Merge pull request #657 from sylr/no-indexbyte

Optimize tag value parsing

tag_value.go

@@ -39,15 +39,36 @@ func (tv *TagValue) init(tag Tag, value []byte) {
 }
 
 func (tv *TagValue) parse(rawFieldBytes []byte) error {
-	sepIndex := bytes.IndexByte(rawFieldBytes, '=')
+	var sepIndex int
 
+	// Most of the Fix tags are 4 or less characters long, so we can optimize
+	// for that by checking the 5 first characters without looping over the
+	// whole byte slice.
+	if len(rawFieldBytes) >= 5 {
+		if rawFieldBytes[1] == '=' {
+			sepIndex = 1
+			goto PARSE
+		} else if rawFieldBytes[2] == '=' {
+			sepIndex = 2
+			goto PARSE
+		} else if rawFieldBytes[3] == '=' {
+			sepIndex = 3
+			goto PARSE
+		} else if rawFieldBytes[4] == '=' {
+			sepIndex = 4
+			goto PARSE
+		}
+	}
+
+	sepIndex = bytes.IndexByte(rawFieldBytes, '=')
 	switch sepIndex {
 	case -1:
 		return fmt.Errorf("tagValue.Parse: No '=' in '%s'", rawFieldBytes)
 	case 0:
 		return fmt.Errorf("tagValue.Parse: No tag in '%s'", rawFieldBytes)
 	}
 
+PARSE:
 	parsedTag, err := atoi(rawFieldBytes[:sepIndex])
 	if err != nil {
 		return fmt.Errorf("tagValue.Parse: %s", err.Error())