inlined unicode decoding

gjson.go

@@ -2,9 +2,10 @@
 package gjson
 
 import (
-	"encoding/json"
 	"reflect"
 	"strconv"
+	"unicode/utf16"
+	"unicode/utf8"
 	"unsafe"
 
 	"github.com/tidwall/match"
@@ -1358,14 +1359,68 @@ func GetBytes(json []byte, path string) Result {
 	return result
 }
 
-// unescape unescapes a string, we'll use the standard go/json library for this.
-func unescape(s string) string {
-	data := make([]byte, 0, len(s)+2)
-	data = append(data, '"')
-	data = append(data, s...)
-	data = append(data, '"')
-	json.Unmarshal(data, &s)
-	return s
+// runeit returns the rune from the the \uXXXX
+func runeit(json string) rune {
+	n, _ := strconv.ParseUint(json[:4], 16, 64)
+	return rune(n)
+}
+
+// unescape unescapes a string
+func unescape(json string) string { //, error) {
+	var str = make([]byte, 0, len(json))
+	for i := 0; i < len(json); i++ {
+		switch {
+		default:
+			str = append(str, json[i])
+		case json[i] < ' ':
+			return string(str)
+		case json[i] == '\\':
+			i++
+			if i >= len(json) {
+				return string(str)
+			}
+			switch json[i] {
+			default:
+				return string(str)
+			case '\\':
+				str = append(str, '\\')
+			case '/':
+				str = append(str, '/')
+			case 'b':
+				str = append(str, '\b')
+			case 'f':
+				str = append(str, '\f')
+			case 'n':
+				str = append(str, '\n')
+			case 'r':
+				str = append(str, '\r')
+			case 't':
+				str = append(str, '\t')
+			case '"':
+				str = append(str, '"')
+			case 'u':
+				if i+5 > len(json) {
+					return string(str)
+				}
+				r := runeit(json[i+1:])
+				i += 5
+				if utf16.IsSurrogate(r) {
+					// need another code
+					if len(json) >= 6 && json[i] == '\\' && json[i+1] == 'u' {
+						// we expect it to be correct so just consume it
+						r = utf16.DecodeRune(r, runeit(json[i+2:]))
+						i += 6
+					}
+				}
+				// provide enough space to encode the largest utf8 possible
+				str = append(str, 0, 0, 0, 0, 0, 0, 0, 0)
+				n := utf8.EncodeRune(str[len(str)-8:], r)
+				str = str[:len(str)-8+n]
+				i-- // backtrack index by one
+			}
+		}
+	}
+	return string(str)
 }
 
 // Less return true if a token is less than another token.