tests/utf8_processor.lua

   1 local lpeg = require("lpeg")
   2
   3 local string = string
   4
   5 module("utf8_processor")
   6
   7 local function encode_utf(codepoint)
   8         if codepoint > 0x10FFFF then
   9                 error("Codepoint > 10FFFF cannot be encoded")
  10         elseif codepoint > 0xFFFF then
  11                 -- Surrogate pair needed
  12                 codepoint = codepoint - 0x10000
  13                 local first, second = codepoint / 0x0400 + 0xD800, codepoint % 0x0400 + 0xDC00
  14                 return ("\\u%.4X\\u%.4X"):format(first, second)
  15         else
  16                 return ("\\u%.4X"):format(codepoint)
  17         end
  18 end
  19
  20 -- decode a two-byte UTF-8 sequence
  21 local function f2 (s)
  22         local c1, c2 = string.byte(s, 1, 2)
  23         return encode_utf(c1 * 64 + c2 - 12416)
  24 end
  25
  26 -- decode a three-byte UTF-8 sequence
  27 local function f3 (s)
  28         local c1, c2, c3 = string.byte(s, 1, 3)
  29         return encode_utf((c1 * 64 + c2) * 64 + c3 - 925824)
  30 end
  31
  32 -- decode a four-byte UTF-8 sequence
  33 local function f4 (s)
  34         local c1, c2, c3, c4 = string.byte(s, 1, 4)
  35         return encode_utf(((c1 * 64 + c2) * 64 + c3) * 64 + c4 - 63447168)
  36 end
  37
  38 local cont = lpeg.R("\128\191")   -- continuation byte
  39
  40 local utf8 = lpeg.R("\0\127") -- Do nothing here
  41         + lpeg.R("\194\223") * cont / f2
  42         + lpeg.R("\224\239") * cont * cont / f3
  43         + lpeg.R("\240\244") * cont * cont * cont / f4
  44
  45 local utf8_decode_pattern = lpeg.Cs(utf8^0) * -1
  46
  47
  48 function process(s)
  49         return utf8_decode_pattern:match(s)
  50 end