		{
 			eskipVAL.lbAlgorithm = eskipDollar[1].token
 			eskipVAL.lbEndpoints = eskipDollar[3].stringvals
 		}
-	case 28:
+	case 33:
 		eskipDollar = eskipS[eskippt-3 : eskippt+1]
 		{
 			eskipVAL.lbAlgorithm = eskipDollar[2].lbAlgorithm
 			eskipVAL.lbEndpoints = eskipDollar[2].lbEndpoints
 		}
-	case 29:
+	case 34:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
-			eskipVAL.backend = eskipDollar[1].stringval
+			eskipVAL.backend = eskipDollar[1].token
 			eskipVAL.shunt = false
 			eskipVAL.loopback = false
 			eskipVAL.dynamic = false
 			eskipVAL.lbBackend = false
 		}
-	case 30:
+	case 35:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
 			eskipVAL.shunt = true
@@ -686,7 +719,7 @@ eskipdefault:
 			eskipVAL.dynamic = false
 			eskipVAL.lbBackend = false
 		}
-	case 31:
+	case 36:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
 			eskipVAL.shunt = false
@@ -694,7 +727,7 @@ eskipdefault:
 			eskipVAL.dynamic = false
 			eskipVAL.lbBackend = false
 		}
-	case 32:
+	case 37:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
 			eskipVAL.shunt = false
@@ -702,7 +735,7 @@ eskipdefault:
 			eskipVAL.dynamic = true
 			eskipVAL.lbBackend = false
 		}
-	case 33:
+	case 38:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
 			eskipVAL.shunt = false
@@ -712,21 +745,11 @@ eskipdefault:
 			eskipVAL.lbAlgorithm = eskipDollar[1].lbAlgorithm
 			eskipVAL.lbEndpoints = eskipDollar[1].lbEndpoints
 		}
-	case 34:
+	case 39:
 		eskipDollar = eskipS[eskippt-1 : eskippt+1]
 		{
 			eskipVAL.numval = convertNumber(eskipDollar[1].token)
 		}
-	case 35:
-		eskipDollar = eskipS[eskippt-1 : eskippt+1]
-		{
-			eskipVAL.stringval = eskipDollar[1].token
-		}
-	case 36:
-		eskipDollar = eskipS[eskippt-1 : eskippt+1]
-		{
-			eskipVAL.regexpval = eskipDollar[1].token
-		}
 	}
 	goto eskipstack /* stack new state and value */
 }
diff --git a/eskip/parser.y b/eskip/parser.y
index 86a7048d23..5378360834 100644
--- a/eskip/parser.y
+++ b/eskip/parser.y
@@ -44,8 +44,6 @@ func convertNumber(s string) float64 {
 	dynamic bool
 	lbBackend bool
 	numval float64
-	stringval string
-	regexpval string
 	stringvals []string
 	lbAlgorithm string
 	lbEndpoints []string
@@ -69,17 +67,42 @@ func convertNumber(s string) float64 {
 %token openarrow
 %token closearrow
 
+%token start_document;
+%token start_predicates;
+%token start_filters;
+
 %%
 
+start:
+	start_document document {
+		eskiplex.(*eskipLex).routes = $2.routes
+	}
+	|
+	start_predicates {
+		// allow empty or comments only
+		eskiplex.(*eskipLex).predicates = nil
+	}
+	|
+	start_predicates predicates {
+		eskiplex.(*eskipLex).predicates = $2.predicates
+	}
+	|
+	start_filters {
+		// allow empty or comments only
+		eskiplex.(*eskipLex).filters = nil
+	}
+	|
+	start_filters filters {
+		eskiplex.(*eskipLex).filters = $2.filters
+	}
+
 document:
 	routes {
 		$$.routes = $1.routes
-		eskiplex.(*eskipLex).routes = $$.routes
 	}
 	|
 	route {
 		$$.routes = []*parsedRoute{$1.route}
-		eskiplex.(*eskipLex).routes = $$.routes
 	}
 
 routes:
@@ -197,22 +220,22 @@ arg:
 		$$.arg = $1.numval
 	}
 	|
-	stringval {
-		$$.arg = $1.stringval
+	stringliteral {
+		$$.arg = $1.token
 	}
 	|
-	regexpval {
-		$$.arg = $1.regexpval
+	regexpliteral {
+		$$.arg = $1.token
 	}
 
 stringvals:
-	stringval {
-		$$.stringvals = []string{$1.stringval}
+	stringliteral {
+		$$.stringvals = []string{$1.token}
 	}
 	|
-	stringvals comma stringval {
+	stringvals comma stringliteral {
 		$$.stringvals = $1.stringvals
-		$$.stringvals = append($$.stringvals, $3.stringval)
+		$$.stringvals = append($$.stringvals, $3.token)
 	}
 
 lbbackendbody:
@@ -232,8 +255,8 @@ lbbackend:
 	}
 
 backend:
-	stringval {
-		$$.backend = $1.stringval
+	stringliteral {
+		$$.backend = $1.token
 		$$.shunt = false
 		$$.loopback = false
 		$$.dynamic = false
@@ -275,14 +298,4 @@ numval:
 		$$.numval = convertNumber($1.token)
 	}
 
