escape unicode as escaped bytes (\x)

1 files changed, 124 insertions, 0 deletions

diff --git a/js/unicode.js b/js/unicode.js
index cbab781..700434d 100644
--- a/js/unicode.js
+++ b/js/unicode.js
@@ -305,6 +305,129 @@ var unicode = (function(){
     }
     return groups
   }
+  
+  // encodes unicode characters as escaped utf16 - \xFFFF
+  // encodes ONLY non-ascii characters
+  function escapeToUtf16 (txt) {
+    var escaped_txt = "", kode
+    for (var i = 0; i < txt.length; i++) {
+      kode = txt.charCodeAt(i)
+      if (kode > 0x7f) {
+        kode = kode.toString(16)
+        switch (kode.length) {
+          case 2:
+            kode = "0" + kode
+          case 3:
+            kode = "0" + kode
+        }
+        escaped_txt += "\\u" + kode
+      }
+      else {
+        escaped_txt += txt[i]
+      }
+    }
+    return escaped_txt
+  }
+
+  // encodes unicode characters as escaped bytes - \xFF
+  // encodes ONLY non-ascii characters
+  function escapeToEscapedBytes (txt) {
+    var escaped_txt = "", kode, utf8_bytes
+    for (var i = 0; i < txt.length; i++) {
+      kode = txt.charCodeAt(i)
+      if (kode > 0x7f) {
+        utf8_bytes = convertUnicodeCodePointToUtf8Bytes(kode)
+        escaped_txt += convertBytesToEscapedString(utf8_bytes, 16)
+      }
+      else {
+        escaped_txt += txt[i]
+      }
+    }
+    return escaped_txt
+  }
+
+  // encodes unicode characters as escaped bytes - \xFF
+  // encodes an ENTIRE string
+  function escapeAllToEscapedBytes(str, base) {
+    var unicode_codes = convertStringToUnicodeCodePoints(str);
+    var data_bytes = convertUnicodeCodePointsToBytes(unicode_codes);
+    return convertBytesToEscapedString(data_bytes, 16);
+  }
+  // [ 0xE3, 0x81, 0x82, 0xE3, 0x81, 0x84 ] => '\xE3\x81\x82\xE3\x81\x84'
+  // [ 0343, 0201, 0202, 0343, 0201, 0204 ] => '\343\201\202\343\201\204'
+  function convertBytesToEscapedString(data_bytes, base) {
+    var escaped = '';
+    for (var i = 0; i < data_bytes.length; ++i) {
+      var prefix = (base == 16 ? "\\x" : "\\");
+      var num_digits = base == 16 ? 2 : 3;
+      var escaped_byte = prefix + formatNumber(data_bytes[i], base, num_digits)
+      escaped += escaped_byte;
+    }
+    return escaped;
+  }
+  // [ 0x3042, 0x3044 ] => [ 0xE3, 0x81, 0x82, 0xE3, 0x81, 0x84 ]
+  function convertUnicodeCodePointsToBytes(unicode_codes) {
+    var utf8_bytes = [];
+    for (var i = 0; i < unicode_codes.length; ++i) {
+      var bytes = convertUnicodeCodePointToUtf8Bytes(unicode_codes[i]);
+      utf8_bytes = utf8_bytes.concat(bytes);
+    }
+    return utf8_bytes;
+  }
+  // 0x3042 => [ 0xE3, 0x81, 0x82 ]
+  function convertUnicodeCodePointToUtf8Bytes(unicode_code) {
+    var utf8_bytes = [];
+    if (unicode_code < 0x80) {  // 1-byte
+      utf8_bytes.push(unicode_code);
+    } else if (unicode_code < (1 << 11)) {  // 2-byte
+      utf8_bytes.push((unicode_code >>> 6) | 0xC0);
+      utf8_bytes.push((unicode_code & 0x3F) | 0x80);
+    } else if (unicode_code < (1 << 16)) {  // 3-byte
+      utf8_bytes.push((unicode_code >>> 12) | 0xE0);
+      utf8_bytes.push(((unicode_code >> 6) & 0x3f) | 0x80);
+      utf8_bytes.push((unicode_code & 0x3F) | 0x80);
+    } else if (unicode_code < (1 << 21)) {  // 4-byte
+      utf8_bytes.push((unicode_code >>> 18) | 0xF0);
+      utf8_bytes.push(((unicode_code >> 12) & 0x3F) | 0x80);
+      utf8_bytes.push(((unicode_code >> 6) & 0x3F) | 0x80);
+      utf8_bytes.push((unicode_code & 0x3F) | 0x80);
+    }
+    return utf8_bytes;
+  }
+  // "あい" => [ 0x3042,  0x3044 ]
+  function convertStringToUnicodeCodePoints(str) {
+    var surrogate_1st = 0;
+    var unicode_codes = [];
+    for (var i = 0; i < str.length; ++i) {
+      var utf16_code = str.charCodeAt(i);
+      if (surrogate_1st != 0) {
+        if (utf16_code >= 0xDC00 && utf16_code <= 0xDFFF) {
+          var surrogate_2nd = utf16_code;
+          var unicode_code = (surrogate_1st - 0xD800) * (1 << 10) + (1 << 16) +
+                             (surrogate_2nd - 0xDC00);
+          unicode_codes.push(unicode_code);
+        } else {
+          // Malformed surrogate pair ignored.
+        }
+        surrogate_1st = 0;
+      } else if (utf16_code >= 0xD800 && utf16_code <= 0xDBFF) {
+        surrogate_1st = utf16_code;
+      } else {
+        unicode_codes.push(utf16_code);
+      }
+    }
+    return unicode_codes;
+  }
+  // 0xff => "ff"
+  // 0xff => "377"
+  function formatNumber(number, base, num_digits) {
+    var str = number.toString(base).toUpperCase();
+    for (var i = str.length; i < num_digits; ++i) {
+      str = "0" + str;
+    }
+    return str;
+  }
+
   return {
     raw: UNICODE_BLOCK_LIST,
     lookup: UNICODE_LOOKUP,
@@ -313,5 +436,6 @@ var unicode = (function(){
     block: block,
     findGroups: findGroups,
     paginate: paginate,
+    escapeToEscapedBytes: escapeToEscapedBytes,
   }
 })()