Converting byte position to character position

> And it works, but ow! what a mess. I could alternatively go through
> the string manually, counting up by byte range (<127 = add one, <0x800
> = add two, etc), but I suspect that'll work out even messier. Is there
> a cleaner way to do this?

Well. With UTF-8, no, not really. Unless you cache it somewhere.

This version is almost exactly as fast for short strings (like your
example) but significantly faster for long strings:

int utf8byte_to_char( string txt, int pos )
{
   int m = String.range(txt)[1];
   if( m < 128 )
     return pos;
   int c;
   if( m < 0x800 )
   {
      for( ; pos; c++,pos-- )
       if( txt[c] > 0x80 ) pos--;
   }
   else if( m < 0x10000 )
   {
      for( ; pos; c++,pos-- ) {
       if( txt[c] > 0x800 ) pos--;
       if( txt[c] > 0x80 ) pos--;
     }
   }
   else for( ; pos; c++,pos-- )
   {
     switch( txt[c] )
     {
	case 0x4000000..: pos--;
	case 0x200000..0x3ffffff: pos--;
	case 0x10000..0x1ffff: pos--;
	case 0x800..0xffff: pos--;
	case 0x80..0x7ff: pos--;
     }
   }
   return c;
}

Of course, your version can also be optimized significantly, at least
when the desired position is rather small and the string is long:

int utf8byte_to_char2( string txt int pos )
{
   return  sizeof(utf8_to_string(string_to_utf8(txt[..pos-1])[..pos-1]));
}

This is the fastest one I have tested yet, but please avoid looking
for byte 10000394 in a 2Mchar long unicode string... :)

-- 
Per Hedbor