-stringval:
-	stringliteral {
-		$$.stringval = $1.token
-	}
-
-regexpval:
-	regexpliteral {
-		$$.regexpval = $1.token
-	}
-
 %%
diff --git a/eskip/parser_test.go b/eskip/parser_test.go
index a48085037e..ffe226d0ef 100644
--- a/eskip/parser_test.go
+++ b/eskip/parser_test.go
@@ -82,14 +82,14 @@ func checkSingleRouteExample(r *parsedRoute, t *testing.T) {
 }
 
 func TestReturnsLexerErrors(t *testing.T) {
-	_, err := parse("invalid code")
+	_, err := parseDocument("invalid code")
 	if err == nil {
 		t.Error("failed to fail")
 	}
 }
 
 func TestParseSingleRoute(t *testing.T) {
-	r, err := parse(singleRouteExample)
+	r, err := parseDocument(singleRouteExample)
 
 	if err != nil {
 		t.Error("failed to parse", err)
@@ -103,7 +103,7 @@ func TestParseSingleRoute(t *testing.T) {
 }
 
 func TestParsingSpecialChars(t *testing.T) {
-	r, err := parse(`Path("newlines") -> inlineContent("Line \ \1\nLine 2\r\nLine 3\a\b\f\n\r\t\v") -> <shunt>`)
+	r, err := parseDocument(`Path("newlines") -> inlineContent("Line \ \1\nLine 2\r\nLine 3\a\b\f\n\r\t\v") -> <shunt>`)
 
 	if err != nil {
 		t.Error("failed to parse", err)
@@ -122,7 +122,7 @@ func TestParsingSpecialChars(t *testing.T) {
 }
 
 func TestParseSingleRouteDef(t *testing.T) {
-	r, err := parse(singleRouteDefExample)
+	r, err := parseDocument(singleRouteDefExample)
 
 	if err != nil {
 		t.Error("failed to parse", err)
@@ -144,14 +144,14 @@ func TestParseInvalidDocument(t *testing.T) {
         route0: Method("GET") -> "https://backend-0.example.com"
         route1: Method("POST") -> "https://backend-1.example.com"`
 
-	_, err := parse(missingSemicolon)
+	_, err := parseDocument(missingSemicolon)
 	if err == nil {
 		t.Error("failed to fail")
 	}
 }
 
 func TestParseDocument(t *testing.T) {
-	r, err := parse(routingDocumentExample)
+	r, err := parseDocument(routingDocumentExample)
 
 	if err != nil {
 		t.Error("failed to parse document", err)
@@ -188,21 +188,21 @@ func TestParseDocument(t *testing.T) {
 }
 
 func TestNumberNotClosedWithDecimalSign(t *testing.T) {
-	_, err := parse(`* -> number(3.) -> <shunt>`)
+	_, err := parseDocument(`* -> number(3.) -> <shunt>`)
 	if err == nil {
 		t.Error("failed to fail")
 	}
 }
 
 func TestNumberStartingWithDecimal(t *testing.T) {
-	_, err := parse(`* -> number(.3) -> <shunt>`)
+	_, err := parseDocument(`* -> number(.3) -> <shunt>`)
 	if err != nil {
 		t.Error("failed to parse number", err)
 	}
 }
 
 func TestNumber(t *testing.T) {
-	_, err := parse(`* -> number(3.14) -> <shunt>`)
+	_, err := parseDocument(`* -> number(3.14) -> <shunt>`)
 	if err != nil {
 		t.Error("failed to parse number", err)
 	}
@@ -221,7 +221,7 @@ func TestRegExp(t *testing.T) {
 }
 
 func testRegExpOnce(t *testing.T, regexpStr string, expectedRegExp string) {
-	routes, err := parse(regexpStr)
+	routes, err := parseDocument(regexpStr)
 	if err != nil {
 		t.Error("failed to parse PathRegexp:"+regexpStr, err)
 		return