Added routines to handle conversion of Java strings between UTF8 and UCS2

representations.

ef/Runtime/System/JavaString.cpp

@@ -26,23 +26,73 @@ static inline JavaArray *newCharArray(Uint32 length)
     return (new (mem) JavaArray(Array::obtain(tkChar), length));
 }
 
+/* Count the number of bytes it would take to encode the given Unicode
+ * string using UTF-8.  Add in the extra byte for the terminating NUL.
+ */
+static int
+countUtf8Chars(const uint16 *ucs2, int ucs2len)
+{
+    int utf8len = 1;    // Need one character for terminating NUL
 
-/* Return the UTF representation of this string. This routine allocates 
+    for (int i = ucs2len-1; i >= 0; i--) {
+        uint16 u = ucs2[i];
+        if (u < 0x80)
+            utf8len += 1;
+        else if (u < 0x800)
+            utf8len += 2;
+        else
+            utf8len += 3;
+    }
+    return utf8len;
+}
+
+/* Convert a Unicode (UCS-2) string to UTF-8 encoding.  The length of
+ * the destination string, in bytes, is given by the utf8len argument.
+ * A NUL character is appended to the destination string, if possible.
+ * Returns: the actual length of the resulting string, in bytes.
+ */
+static int
+convertUnicodeToUtf8(char *utf8, const uint16* ucs2, int utf8len)
+{
+    char* start_utf8 = utf8;
+    char* lastchar = utf8 + utf8len - 1;
+
+    while (utf8 < lastchar) {
+        uint16 u = *ucs2++;
+        if (u < 0x80) {
+            *utf8++ = (char)u;
+        } else if (u < 0x800) {
+            if (utf8 >= (lastchar - 1))
+                break;
+            *utf8++ = 0xc0 | ((u >> 6) & 0x1f);
+            *utf8++ = 0x80 | (u & 0x3f);
+        } else {
+            if (utf8 >= (lastchar - 2))
+                break;
+            *utf8++ = 0xe0 | ((u >> 12) & 0x0f);
+            *utf8++ = 0x80 | ((u >> 6) & 0x3f);
+            *utf8++ = 0x80 | (u & 0x3f);
+        }
+    }
+    if (utf8 <= lastchar)
+        *utf8 = 0;
+
+    return utf8 - start_utf8;
+}
+
+/* Return the UTF8 representation of this string. This routine allocates 
  * enough memory for the conversion; this memory can be freed using 
  * JavaString::freeUtf()
  */
 char *JavaString::convertUtf()
 {
-  /* XXX Fixme For now, we just copy the string over byte by byte... */
-  const int16 *chars = getStr();
-  char *copy = new char[count+1];
+    const uint16 *chars = getStr();
+    int utf8len = countUtf8Chars(chars, count);
+    char *utf8 = new char[utf8len];
 
-  int32 i;
-  for (i = 0; i < count; i++)
-    copy[i] = (char) chars[i];
+    convertUnicodeToUtf8(utf8, chars, utf8len);
 
-  copy[i] = 0;
-  return copy;
+    return utf8;
 }
 
 void JavaString::freeUtf(char *str)
@@ -50,29 +100,96 @@ void JavaString::freeUtf(char *str)
   delete [] str;
 }
 
+/* Count the number of Unicode characters in a NUL-terminated
+ * UTF8 string.  Don't count the final NUL character.
+ */
+static int
+countUnicodeChars(const char *utf8)
+{
+    signed char c;
+    int length = 0;
+    
+    // Unicode characters are encoded as 1, 2, or 3 bytes in a UCS-2 string
+    while (c = *utf8) {
+        length++;
+
+        if (c >= 0) {
+            // Characters in the range of 0..0x7f are encoded using one byte
+            // b0xxxxxxx
+            utf8++;
+        } else if ((c & 0xe0) == 0xc0) {
+            // Characters in the range 0x80..0x7ff are encoded using two bytes
+            // b110xxxxx b10yyyyyy
+            utf8 += 2;
+        } else {
+            // Characters in the range 0x800..0xffff are encoded using three bytes
+            // b1110xxxx b10yyyyyy b10zzzzzz
+            PR_ASSERT((c & 0xf0) == 0xe0);
+            utf8 += 3;  
+        }
+    }
+    return length;
+}
+
+/* Convert a UTF-8 encoded string to Unicode (UCS-2) representation.  The
+ * length of the destination string, in 16-bit characters, is given by the
+ * ucs2 argument.  The result is *not* NUL-terminated.
+ * Returns: the actual length of the resulting string, in characters.
+ */
+static int
+convertUTF8ToUnicode(uint16 *ucs2, const char *utf8, int ucs2len)
+{
+    signed char c;
+    int length = 0;
+    
+    // Unicode characters are encoded as 1, 2, or 3 bytes in a UCS-2 string
+    while ((c = *utf8) != 0) {
+        length++;
+        if (length > ucs2len)
+            return ucs2len;
+
+        if (c >= 0) {
+            // Characters in the range of 0..0x7f are encoded using one byte
+            // b0xxxxxxx
+            *ucs2 = c;
+            utf8++;
+        } else if ((c & 0xe0) == 0xc0) {
+            // Characters in the range 0x80..0x7ff are encoded using two bytes
+            // b110xxxxx b10yyyyyy
+            *ucs2 = ((c & 0x1f) << 6) | (utf8[1] & 0x3f);
+            utf8 += 2;
+        } else {
+            // Characters in the range 0x800..0xffff are encoded using three bytes
+            // b1110xxxx b10yyyyyy b10zzzzzz
+            PR_ASSERT((c & 0xf0) == 0xe0);
+            *ucs2 = ((c & 0x0f) << 12) | ((utf8[1] & 0x3f) << 6) | (utf8[2] & 0x3f);
+            utf8 += 3;
+        }
+        ucs2++;
+    }
+    return length;
+}
+
+
 /* Create a new JavaString from a char array that represents the string in UTF-8
  * format.
  */
 JavaString::JavaString(const char *str) : JavaObject(*strType)
 {
-  count = PL_strlen(str);
+  count = countUnicodeChars(str);
 
   offset = 0;
 
-  /* Let's keep the string zero-terminated anyway */
-  value = (JavaArray *) newCharArray(count+1);
-  int16 *chars = const_cast<int16 *>(getStr());
+  value = (JavaArray *) newCharArray(count);
+  uint16 *chars = const_cast<uint16 *>(getStr());
 
-  for (int32 i = 0; i < count; i++)
-    chars[i] = str[i];
-
-  chars[count] = 0;
+  convertUTF8ToUnicode(chars, str, count);
 }
 
 /* print a textual representation of this string */
 void JavaString::dump()
 {
-  const int16 *chars = getStr();
+  const uint16 *chars = getStr();
 
   for (int16 i = 0; i < count; i++)
     putchar(chars[i]);