header.go Referer() optimize (valyala#1313)

* args.go GetBool(): use switch with string casting

This should be optimized by Go compiler itself so the b2s() call is not needed.

It was previously done by this but changed in
1e7885e

* header.go Referer() optimize

Use direct peekArgBytes() instead of PeekBytes() that will check for special headers

* header_timing_test.go BenchmarkRequestHeaderPeekBytesSpecialHeader

The old BenchmarkRequestHeaderPeekBytesCanonical and BenchmarkRequestHeaderPeekBytesNonCanonical are in fact just measured the header normalization.
But it's anyway is benchmarked separately.
Results was almost the same: 1.5 ns/op.

Instead, let's reuse the benches to find a difference between peeking of special (Host, CT) and custom headers.

Author: stokito

Diff for: args.go

@@ -343,7 +343,7 @@ func (a *Args) GetUfloatOrZero(key string) float64 {
 // true is returned for "1", "t", "T", "true", "TRUE", "True", "y", "yes", "Y", "YES", "Yes",
 // otherwise false is returned.
 func (a *Args) GetBool(key string) bool {
-	switch b2s(a.Peek(key)) {
+	switch string(a.Peek(key)) {
 	// Support the same true cases as strconv.ParseBool
 	// See: https://github.com/golang/go/blob/4e1b11e2c9bdb0ddea1141eed487be1a626ff5be/src/strconv/atob.go#L12
 	// and Y and Yes versions.

Diff for: header.go

@@ -586,7 +586,7 @@ func (h *RequestHeader) SetUserAgentBytes(userAgent []byte) {
 
 // Referer returns Referer header value.
 func (h *RequestHeader) Referer() []byte {
-	return h.PeekBytes(strReferer)
+	return peekArgBytes(h.h, strReferer)
 }
 
 // SetReferer sets Referer header value.

Diff for: header_timing_test.go

@@ -12,6 +12,9 @@ import (
 
 var strFoobar = []byte("foobar.com")
 
+// it has the same length as Content-Type
+var strNonSpecialHeader = []byte("Dontent-Type")
+
 type benchReadBuf struct {
 	s []byte
 	n int
@@ -96,26 +99,55 @@ func BenchmarkResponseHeaderWrite(b *testing.B) {
 	})
 }
 
-func BenchmarkRequestHeaderPeekBytesCanonical(b *testing.B) {
+// Result: 2.2 ns/op
+func BenchmarkRequestHeaderPeekBytesSpecialHeader(b *testing.B) {
 	b.RunParallel(func(pb *testing.PB) {
 		var h RequestHeader
-		h.SetBytesV("Host", strFoobar)
+		h.SetContentTypeBytes(strFoobar)
 		for pb.Next() {
-			v := h.PeekBytes(strHost)
+			v := h.PeekBytes(strContentType)
 			if !bytes.Equal(v, strFoobar) {
 				b.Fatalf("unexpected result: %q. Expected %q", v, strFoobar)
 			}
 		}
 	})
 }
 
-func BenchmarkRequestHeaderPeekBytesNonCanonical(b *testing.B) {
+// Result: 2.9 ns/op
+func BenchmarkRequestHeaderPeekBytesNonSpecialHeader(b *testing.B) {
 	b.RunParallel(func(pb *testing.PB) {
 		var h RequestHeader
-		h.SetBytesV("Host", strFoobar)
-		hostBytes := []byte("HOST")
+		h.SetBytesKV(strNonSpecialHeader, strFoobar)
+		for pb.Next() {
+			v := h.PeekBytes(strNonSpecialHeader)
+			if !bytes.Equal(v, strFoobar) {
+				b.Fatalf("unexpected result: %q. Expected %q", v, strFoobar)
+			}
+		}
+	})
+}
+
+// Result: 2.3 ns/op
+func BenchmarkResponseHeaderPeekBytesSpecialHeader(b *testing.B) {
+	b.RunParallel(func(pb *testing.PB) {
+		var h ResponseHeader
+		h.SetContentTypeBytes(strFoobar)
+		for pb.Next() {
+			v := h.PeekBytes(strContentType)
+			if !bytes.Equal(v, strFoobar) {
+				b.Fatalf("unexpected result: %q. Expected %q", v, strFoobar)
+			}
+		}
+	})
+}
+
+// Result: 2.9 ns/op
+func BenchmarkResponseHeaderPeekBytesNonSpecialHeader(b *testing.B) {
+	b.RunParallel(func(pb *testing.PB) {
+		var h ResponseHeader
+		h.SetBytesKV(strNonSpecialHeader, strFoobar)
 		for pb.Next() {
-			v := h.PeekBytes(hostBytes)
+			v := h.PeekBytes(strNonSpecialHeader)
 			if !bytes.Equal(v, strFoobar) {
 				b.Fatalf("unexpected result: %q. Expected %q", v, strFoobar)
 			}