lib/json: Improved performance of Unicode-escaping in a `Text`

Signed-off-by: Lucas Bajolet <r4pass@hotmail.com>

diff --git a/lib/core/text/abstract_text.nit b/lib/core/text/abstract_text.nit
index 1de1aa1..0b07edb 100644 (file)
--- a/lib/core/text/abstract_text.nit
+++ b/lib/core/text/abstract_text.nit
@@ -743,21 +743,32 @@ abstract class Text
        #     assert "\\ud800\\udfd3".from_utf16_escape == '𐏓'
        #     assert "\\u00e8".from_utf16_escape == 'è'
        #     assert "\\u3042".from_utf16_escape == 'あ'
-       fun from_utf16_escape: Char do
-               var ln = length
-               if ln != 6 and ln != 12 then return 0xFFFD.code_point
-               var cphi = substring(2, 4).to_hex
-               if cphi < 0xD800 then return cphi.code_point
-               if cphi > 0xDFFF then return cphi.code_point
-               if cphi > 0xDBFF then return 0xFFFD.code_point
-               var cp = 0
-               cp += (cphi - 0xD800) << 10
-               var cplo = substring(8, 4).to_hex
+       fun from_utf16_escape(pos, ln: nullable Int): Char do
+               if pos == null then pos = 0
+               if ln == null then ln = length - pos
+               if ln < 6 then return 0xFFFD.code_point
+               var cp = from_utf16_digit(pos + 2)
+               if cp < 0xD800 then return cp.code_point
+               if cp > 0xDFFF then return cp.code_point
+               if cp > 0xDBFF then return 0xFFFD.code_point
+               if ln == 6 then return 0xFFFD.code_point
+               if ln < 12 then return 0xFFFD.code_point
+               cp <<= 16
+               cp += from_utf16_digit(pos + 8)
+               var cplo = cp & 0xFFFF
                if cplo < 0xDC00 then return 0xFFFD.code_point
                if cplo > 0xDFFF then return 0xFFFD.code_point
-               cp += cplo - 0xDC00
-               cp += 0x10000
-               return cp.code_point
+               return cp.from_utf16_surr.code_point
+       end
+
+       # Returns a UTF-16 escape value
+       #
+       #     var s = "\\ud800\\udfd3"
+       #     assert s.from_utf16_digit(2) == 0xD800
+       #     assert s.from_utf16_digit(8) == 0xDFD3
+       fun from_utf16_digit(pos: nullable Int): Int do
+               if pos == null then pos = 0
+               return to_hex(pos, 4)
        end
 
        # Encode `self` to percent (or URL) encoding

diff --git a/lib/json/static.nit b/lib/json/static.nit
index 15a4355..7c0e2cf 100644 (file)
--- a/lib/json/static.nit
+++ b/lib/json/static.nit
@@ -138,7 +138,8 @@ redef class Text
        protected fun json_to_nit_string: String do
                var res = new FlatBuffer.with_capacity(bytelen)
                var i = 0
-               while i < self.length do
+               var ln = self.length
+               while i < ln do
                        var char = self[i]
                        if char == '\\' then
                                i += 1
@@ -154,21 +155,19 @@ redef class Text
                                else if char == 't' then
                                        char = '\t'
                                else if char == 'u' then
-                                       var code = substring(i + 1, 4)
-                                       var hx = code.to_hex
-                                       if hx >= 0xD800 and hx <= 0xDFFF then
-                                               var lostr = substring(i + 7, 4)
-                                               if lostr.length < 4 then
-                                                       hx = 0xFFFD
+                                       var u16_esc = from_utf16_digit(i + 1)
+                                       char = u16_esc.code_point
+                                       if char.is_surrogate and i + 10 < ln then
+                                               if self[i + 5] == '\\' and self[i + 6] == 'u' then
+                                                       u16_esc <<= 16
+                                                       u16_esc += from_utf16_digit(i + 7)
+                                                       char = u16_esc.from_utf16_surr.code_point
+                                                       i += 6
                                                else
-                                                       hx <<= 16
-                                                       hx += lostr.to_hex
-                                                       hx = hx.from_utf16_surr
+                                                       char = 0xFFFD.code_point
                                                end
-                                               i += 6
                                        end
                                        i += 4
-                                       char = hx.code_point
                                end
                                # `"`, `/` or `\` => Keep `char` as-is.
                        end